1: 

2: \magnification=\magstep1

3: 

4: \input amstex

5: \documentstyle{myamsppt}

6: 

7: % Comment out EITHER the next 3 lines (PicTeX) ...

8: 

9: %\input pictex

10: %\setcoordinatesystem units <1in,1in>

11: %\linethickness 0.7pt

12: 

13: % ... OR the next 4 lines (Eplain).

14: 

15: %\let\noarrow=0    % comment out if using "\commdiag"

16: \input eplain

17: \input ruled

18: \let\ref=\eplainref  %see Eplain document (v2.6) page 4.

19: 

20: \TagsOnRight

21: \parskip 1ex

22: 

23: \hsize=6.5truein

24: \vsize=9truein

25: 

26: \loadbold

27: \loadeurm

28: 

29: \font\bbbf=cmbx12 scaled\magstep1

30: \font\bbf=cmbx12

31: \font\bbigbf=cmbx12 scaled\magstep1

32: \font\bi=cmbxti10	% bold italic

33: \font\bigbf=cmbx12

34: \font\bigbi=cmbxti10 scaled\magstep1	%big bold italic

35: \font\bigit=cmti12

36: \font\bigrm=cmr12 

37: \font\sc=cmcsc10

38: \font\smallrm=cmr9

39: \font\smallit=cmti9

40: \font\smallbf=cmbx9

41: \font\smallbi=cmbxti10 at 8pt

42: \font\sf=cmss10

43: \font\script=eurm10

44: \font\bigsf=cmss12

45: \font\bigtt=cmtt12

46: \font\smalltt=cmtt8

47: \font\ssf=cmss8 

48: \font\ssmallrm=cmr8

49: \font\smallsl=cmsl8

50: \document

51: 

52: \define\de{\define}

53: \de\AAA{{\tx{${\scr A}$}}}

54: \de\AAa{{\tx{${\Bbb A}$}}}

55: \de\ABBB{{\tx{$A^\BBB$}}}

56: \de\ABN{{\tx{$A^{B,N}$}}}

57: \de\AN{{\tx{$A^N$}}}

58: \de\Ab{{\bf A}}

59: \de\Abi{{\bi A}}

60: \de\Abar{{\tx{$\bar A$}}}

61: \de\Abars{\tx{$\Abar_s$}}

62: \de\Abaru{\tx{$\Abar^u$}}

63: \de\AinKK{\tx{$A \in \KK$}}

64: \de\AinAlgSig{\tx{$A \in \AlgSig$}}

65: \de\AinNStdAlgSig{\tx{$A \in \NStdAlgSig$}}

66: \de\AinStdAlgSig{\tx{$A \in \StdAlgSig$}}

67: \de\Akvec{\tx{$A[\kvec]$}}

68: \de\Akxvec{\tx{$A[\kxvec]$}}

69: \de\Aklxvec{\tx{$A[\kvec,\lxvec]$}}

70: \de\Alg{\tx{\bi Alg}}

71: \de\AlgSig{\Alg\,(\Sig)}

72: \de\Alvec{\tx{$A[\lvec\,]$}}

73: \de\Alxvec{\tx{$A[\lxvec]$}}

74: \de\Aps{\tx{\sf Ap}}

75: \de\As{\tx{$A_s$}}

76: \de\Asbar{\tx{$\Abar_s$}}

77: \de\Asx{\tx{$A_s^*$}}

78: \de\AssLang{\tx{\bi Ass}\Lang}

79: \de\Assumption{\pr{Assumption}\sl}

80: \de\Assumptionn#1{\pr{Assumption #1}\sl}

81: \de\Assumptions{\pr{Assumptions}\sl}

82: \de\Asuu{\tx{$A^\uu_s$}}

83: \de\At{\tx{\bi At}}

84: \de\AtSig{\At(\Sig)}

85: \de\AtSt{\tx{\bi AtSt}}

86: \de\AtStSig{\tx{$\AtSt(\Sig)$}}

87: \de\Atil{{\tx{$\til A$}}}

88: \de\Au{\tx{$A^u$}}

89: \de\Aul{{\tx{$\ul A$}}}

90: \de\Auu{\tx{$A^\uu$}}

91: \de\AuuN{\tx{$A^{\uu,N}$}}

92: \de\Av{\tx{$A^v$}}

93: \de\Aw{\tx{$A^w$}}

94: \de\Ax{{\tx{$A^*$}}}

95: \de\Axs{\tx{$A^*_s$}}

96: \de\Axx{(\Ax)\str}

97: \de\Bbi{\tx{\bi B}}

98: \de\BB{{\tx{${\Bbb B}$}}}

99: \de\BBB{{\tx{${\scr B}$}}}

100: \de\BBbar{\tx{$\ol \BB$}}

101: \de\BBuu{\tx{${\Bbb B}^\uu$}}

102: \de\Bb{\tx{\bf B}}

103: \de\Bool{\tx{\bi Bool}}

104: \de\BoolSig{\tx{$\Bool(\Sig)$}}

105: \de\Bs{\tx{\sf B}}

106: \de\Bss{{\tx{\ssf B}}}

107: \de\Btil{\tx{$\til B$}}

108: \de\Bul{\tx{$\ul B$}}

109: \de\CC{\tx{${\Bbb C}$}}

110: \de\CCn{\tx{$\CC^n$}}

111: \de\CCC{\tx{$\scr C$}}

112: \de\CCCL{\tx{${\CCC}^<$}}

113: \de\CR{\tx{\bi CR}}

114: \de\CRSig{\tx{$\CR(\Sig)$}}

115: \de\CT{{\tx{\bi CT}}}

116: \de\CTSig{\CT(\Sig)}

117: \de\CTsmall{{\tx{\smallbi CT}}}

118: \de\Cbar{\tx{$\ol C$}}

119: \de\Cf{{\it Cf\.}}

120: \de\Chead#1{\bn\cbb{#1}\bn}

121: \de\Comp{\tx{\bi Comp}}

122: \de\CompA{\tx{$\Comp^A$}}

123: \de\CompLengthA{\tx{$\Comp\Length^A$}}

124: \de\CompSeqA#1{\tx{$\Comp\Seq^A(#1)$}}

125: \de\Compu{\tx{\bi Compu}}

126: \de\Con{\tx{\bi Con}}

127: \de\Cond{\tx{\bi Cond}}

128: \de\CondSig{\tx{$\Cond(\Sig)$}}

129: \de\Constrn#1{\pr{Construction \ #1}\rm}

130: \de\Cont{\tx{\bi Cont}}

131: \de\ContA{\tx{$\Cont(A)$}}

132: \de\Conventionn#1{\pr{Convention \ #1}\sl}

133: \de\Cor{\pr{Corollary}\sl}

134: \de\Corn#1{\prn{Corollary #1}\sl}

135: \de\Cors{\pr{Corollaries}\sl}

136: \de\Cs{\tx{\sf C}}

137: \de\Cul{\tx{$\ul C$}}

138: \de\DD{{\tx{${\Bbb D}$}}}

139: \de\DDD{{\tx{$\scr D$}}}

140: \de\Dbar{\tx{$\ol{D}$}}

141: \de\Dbars{\tx{$\ol{D}_s$}}

142: \de\Def{\pr{Definition}\rm}

143: \de\Defn#1{\prn{Definition #1}\rm}

144: \de\Defs{\pr{Definitions}\rm}

145: \de\Defsn#1{\prn{Definitions #1}\rm}

146: \de\Del{\tx{$\Delta$}}

147: \de\Discussion{\pr{Discussion}\rm}

148: \de\Discussionn#1{\pr{Discussion #1}\rm}

149: \de\Ds{\tx{\sf D}}

150: \de\Dss{{\tx{\ssf D}}}

151: \de\Ebar{\tx{$\ol E$}}

152: \de\Eg{{\it E.g.}}

153: \de\ElemInd{\tx{\sf ElemInd}}

154: \de\ElemIndSig{\tx{$\ElemInd(\Sig)$}}

155: \de\EqSort{\tx{\bi EqSort\/}}

156: \de\EqSortSig{\EqSort(\Sig)}

157: \de\Eqb{\tx{\bi Eq}}

158: \de\Eqs{\tx{\sf Eq}}

159: \de\Equiv{\tx{$\pmb{Equiv}$}}

160: \de\Etil{\tx{$\til{E}$}}

161: \de\Example{\pr{Example}}

162: \de\Examplen#1{\prn{Example {#1}}}

163: \de\Examples{\pr{Examples}}

164: \de\Examplesn#1{\prn{Examples {#1}}}

165: \de\Exercise{\pr{Exercise}}

166: \de\Exercisen#1{\prn{Exercise {#1}}}

167: \de\Exercises{\pr{Exercises}}

168: \de\Exercisesn#1{\prn{Exercises {#1}}}

169: \de\F{\tx{\bi F}}

170: \de\FA{\tx{$F^A$}}

171: \de\FAuu{\tx{$F^{A,\uu}$}}

172: \de\FFF{\tx{$\scr F$}}

173: \de\FFt{\tx{\tt F}}

174: \de\FinFuncSig{\tx{$F \in \FuncSig$}}

175: \de\First{\tx{\bi First}}

176: \de\For{\tx{\bi For}}

177: \de\ForA{\tx{$\For(A)$}}

178: \de\ForN{\tx{$\For^N$}}

179: \de\ForNA{\tx{$\ForN(A)$}}

180: \de\ForNSig{\tx{$\ForN(\Sig)$}}

181: \de\ForSig{\tx{$\For(\Sig)$}}

182: \de\ForSigN{\tx{$\For(\SigN)$}}

183: \de\ForSigx{\tx{$\For(\Sigx)$}}

184: \de\Form{\tx{\bi Form}}

185: \de\FormSig{\tx{$\Form(\Sig)$}}

186: \de\Foro{\tx{$\For_0$}}

187: \de\Forolam{\tx{$\For_0^{\lam}$}}

188: \de\Forx{\tx{$\For^\starb$}}

189: \de\ForxA{\tx{$\Forx(A)$}}

190: \de\ForxSig{\tx{$\Forx(\Sig)$}}

191: \de\Fs{{\tx{\smallbi F}}}

192: \de\Ftil{\tx{$\til{F}$}}

193: \de\Func{\tx{\bi Func}}

194: \de\FuncSig{\Func\,(\Sig)}

195: \de\FuncSigp{\tx{$\Func(\Sigp)$}}

196: \de\FuncSigus{\tx{$\FuncSig_\utos$}}

197: \de\GG{\tx{${\Bbb G}$}}

198: \de\GGG{\tx{$\scr G$}}

199: \de\Gtt{\tx{\tt G}}

200: \de\Gam{\tx{$\Gamma$}}

201: \de\Gbar{\tx{$\ol G$}}

202: \de\Gs{{\tx{\sf G}}}

203: \de\Halt{\tx{\bi Halt}}

204: \de\HaltTests{\tx{\sf HaltTest}}

205: \de\HaltA#1{\tx{$\Halt^A\,(#1)$}}

206: \de\Hbar{\tx{$\bar H$}}

207: \de\Hh{\tx{\bi H}}

208: \de\Hs{{\tx{\sf H}}}

209: \de\Hss{{\tx{\ssf H}}}

210: \de\Htt{\tx{\tt H}}

211: \de\Ib{\tx{{\bf I}}}

212: \de\II{{\tx{${\Bbb I}$}}}

213: \de\III{\tx{${\scr I}$}}

214: \de\IIIN{{\tx{$\III^N$}}}

215: \de\IL{\tx{\bi IL}}

216: \de\IO{\tx{\bi IO}}

217: \de\Ibar{{\tx{$\bar I$}}}

218: \de\Ie{{\it I.e.}}

219: \de\Ind{\tx{\sf Ind}}

220: \de\IndSig{\tx{$\Ind(\Sig)$}}

221: \de\Indi{\tx{$\Ind_i$}}

222: \de\IndiSig{\Indi(\Sig)}

223: \de\Init{\tx{\bi Init}}

224: \de\Is{\tx{\sf I}}

225: \de\Iss{{\tx{\ssf I}}}

226: \de\JJJ{\tx{${\scr J}$}}

227: \de\Jbar{{\tx{$\bar J$}}}

228: \de\KK{{\tx{${\Bbb K}$}}}

229: \de\KKbar{\tx{$\ol{\KK}$}}

230: \de\KKN{\tx{${\Bbb K}^N$}}

231: \de\KKuu{\tx{${\Bbb K}^\uu$}}

232: \de\KKx{\tx{${\Bbb K}^*$}}

233: \de\LL{\tx{${\Bbb L}$}}

234: \de\LLL{\tx{${\scr L}$}}

235: \de\Lam{\tx{$\Lambda$}}

236: \de\Lang{\tx{\bi Lang}}

237: \de\LangSig{\tx{$\Lang(\Sig)$}}

238: \de\LangSigx{\tx{$\Lang(\Sigx)$}}

239: \de\Langx{\tx{$\Lang^*$}}

240: \de\LangxSig{\tx{$\Langx(\Sig)$}}

241: \de\Lemma{\pr{Lemma}\sl}

242: \de\Lemman#1{\prn{Lemma #1}\sl}

243: \de\Lemmas{\pr{Lemmas}\sl}

244: \de\Lemmasn#1{\prn{Lemmas #1}\sl}

245: \de\Length{\tx{\bi Length}}

246: \de\Lgths{\tx{\sf Lgth}}

247: \de\Lss{{\tx{\ssf L}}}

248: \de\MMM{\tx{${\scr M}$}}

249: \de\Min{\tx{\bi Min}}

250: \de\MinNStdAlg{\Min\NStdAlg}

251: \de\Mod{\tx{\bi Mod}}

252: \de\NN{{\tx{${\Bbb N}$}}}

253: \de\NNN{{\tx{${\scr N}$}}}

254: \de\NNNB{{\tx{${\NNN^B}$}}}

255: \de\NNNo{{\tx{$\NNN_0$}}}

256: \de\NNu{\tx{${\Bbb N}^\uu$}}

257: \de\NStdAlg{\tx{\bi NStdAlg\/}}

258: \de\NStdAlgSig{\tx{$\NStdAlg\,(\Sig)$}}

259: \de\Nbd{\tx{\bi Nbd}}

260: \de\Nbi{{\tx{\bi N}}}

261: \de\Nbismall{{\tx{\smallbi N}}}

262: \de\Newlengths{\tx{\sf Newlength}}

263: \de\Notation{\pr{Notation}\rm}

264: \de\Notationn#1{\pr{Notation \ #1}\rm}

265: \de\Note{\pr{Note}\rm}

266: \de\Noten#1{\pr{Note #1}\rm}

267: \de\Notes{\pr{Notes}\rm}

268: \de\Notesn#1{\pr{Notes #1}\rm}

269: \de\Ns{\tx{\sf N}}

270: \de\Nss{{\tx{\ssf N}}}

271: \de\Nnu{\tx{$\Nn^\uu$}}

272: \de\Nulls{\tx{\sf Null}}

273: \de\Nullvec{\tx{$\vec{\Nulls}$}}

274: \de\Om{{\tx{$\Omega$}}}

275: \de\Pp{\tx{\bk P}}

276: \de\PA{\tx{$P^A$}}

277: \de\PE{\tx{\bi PE}}

278: \de\PL#1{\tx{$\ProgLang_{#1}$}}

279: \de\PPP{\tx{$\scr P$}}

280: \de\PR{\tx{PR}}

281: \de\PRA{\tx{$\PR(A)$}}

282: \de\PRAutos{\tx{$\PRA_\utos$}}

283: \de\PRSig{\tx{$\PR(\Sig)$}}

284: \de\PRSigb{\tx{$\PRb(\Sigb)$}}

285: \de\PRSigutos{\tx{$\PRSig_\utos$}}

286: \de\PRb{\tx{\bf PR}}

287: \de\PRmin{\tx{$\PR_-$}}

288: \de\PRo{\tx{$\PR_0$}}

289: \de\PRolam{\tx{$\PR_0^{\lam}$}}

290: \de\PRutos{\tx{$\PR_\utos$}}

291: \de\PRx{\PR\str}

292: \de\PRxA{\tx{$\PRx(A)$}}

293: \de\PRxSig{\tx{\PRx(\Sig)}}

294: \de\PRxSigb{\tx{\PRxb(\Sigb)}}

295: \de\PRxb{\PRb\strb}

296: \de\PTE{\tx{\bi PTE}}

297: \de\PTEA{\tx{$\PTE^A$}}

298: \de\PTerm{\tx{\bi PTerm}}

299: \de\PTermSig{\tx{$\PTerm(\Sig)$}}

300: \de\Pbi{\tx{\bi P}}

301: \de\Pf{\n{\bf Proof:\ \,}}

302: \de\Ph{\tx{$\Phi$}}

303: \de\PreStr{\tx{\bi PreStr}}

304: \de\PreStrSig{\PreStr\,(\Sig)}

305: \de\Problem{\pr{Problem}}

306: \de\Problemn#1{\pr{Problem {#1}}}

307: \de\Problems{\pr{Problems}}

308: \de\Problemsn#1{\pr{Problems {#1}}}

309: \de\Propsn#1{\pr{Propositions {#1}}}

310: \de\Proc{\tx{\bi Proc}}

311: \de\ProcN{\tx{$\Proc^N$}}

312: \de\ProcNSig{\ProcN(\Sig)}

313: \de\ProcSig{\Proc(\Sig)}

314: \de\ProcSigN{\Proc(\SigN)}

315: \de\ProcSiguv{\tx{$\Proc(\Sig)_\utov$}}

316: \de\ProcSigx{\tx{$\Proc(\Sigx)$}}

317: \de\Procuv{\tx{$\Proc_\utov$}}

318: \de\ProcuvSig{\tx{$\Procuv(\Sig)$}}

319: \de\Procx{\Proc\str}

320: \de\ProcxSig{\Procx(\Sig)}

321: \de\ProcxSiguv{\tx{$\ProcxSig_\utov$}}

322: \de\Procxuv{\tx{$\Proc^*_\utov$}}

323: \de\Prod{\tx{\bi Prod}}

324: \de\ProdType{\tx{\bi ProdType}}

325: \de\ProdTypeSig{\tx{$\ProdType(\Sig)$}}

326: \de\Prog{\tx{\bi Prog}}

327: \de\ProgLang{\Prog\Lang}

328: \de\ProgTerm{\Prog\Term}

329: \de\ProgTermSig{\tx{$\ProgTerm(\Sig)$}}

330: \de\Prop{\pr{Proposition}\sl}

331: \de\Propn#1{\prn{Proposition #1}\sl}

332: \de\Propp{\tx{\bi Prop}}

333: \de\Props{\pr{Propositions}\sl}

334: \de\Ps{\tx{$\Psi$}}

335: \de\Q{{\bf Q}}

336: \de\QQ{\tx{${\Bbb Q}$}}

337: \de\Qs{\tx{\sf Q}}

338: \de\Question{\pr{Question}\sl}

339: \de\RR{{\tx{${\Bbb R}$}}}

340: \de\RRR{{\tx{$\scr R$}}}

341: \de\RRRB{\tx{$\RRR^B$}}

342: \de\RRRL{\tx{${\RRR}^<$}}

343: \de\RRRN{\tx{$\RRR^N$}}

344: \de\RRn{\tx{$\RR^n$}}

345: \de\RRx{\tx{$\RR\str$}}

346: \de\RRq{\tx{$\RR^q$}}

347: \de\Ref{\tx{\bf Ref:\/}}

348: \de\Rem{\tx{\bi Rem}}

349: \de\RemA{\tx{$\Rem^A$}}

350: \de\RemSeqA#1{\tx{$\Rem\Seq^A(#1)$}}

351: \de\RemSet#1{\tx{\bi RemSet\,$(#1)$}}

352: \de\Remark{\pr{Remark}\rm}

353: \de\Remarkn#1{\prn{Remark #1}\rm}

354: \de\Remarks{\pr{Remarks}\rm}

355: \de\Remarksn#1{\prn{Remarks #1}\rm}

356: \de\Rep{\tx{\bi Rep}}

357: \de\Rest{\tx{\bi Rest}}

358: \de\Rs{\tx{\sf R}}

359: \de\Rss{{\tx{\ssf R}}}

360: \de\SA{\tx{$\bb{S}^A$}}

361: \de\SS{{\tx{${\Bbb S}$}}}

362: \de\SSS{{\tx{${\scr S}$}}}

363: \de\Sarrow{\tx{$S^{\to}$}}

364: \de\Sat{\tx{$S_{\tx{\ssf at}}$}}

365: \de\Sbar{\tx{$\ol \Sbi$}}

366: \de\Sbi{\tx{\bi S}}

367: \de\Seq{\tx{\bi Seq}}

368: \de\Shead#1#2{\bn{\bigbf {#1}\ \ \ {#2}}\sn}

369: \de\Sheads#1#2{\bn{\bigbf {#1}\ \ \ {#2}}}

370: \de\Sig{{\tx{$\varSigma$}}}

371: \de\SigA{\tx{$\Sig(A)$}}

372: \de\SigBBB{\tx{$\Sig^\BBB$}}

373: \de\SigBN{\tx{$\Sig^{B,N}$}}

374: \de\SigN{\tx{$\Sig^N$}}

375: \de\SigT{\tx{$(\Sig,T)$}}

376: \de\Sigb{\tx{$\bs\varSigma$}}

377: \de\Sigbar{\tx{$\ol \Sig$}}

378: \de\Sigo{\tx{$\varSigma_0$}}

379: \de\Sigox{\tx{$\varSigma_0^*$}}

380: \de\Sigone{\tx{$\Sigb_1$}}

381: \de\SigoneInd{\tx{$\Sigb_{\pmb1}$-$\pmb{Ind}\,$}}

382: \de\Sigonex{\tx{$\Sigb_1^*$}}

383: \de\SigonexInd{\tx{$\Sigb_{\pmb1}^{\pmb*}$-$\pmb{Ind}\,$}}

384: \de\SigonexIndl{\tx{$\Sigb_{\pmb1}^{\pmb*}$-$\pmb{Ind_\ell}\,$}}

385: \de\SigonexIndc{\tx{$\Sigb_{\pmb1}^{\pmb*}$-$\pmb{Ind_c}\,$}}

386: \de\SigonexIndi{\tx{$\Sigb_{\pmb1}^{\pmb*}$-$\pmb{Ind_i}\,$}}

387: \de\Sigp{{\tx{$\Sig'$}}}

388: \de\SigpTp{\tx{$(\Sig',T')$}}

389: \de\Sigpp{{\tx{$\Sig''$}}}

390: \de\SigppTpp{\tx{$(\Sig'',T'')$}}

391: \de\Sigstk{\tx{$\Sig^\stkss$}}

392: \de\Siguu{\tx{$\varSigma^\uu$}}

393: \de\SiguuN{\tx{$\varSigma^{\uu,N}$}}

394: \de\Sigx{\tx{$\varSigma^*$}}

395: \de\Snap{\tx{\bi Snap}}

396: \de\SnapA{\tx{$\Snap^A$}}

397: \de\SnapSeqA#1{\tx{$\Snap\Seq^A(#1)$}}

398: \de\Sort{\tx{\bi Sort\/}}

399: \de\SortSig{\Sort(\Sig)}

400: \de\SortSigsmall{\Sortsmall\,(\Sig)}

401: \de\SortSigp{\tx{$\Sort(\Sigp)$}}

402: \de\SortSigx{\tx{$\Sort(\Sigx)$}}

403: \de\Sortsmall{\tx{\smallbi Sort}}

404: \de\Spec{\tx{\sf Spec}}

405: \de\Ss{\tx{\sf S}}

406: \de\Sss{{\tx{\ssf S}}}

407: \de\State{\tx{\bi State}}

408: \de\StateA{\tx{$\State(A)$}}

409: \de\Step{\tx{\bi Step}}

410: \de\Stmt{\tx{\bi Stmt}}

411: \de\StmtSig{\tx{$\Stmt(\Sig)$}}

412: \de\StmtStepA{\tx{$\Stmt\Step^A$}}

413: \de\Std{\tx{\bi Std\/}}

414: \de\StdAlg{\Std\Alg}

415: \de\StdAlgSig{\tx{$\StdAlg\,(\Sig)$}}

416: \de\StdAlgSigx{\tx{$\StdAlg\,(\Sigx)$}}

417: \de\StdMod{\Std\Mod}

418: \de\Str{\tx{\bi Str}}

419: \de\StrSig{\Str\,(\Sig)}

420: \de\SubalgA#1{\tx{${\bi \ Subalg}_A(#1)$}}

421: \de\Tb{\tx{\bf T}}

422: \de\Tbi{\tx{\bi T}}

423: \de\TComp{\tx{\bi TComp}}

424: \de\TE{\tx{\bi TE}}

425: \de\TEA{\tx{$\TE^A$}}

426: \de\TOL{\TagsOnLeft}

427: \de\TOR{\TagsOnRight}

428: \de\TSig{\Tbi(\Sig)}

429: \de\TT{\tx{${\Bbb T}$}}

430: \de\TTT{\tx{$\scr T$}}

431: \de\Tab{\tx{\bi Tab}}

432: \de\TabSig{\Tab(\Sig)}

433: \de\Tbar{\tx{$\ol T$}}

434: \de\Term{\tx{\bi Term}}

435: \de\TermSig{\tx{$\Term(\Sig)$}}

436: \de\TermTup{\Term\Tup}

437: \de\TermTupSig{\tx{$\TermTup(\Sig)$}}

438: \de\Terminology{\pr{Terminology}\rm}

439: \de\Th{\tx{\bi Th}}

440: \de\Thesisn#1{\pr{Thesis \ #1}\sl}

441: \de\Thm{\pr{Theorem}\sl}

442: \de\Thmn#1{\prn{Theorem #1}\sl}

443: \de\Thms{\pr{Theorems}\sl}

444: \de\Tm{\tx{\bi Tm}}

445: \de\TmSig{\tx{$\Tm(\Sig)$}}

446: \de\Tms{\tx{$\Tm_s$}}

447: \de\TmsSig{\tx{$\Tms(\Sig)$}}

448: \de\Ts{\tx{$\Tbi_s$}}

449: \de\TsSig{\tx{$\Ts(\Sig)$}}

450: \de\Tup{\tx{\bi Tup}}

451: \de\Ub{\tx{\bf U}}

452: \de\Ubi{\tx{\bi U}}

453: \de\Univ{\tx{\bi Univ}}

454: \de\Univs{\tx{\sf Univ}}

455: \de\Univuv{\tx{$\Univs_\utov$}}

456: \de\UnivuvA{\tx{$\Univs_\utov^A$}}

457: \de\Univuvx{\tx{$\Univs_\utov^*$}}

458: \de\UnivuvxA{\tx{$\Univs_\utov^{*,A}$}}

459: \de\Univus{\tx{$\Univ_\utos$}}

460: \de\UnivusA{\tx{$\Univ_\utos^A$}}

461: \de\Unspecs{\tx{\sf Unspec}}

462: \de\Updates{\tx{\sf Update}}

463: \de\UpdateD{\tx{$\Updates^D$}}

464: \de\Us{\tx{\sf U}}

465: \de\VVV{\tx{$\scr V$}}

466: \de\Val{\tx{\bi Val}}

467: \de\Var{\tx{\bi Var}}

468: \de\VarSig{\tx{$\Var(\Sig)$}}

469: \de\VarTup{\Var\Tup}

470: \de\VarTupSig{\tx{$\VarTup(\Sig)$}}

471: \de\Vars{\tx{$\Var_s$}}

472: \de\VarsSig{\tx{$\Vars(\Sig)$}}

473: \de\Vbb{\tx{\bf V}}

474: \de\WWW{\tx{${\scr W}$}}

475: \de\Wb{\tx{\bf W}}

476: \de\While{\tx{\bi While}}

477: \de\WhileA{\tx{$\While(A)$}}

478: \de\WhileN{\tx{$\While^\Nbismall$}}

479: \de\WhileNA{\tx{$\WhileN(A)$}}

480: \de\WhileNSig{\tx{$\WhileN(\Sig)$}}

481: \de\WhileSig{\tx{$\While(\Sig)$}}

482: \de\WhileSigN{\tx{$\While(\SigN)$}}

483: \de\WhileSigx{\tx{$\While(\Sigx)$}}

484: \de\Whilebig{\tx{\bigbi While}}

485: \de\WhilebigSig{\tx{$\Whilebig(\Sigb)$}}

486: \de\WhilebigN{\tx{$\Whilebig^\Nbi$}}

487: \de\Whilebigx{\tx{$\Whilebig^\starb$}}

488: \de\Whilex{\tx{$\While^\starb$}}

489: \de\WhilexA{\Whilex(A)}

490: \de\WhilexSig{\Whilex(\Sig)}

491: \de\WhilexSigN{\Whilex(\SigN)}

492: \de\Whilexbig{\tx{$\Whilebig^{\dz{\starb}}$}}

493: \de\XXX{{\tx{${\scr X}$}}}

494: \de\ZZ{\tx{${\Bbb Z}$}}

495: \de\ZZZ{{\tx{${\scr Z}$}}}

496: \de\Zbi{\tx{\bi Z}}

497: \de\Zs{\tx{\sf Z}}

498: \de\Zss{{\tx{\ssf Z}}}

499: \de\Zz{\tx{\sf Z}}

500: \de\abi{\tx{\bi a}}

501: \de\aas{\tx{\sf a}}

502: \de\abar{\tx{$\bar a$}}

503: \de\adt{{\bi adt}}

504: \de\aee{\tx{\bi ae}}

505: \de\ainAu{\tx{$a \in \Au$}}

506: \de\ainAv{\tx{$a \in \Av$}}

507: \de\ainAs{\tx{$a \in \As$}}

508: \de\al{{\tx{$\alpha$}}}

509: \de\alA{\tx{$\alpha^A$}}

510: \de\alb{\tx{$\bs\alpha$}}

511: \de\albar{{\tx{$\ol \alpha$}}}

512: \de\algebras{\tx{\sf algebra}}

513: \de\alhat{{\tx{$\hat{\alpha}$}}}

514: \de\aliass{\tx{\sf alias}}

515: \de\aliasss{{\tx{\ssf alias}}}

516: \de\all{\forall}

517: \de\alvec{\tx{$\vec{\alpha}$}}

518: \de\ands{\tx{\sf and}}

519: \de\ang#1{\tx{$\langle #1 \rangle$}}

520: \de\angg#1{\tx{$\langg #1 \rangg$}}

521: \de\arity#1{\tx{\rm arity$(#1)$}}

522: \de\arr{\arrow <0.1in> [0.2,0.5]}

523: \de\att{{\tx{\tt a}}}

524: \de\atts{{\tx{\smalltt a}}}

525: \de\attx{\att\str}

526: \de\auxs{\tx{\sf aux}}

527: \de\avec{\tx{$\vec{a}$}}

528: \de\ax{\tx{$a^*$}}

529: \de\bA{\tx{$\bb{b}^A$}}

530: \de\bB{\bk{B}}

531: \de\bFor{\tx{\bigbi For}}

532: \de\bForN{\tx{$\bFor^\Nbi$}}

533: \de\bForx{\tx{$\bFor^\starb$}}

534: \de\bProc{\tx{\bigbi Proc}}

535: \de\bb#1{\tx{$\lbb #1 \rbb$}}

536: \de\bbar{[\ns]}

537: \de\bbi{\tx{\bi b}}

538: \de\bbs{\tx{\sf b}}

539: \de\bdot{{\bs\cdot}}

540: \de\be{{\tx{$\beta$}}}

541: \de\beA{\tx{$\beta^A$}}

542: \de\beb{\tx{$\bs\beta$}}

543: \de\begins{\tx{\sf begin}}

544: \de\bigcaplim#1{\underset#1\to\bigcap}

545: \de\bigcon{\bigwedge}

546: \de\bigdis{\bigvee}

547: \de\bigskipn{\bigskip\nin}

548: \de\biu{\tx{\bi u}}

549: \de\bk{\boldkey}

550: \de\bn{\bigskip\nin}

551: \de\bools{\tx{\sf bool}}

552: \de\boolss{{\tx{\ssf bool}}}

553: \de\br{\ |\ }

554: \de\bs{\boldsymbol}

555: \de\btt{{\tx{\tt b}}}

556: \de\btts{{\tx{\smalltt b}}}

557: \de\bttx{\btt\str}

558: \de\bu{$\bullet$}

559: \de\bul{\item{\bu}}

560: \de\bull{\itemm{\bu}}

561: \de\bulll{\itemmm{\bu}}

562: \de\bx{\tx{$b^*$}}

563: \de\capt#1#2{\ce{{\sc #1}. \ \ #2}}

564: \de\card#1{\tx{\bi card\,$(#1)$}}

565: \de\carrierss{\tx{\sf carriers}}

566: \de\cart#1{\tx{\bi cart\, $(#1)$}}

567: \de\cb#1{\ce{\bf #1}}

568: \de\cbb#1{\ce{\bbf #1}}

569: \de\cbbb#1{\ce{\bbigbf #1}}

570: \de\cbi#1{\ce{\bi #1}}

571: \de\cbbi#1{\ce{\bigbi #1}}

572: \de\ccc{{\tx{\bi c}}}

573: \de\ce{\centerline}

574: \de\cf{{\it cf\.}}

575: \de\chead#1{\bn\cb{#1}\mn}

576: \de\chooses{\tx{\sf choose}}

577: \de\cnr#1{\tx{$\ulc #1 \urc$}}

578: \de\comlabs{\tx{\com\bf-\labs}}

579: \de\comlam{\tx{$\com^\lam$}}

580: \de\compb{\tx{\bi comp}}

581: \de\compl#1{\tx{\bi compl\,$(#1)$}}

582: \de\complexs{\tx{\sf complex}}

583: \de\complexss{{\tx{\ssf complex}}}

584: \de\comps{\tx{\sf comp}}

585: \de\con{\land}

586: \de\concon{{\bigcon\nss\nss\nss\bigcon}}

587: \de\conconlim#1{\underset#1\to\concon}

588: \de\conconlims#1#2{\overset{#2}\to{\conconlim{#1}}}

589: \de\cons{\tx{\sf con}}

590: \de\conss{{\tx{\ssf con}}}

591: \de\constantss{\tx{\sf constants}}

592: \de\consts{\tx{\sf const}}

593: \de\constss{{\tx{\ssf const}}}

594: \de\cpxs{\tx{\sf complex}}

595: \de\ccs{\tx{\sf c}}

596: \de\css{{\tx{\ssf c}}}

597: \de\ctt{{\tx{\tt c}}}

598: \de\ctts{{\tx{\smalltt c}}}

599: \de\cttx{\ctt\str}

600: \de\curl#1{\tx{$\{#1\}$}}

601: \de\curly#1{\tx{$\{\,#1\,\}$}}

602: \de\cvals{\tx{\sf cval}}

603: \de\cvec{\tx{$\vec{c}$}}

604: \de\cx{\tx{$c^*$}}

605: \de\da{\downarrow}

606: \de\dash{\item{---}}

607: \de\dashh{\itemm{---}}

608: \de\dashhh{\itemmm{---}}

609: \de\datas{{\tx{\sf data}}}

610: \de\datass{{\tx{\ssf data}}}

611: \de\dcs{\tx{\sf dc}}

612: \de\ddis{\lor\!\!\!\!\!\lor}

613: \de\defaults{\tx{\sf default}}

614: \de\degg#1{\tx{\bi deg$(#1)$}}

615: \de\del{{\tx{$\delta$}}}

616: \de\delb{\tx{$\bs\delta$}}

617: \de\delhat{\hat{\delta}}

618: \de\depth#1{\tx{\bi depth$(#1)$}}

619: \de\dis{\lor}

620: \de\disdis{{\bigdis\nss\nss\nss\bigdis}}

621: \de\disdislim#1{\underset#1\to\disdis}

622: \de\disdislims#1#2{\overset{#2}\to{\disdislim{#1}}}

623: \de\displayquote{\narrower\narrower\smallrm\nin}

624: \de\displaytext{\narrower\narrower\sl\nin}

625: \de\diss{\tx{\sf dis}}

626: \de\divN{\tx{$\divs_{\natss}$}}

627: \de\divNR{\tx{$\divs_\natss^\RRR$}}

628: \de\divs{\tx{\sf div}}

629: \de\dos{\tx{\sf do}}

630: \de\dom#1{\tx{\bi dom$(#1)$}}

631: \de\down{\tx{$^\vee$}}

632: \de\ds{{\tx{\sf d}}}

633: \de\dt{\tx{\bi dt}}

634: \de\dtt{{\tx{\tt d}}}

635: \de\dz{\dsize}

636: \de\ebar{{\tx{$\bar e$}}}

637: \de\eg{{\it e.g.}}

638: \de\ehat{tx{$\hat e$}}

639: \de\el{\tx{$\ell$}}

640: \de\elses{\tx{\sf else}}

641: \de\endpf{\qed\smskip}

642: \de\endpr{\rm\smskip}

643: \de\ends{\tx{\sf end}}

644: \de\enum{\tx{\bi enum}}

645: \de\eps{{\tx{$\epsilon$}}}

646: \de\epsb{{\tx{$\bs\epsilon$}}}

647: \de\eqbool{\tx{$\eqs_{\boolss}$}}

648: \de\eqdata{\tx{$\eqs_\datass$}}

649: \de\eqdf{=_{df}}

650: \de\eqint{\tx{$\eqs_{\intss}$}}

651: \de\eqintvl{\tx{$\eqs_{\intvlss}$}}

652: \de\eqnat{\tx{$\eqs_{\natss}$}}

653: \de\eqnatN{\tx{$\eqs_{\natss}^\NNN$}}

654: \de\eqreal{\tx{$\eqs_{\realss}$}}

655: \de\eqrealR{\tx{$\eqs_{\realss}^\RRR$}}

656: \de\eqs{\tx{\sf eq}}

657: \de\eqss{\tx{$\eqs_s$}}

658: \de\equal{\ &= \ }

659: \de\es{\tx{\sf e}}

660: \de\etal{{\it et al.}}

661: \de\ett{{\tx{\tt e}}}

662: \de\eval{\tx{\bi eval}}

663: \de\evals{\tx{\sf eval}}

664: \de\evec{\tx{$\vec{e}$}}

665: \de\ex{\exists}

666: \de\expands{\tx{\sf expand}}

667: \de\fA{\tx{$f^A$}}

668: \de\falses{\tx{\sf false}}

669: \de\fbi{\tx{\bi f}}

670: \de\fff{\tx{{\sf f}\!{\sf f}}}

671: \de\ffrom{\tx{$\Leftarrow$}}

672: \de\fcheck{\tx{$\check f$}}

673: \de\fhat{\tx{$\hat f$}}

674: \de\first{\tx{\bi first}}

675: \de\fis{\tx{\sf fi}}

676: \de\floor#1{\tx{$\llc #1 \lrc$}}

677: \de\fn{\footnote}

678: \de\fors{\tx{\sf for}}

679: \de\free{\tx{\bi free}}

680: \de\from{\leftarrow}

681: \de\fs{{\tx{\sf f}}}

682: \de\fss{{\tx{\ssf f}}}

683: \de\ftil{\tx{$\til f$}}

684: \de\funcs{\tx{\sf func}}

685: \de\functionss{\tx{\sf functions}}

686: \de\fuu{\tx{$f^\uu$}}

687: \de\fvec{\tx{$\vec{f}$}}

688: \de\gam{{\tx{$\gamma$}}}

689: \de\gamA{\tx{$\gam^A$}}

690: \de\gamb{\tx{$\bs\gamma$}}

691: \de\gamx{\tx{$\gam^*$}}

692: \de\ghat{\tx{$\hat g$}}

693: \de\gtil{\tx{$\til g$}}

694: \de\gn{\tx{\bi gn}}

695: \de\gotos{\tx{\sf goto}}

696: \de\graph{\tx{\bi graph}}

697: \de\gs{{\tx{\sf g}}}

698: \de\gtt{{\tx{\tt g}}}

699: \de\gvec{\tx{$\vec{g}$}}

700: \de\halt{\tx{\bi halt}}

701: \de\hs{{\tx{\sf h}}}

702: \de\htil{\tx{$\til h$}}

703: \de\htt{{\tx{\tt h}}}

704: \de\ibar{{\tx{$\bar{\imath}$}}}

705: \de\id{\tx{\bi id}}

706: \de\ident{\equiv}

707: \de\idi{\tx{$\id_i$}}

708: \de\ie{{\it i.e.}}

709: \de\ifbool{\tx{$\ifs_{\boolss}$}}

710: \de\ifdata{\tx{$\ifs_{\datass}$}}

711: \de\ifff{\ \,\llongtofrom\ }

712: \de\ifffdf{\ \ \tx{$\llongtofrom_{df}$}\ }

713: \de\ifint{\tx{$\ifs_{\intss}$}}

714: \de\ifintvl{\tx{$\ifs_{\intvlss}$}}

715: \de\ifnat{\tx{$\ifs_{\natss}$}}

716: \de\ifnatN{\tx{$\ifs_{\natss}^\NNN$}}

717: \de\ifreal{\tx{$\ifs_{\realss}$}}

718: \de\ifrealR{\tx{$\ifs_\realss^\RRR$}}

719: \de\ifs{\tx{\sf if}}

720: \de\ifss{\tx{$\ifs_s$}}

721: \de\ifstk{\tx{$\ifs_{\stkss}$}}

722: \de\ift#1{&\tx{if \ {#1}}}

723: \de\ii{{\tx{\bi i}}}

724: \de\iii{{\tx{\sf i}}}

725: \de\iinI{{\tx{$i \in I$}}}

726: \de\ims{\tx{\sf im}}

727: \de\imp{\to}

728: \de\impp{\ \,\llongto\ }

729: \de\imps{\tx{\sf imp}}

730: \de\imports{\tx{\sf import}}

731: \de\inbull{\itemitem{\bu}}

732: \de\indash{\itemitem{---}}

733: \de\indentt{\indent\quad}

734: \de\ins{\tx{\sf in}}

735: \de\ints{{\tx{\sf int}}}

736: \de\intss{{\tx{\ssf int}}}

737: \de\intvls{{\tx{\sf intvl}}}

738: \de\intvlss{{\tx{\ssf intvl}}}

739: \de\io{{\tx{$\iota$}}}

740: \de\is{\tx{\sf i}}

741: \de\isom{\cong}

742: \de\itt{{\tx{\tt i}}}

743: %\de\itemm{\itemitem}

744: \de\ivec{\tx{$\vec{\imath}$}}

745: \de\ivecstrut{\vec{\mathstrut i}}

746: \de\jbar{{\tx{$\bar{\jmath}$}}}

747: \de\js{\tx{\sf j}}

748: \de\jvec{\tx{$\vec{\jmath}$}}

749: \de\jvecstrut{\vec{\mathstrut j}}

750: \de\kap{{\tx{$\kappa$}}}

751: \de\kapb{\tx{$\bs\kappa$}}

752: \de\kbar{\tx{$\bar k$}}

753: \de\kinNN{\tx{$k\in\NN$}}

754: \de\ktt{{\tx{\tt k}}}

755: \de\kvec{\tx{$\vec{k}$}}

756: \de\kvecstrut{\vec{\mathstrut k}}

757: \de\kxvec{\tx{$\kvec^*$}}

758: \de\labs{\lamb\tx{\bi abs}}

759: \de\lam{{\tx{$\lambda$}}}

760: \de\lamPR{\lam\PR}

761: \de\lamPRmin{\lam\PRmin}

762: \de\lamPRx{\lam\PRx}

763: \de\lamb{\tx{$\bs\lambda$}}

764: \de\lammuPR{\lam\muu\PR}

765: \de\lammuPRmin{\lam\muu\PRmin}

766: \de\lammuPRx{\lam\muu\PRx}

767: \de\langg{\langle\!|}

768: \de\lb{\linebreak}

769: \de\lbar{\tx{$\ol l$}}

770: \de\lbb{[\![}

771: \de\leasts{\tx{\sf least}}

772: \de\lele#1#2#3{{\tx{$#1 \le #2 \le #3$}}}

773: \de\lel#1#2#3{{\tx{$#1 \le #2 < #3$}}}

774: \de\lgth#1{\tx{\bi lgth$(#1)$}}

775: \de\llc{\llcorner}

776: \de\lle#1#2#3{{\tx{$#1 < #2 \le #3$}}}

777: \de\llongfrom{\tx{$\Longleftarrow$}}

778: \de\llongto{\tx{$\Longrightarrow$}}

779: \de\llongtofrom{\tx{$\Longleftrightarrow$}}

780: \de\loccit{{\it loc\. cit.}}

781: \de\locs{\tx{\sf loc}}

782: \de\longfrom{\longleftarrow}

783: \de\longimp{\longrightarrow}

784: \de\longto{\longrightarrow}

785: \de\longtofrom{\longleftrightarrow}

786: \de\loops{\tx{\sf loop}}

787: \de\lowerbox#1#2{{\lower#1pt\hbox{$#2$}}}

788: \de\lrc{\lrcorner}

789: \de\lss{\tx{\sf less}}

790: \de\lsint{\tx{$\lss_\intss$}}

791: \de\lsintvl{\tx{$\lss_\intvlss$}}

792: \de\lsnat{\tx{$\lss_\natss$}}

793: \de\lsnatN{\tx{$\lss_\natss^\NNN$}}

794: \de\lsreal{\tx{$\lss_\realss$}}

795: \de\lsrealR{\tx{$\lss_\realss^\RRR$}}

796: \de\lvec{\tx{$\vec{l}$}}

797: \de\lxvec{\tx{$\vec{l}^*$}}

798: \de\lxx{\lower.2ex}

799: \de\lxxx{\lower.3ex}

800: \de\lxxxx{\lower.4ex}

801: \de\macros{\tx{\sf macro}}

802: \de\macross{{\tx{\ssf macro}}}

803: \de\magn#1{\magnification=\magstep#1}

804: \de\mapdownl#1{\mapdown\lft{\dz{#1}}}

805: \de\mapdownlr#1#2{\mapdown\lft{\dz{#1}}\rt{\dz{#2}}}

806: \de\mapdownr#1{\mapdown\rt{\dz{#1}}}

807: \de\maprightd#1{\mapright_{\dz{#1}}}

808: \de\maprightdlower#1#2{\mapright_{\lower#1pt\hbox{$\dz{#2}$}}}

809: \de\maprightu#1{\mapright^{\dz{#1}}}

810: \de\maprightud#1#2{\mapright^{\dz{#1}}_{\dz{#2}}}

811: \de\maprighturaise#1#2{\mapright^{\raise#1pt\hbox{$\dz{#2}$}}}

812: \de\maprighturaised#1#2#3{\mapright^{\raise#1pt\hbox{$\dz{#2}$}}_{\dz{#3}}}

813: \de\maprightudlower#1#2#3{\mapright^{\dz{#1}}_{\lower#2pt\hbox{$\dz{#3}$}}}

814: \de\mapupl#1{\mapup\lft{\dz{#1}}}

815: \de\mapuplr#1#2{\mapup\lft{\dz{#1}}\rt{\dz{#2}}}

816: \de\mapupr#1{\mapup\rt{\dz{#1}}}

817: %\de\mapupud#1#2{\mapup^{\dz{#1}}_{\dz{#2}}}

818: \de\mbar{\tx{$\bar m$}}

819: \de\medskipn{\medskip\nin}

820: \de\mg#1{\tx{\bi mg$(#1)$}}

821: \de\mi{\tx{$^-$}}

822: \de\mins{\tx{\sf min}}

823: \de\mn{\medskip\nin}

824: \de\mtt{{\tx{\tt m}}}

825: \de\muCR{\mub\CR}

826: \de\muCRA{\tx{$\muCR(A)$}}

827: \de\muCRSig{\tx{$\muCR(\Sig)$}}

828: \de\muPR{\muu\PR}

829: \de\muPRA{\tx{$\muPR(A)$}}

830: \de\muPRSig{\tx{$\muPR(\Sig)$}}

831: \de\muPRSigb{\tx{$\muPRb(\Sigb)$}}

832: \de\muPRb{\mub\PRb}

833: \de\muPRmin{\muu\PRmin}

834: \de\muPRo{\muu\PRo}

835: \de\muPRolam{\muu\PRolam}

836: \de\muPRx{\muPR\str}

837: \de\muPRxb{\mub\PRxb}

838: \de\muPRxA{\tx{$\muPRx(A)$}}

839: \de\muPRxSig{\tx{$\muPRx(\Sig)$}}

840: \de\muPRxSigutos{\tx{$\muPRxSig_\utos$}}

841: \de\muPRxSigb{\tx{$\muPRxb(\Sigb)$}}

842: \de\muPRxutos{\tx{$\muPRx_\utos$}}

843: \de\mub{\tx{$\bs\mu$}}

844: \de\muu{{\tx{$\mu$}}}

845: \de\mvec{\tx{$\vec{m}$}}

846: \de\mxvec{\tx{$\mvec^*$}}

847: \de\n{\noindent}

848: \de\names{\tx{\sf name}}

849: \de\namess{{\tx{\ssf name}}}

850: \de\nats{\tx{\sf nat}}

851: \de\natss{{\tx{\ssf nat}}}

852: \de\nbar{\tx{$\bar n$}}

853: \de\negs{\tx{\sf neg}}

854: \de\news{\tx{\sf new}}

855: \de\newss{{\tx{\ssf new}}}

856: \de\nil{\tx{$\emptyset$}}

857: \de\nin{\noindent}

858: \de\ninNN{\tx{$n\in\NN$}}

859: \de\nl{\newline}

860: \de\nneg{\neg\!\!\!\!\!\neg}

861: \de\notover{\tx{\bi notover}}

862: \de\notimplies{\tx{$\ \not\nsss\llongto$}}

863: \de\notimplied{\tx{$\ \not\nsss\llongfrom$}}

864: \de\nots{\tx{\sf not}}

865: \de\npn{\NoPageNumbers}

866: \de\ns{\negthinspace}

867: \de\nss{\negthickspace}

868: \de\nsss{\nss\nss}

869: \de\nssss{\nsss\nsss}

870: \de\ntt{{\tx{\tt n}}}

871: \de\nub{\tx{$\bs\nu$}}

872: \de\nx{\tx{$n^*$}}

873: \de\ol{\overline}

874: \de\opcit{{\it op\. cit\.}}

875: \de\ods{\tx{\sf od}}

876: \de\om{{\tx{$\omega$}}}

877: \de\omb{{\tx{$\bs\omega$}}}

878: \de\onto{\twoheadrightarrow}

879: \de\ors{\tx{\sf or}}

880: \de\outs{\tx{\sf out}}

881: \de\ow{&\tx{otherwise}}

882: \de\pa{{\it p.a.}}

883: \de\pairs{\tx{\sf pair}}

884: \de\pe{\tx{\bi pe}}

885: \de\ph{{\tx{$\varphi$}}}

886: \de\pie{\tx{$\pi$}}

887: \de\pmapsto{{\,\overset\ {\cdot} \to\mapsto\,}}

888: \de\pr#1{\mn{\bf #1}. \,}

889: \de\prb#1{\bn{\bbf #1}.\sn}

890: \de\prc#1{\sn{\sc #1}.\ \ }

891: \de\prn#1{\mn{\bf #1}.\,}

892: \de\prims{\tx{\sf prim}}

893: \de\procs{\tx{\sf proc}}

894: \de\procss{{\tx{\ssf proc}}}

895: \de\proj{\tx{\bi proj}}

896: \de\projs{\tx{\sf proj}}

897: \de\ps{{\tx{$\psi$}}}

898: \de\pte{\tx{\bi pte}}

899: \de\pto{{\,\overset\ {\cdot} \to\longto\,}}

900: \de\ptt{{\tx{\tt p}}}

901: \de\pttx{\ptt\str}

902: \de\px{\tx{$p^*$}}

903: \de\qqquad{\qquad \qquad}

904: \de\qqqquad{\qqquad \qqquad}

905: \de\qfors{`\,\fors'}

906: \de\qforxs{`$\,\fors^*$'}

907: \de\qtt{{\tx{\tt q}}}

908: \de\qwhiles{`\,\whiles'}

909: \de\qwhilex{`$\,\whiles^*$'}

910: \de\rR{\tx{$\bk R$}}

911: \de\raisebox#1#2{{\raise#1pt\hbox{$#2$}}}

912: \de\ran#1{\tx{\bi ran$(#1)$}}

913: \de\rand{\ \tx{\rm and}\ }

914: \de\rangg{|\!\rangle}

915: \de\rats{\tx{\sf rat}}

916: \de\ratss{{\tx{\ssf rat}}}

917: \de\rbb{]\!]}

918: \de\reals{\tx{\sf real}}

919: \de\realss{{\tx{\ssf real}}}

920: \de\rec{\tx{\bi rec}}

921: \de\redSig{\,|\,_{\Sig}}

922: \de\reffs{\tx{\sf ref}}

923: \de\refss{{\tx{\ssf ref}}}

924: \de\rel#1{\tx{\rm (rel $#1$)}}

925: \de\rem{\tx{\bi rem}}

926: \de\res{\tx{\sf re}}

927: \de\rest{\restriction}

928: \de\restt{\tx{\bi rest}}

929: \de\rh{\tx{$\rho$}}

930: \de\rtt{{\tx{\tt r}}}

931: \de\rvec{\tx{$\vec{r}$}}

932: \de\rxx{\raise.2ex}

933: \de\rxxx{\raise.3ex}

934: \de\rxxxx{\raise.4ex}

935: \de\sbar{{\tx{$\bar s$}}}

936: \de\scr{\Cal}

937: %\de\scr{\fam\eusbfam}

938: \de\se{\tx{\bi se}}

939: \de\seq{\tx{\bi seq}}

940: \de\seqt{\longmapsto}

941: \de\setss{{\tx{\ssf set}}}

942: \de\shead#1#2{\mn{\bf {#1}\ \ \ {#2}}\sn}

943: \de\sheadb#1#2{\mn{\bf {#1}\ \ \ {#2}}\nl}

944: \de\sheadrun#1#2{\mn{\bf {#1}\ \ \ {#2}.\ \ }}

945: \de\sheads#1#2{\mn{\bf {#1}\ \ \ {#2}}}

946: \de\sig{{\tx{$\sigma$}}}

947: \de\signatures{\tx{\sf signature}}

948: \de\sigo{\tx{$\sig_0$}}

949: \de\sigp{\tx{$\sig'$}}

950: \de\sinSortSig{{\tx{$s \in \SortSig$}}}

951: \de\sinSortSigx{\tx{$s \in \SortSigx$}}

952: \de\skips{\tx{\sf skip}}

953: \de\smskip{\smallskip}

954: \de\smskipn{\smallskip\nin}

955: \de\sn{\smallskip\nin}

956: \de\snap{\tx{\bi snap}}

957: \de\sort#1{\tx{\bi sort\,{$(#1)$}}}

958: \de\sortss{\tx{\sf sorts}}

959: \de\sq{\simeq}

960: \de\sqsseq{\sqsubseteq}

961: \de\srchs{\tx{\sf srch}}

962: \de\ssA{{\ssz A}}

963: \de\ssB{{\ssz B}}

964: \de\ssbe{{\ssz \beta}}

965: \de\sset{\subset}

966: \de\sseq{\subseteq}

967: \de\sshead#1#2{{\bf {#1}\ \ \ {#2}}\sn\!}

968: \de\ssheadrun#1#2{{\bf {#1}\ \ \ {#2}.\ \ }}

969: \de\ssheads#1#2{{\bf {#1}\ \ \ {#2}}}

970: \de\sss{\tx{\sf s}}

971: \de\sz{\ssize}

972: \de\ssz{\sssize}

973: \de\starb{{\tx{$\bk *$}}}

974: \de\state{\tx{\bi state}}

975: \de\stks{{\tx{\sf stk}}}

976: \de\stkss{{\tx{\ssf stk}}}

977: \de\str{\tx{$^*$}}

978: \de\stt{{\tx{\tt s}}}

979: \de\strb{\tx{$^{\bk*}$}}

980: \de\subex{\tx{\bi subex}}

981: \de\subst{\tx{\bi subst}}

982: \de\sucs{\tx{\sf suc}}

983: \de\suu{{\tx{$s^\uu$}}}

984: \de\sx{{\tx{$s^*$}}}

985: \de\tA{\tx{$\bb{t}^A$}}

986: \de\ta{\tx{$\tau$}}

987: \de\tagx{\tag{$*$}}

988: \de\tagxx{\tag{$**$}}

989: \de\tagxxx{\tag{${*}{*}{*}$}}

990: \de\tagxxxx{\tag{${*}{*}{*}{*}$}}

991: \de\tagxxxxx{\tag{${*}{*}{*}{*}{*}$}}

992: \de\tagxxxxxx{\tag{${*}{*}{*}{*}{*}{*}$}}

993: \de\tbar{\tx{$\bar t$}}

994: \de\te{\tx{\bi te}}

995: \de\term{\tx{\bi term}}

996: \de\that{\tx{$\hat t$}}

997: \de\thens{\tx{\sf then}}

998: \de\tif#1{&\tx{if \ {#1}}}

999: \de\til{\tilde}

1000: \de\tiltil#1{\tilde{\tilde#1}}

1001: \de\tinT{\tx{$t\in\TT$}}

1002: \de\th{\tx{$\theta$}}

1003: \de\tofrom{\leftrightarrow}

1004: \de\tos{\tx{\sf to}}

1005: \de\toto{\rightrightarrows}

1006: \de\trues{\tx{\sf true}}

1007: \de\tstile{\vdash}

1008: \de\tto{\tx{$\Rightarrow$}}

1009: \de\ttofrom{\tx{$\Leftrightarrow$}}

1010: \de\ttstile{\models}

1011: \de\ttt{\tx{{\sf t}\!{\sf t}}}

1012: \de\tup#1#2#3{\tx{$#1_{#2},\dots,#1_{#3}$}}

1013: \de\tuptimes#1#2#3{\tx{$#1_{#2} \times \dots \times #1_{#3}$}}

1014: \de\tvec{\tx{$\vec{t}$}}

1015: \de\tvecstrut{\vec{\mathstrut t}}

1016: \de\twiddle{{\tx{$\sim$}}}

1017: \de\twiddlek{\tx{$\sim_k$}}

1018: \de\tx{\text}

1019: \de\txvec{\tx{$\vec{t^*}$}}

1020: \de\type#1{\tx{\bi type$(#1)$}}

1021: \de\ua{{\tx{$\uparrow$}}}

1022: \de\ub#1{\tx{\underbar{#1}}}

1023: \de\ubi{{\tx{\bi u}}}

1024: \de\ubrivec{\tx{$u | \ivec$}}

1025: \de\uinProdTypeSig{\tx{$u \in \ProdTypeSig$}}

1026: \de\ul{\underline}

1027: \de\ulc{\ulcorner}

1028: \de\up{\tx{$^\wedge$}}

1029: \de\urc{\urcorner}

1030: \de\unspecs{\tx{\sf unspec}}

1031: \de\unspecN{\tx{$\unspecs_\natss$}}

1032: \de\unspecB{\tx{$\unspecs_\boolss$}}

1033: \de\ups{{\tx{$\upsilon$}}}

1034: \de\utos{{\tx{$u \to s$}}}

1035: \de\utov{{\tx{$u \to v$}}}

1036: \de\uu{{\tx{\ssf u}}}

1037: \de\uuu{\tx{{\sf u}\!\i}}

1038: \de\uuus{\tx{$\uuu_s$}}

1039: \de\uvec{\tx{$\vec{u}$}}

1040: \de\ux{{\tx{$u^*$}}}

1041: \de\val{\tx{\bi val}}

1042: \de\vals{\tx{\sf val}}

1043: \de\valss{{\tx{\ssf val}}}

1044: \de\var#1{\tx{\bi var$(#1)$}}

1045: \de\vars{\tx{\sf var}}

1046: \de\vects{\tx{\sf vect}}

1047: \de\vinProdTypeSig{\tx{$v \in \ProdTypeSig$}}

1048: \de\vvec{\tx{$\vec{v}$}}

1049: \de\vvecstrut{\vec{\mathstrut v}}

1050: \de\vx{{\tx{$v^*$}}}

1051: \de\vxvec{\tx{$\vec{v}\,^*$}}

1052: \de\whiles{\tx{\sf while}}

1053: \de\wrt{w.r.t\.}

1054: \de\wvec{\tx{$\vec{w}$}}

1055: \de\wx{{\tx{$w^*$}}}

1056: \de\x{\tx{\bi x}}

1057: \de\xbar{\tx{$\bar x$}}

1058: \de\xib{\tx{$\bs\xi$}}

1059: \de\xii{\tx{$\xi$}}

1060: \de\xinAbaru{\tx{$x \in \Abaru$}}

1061: \de\xinAu{\tx{$x \in \Au$}}

1062: \de\xinAw{\tx{$x \in \Aw$}}

1063: \de\xtt{{\tx{\tt x}}}

1064: \de\xttx{\tx{$\xtt^*$}}

1065: \de\xtts{{\tx{\smalltt x}}}

1066: \de\xttvec{\tx{$\vec{\xtt}$}}

1067: \de\xvec{\tx{$\vec{x}$}}

1068: \de\xx{\tx{$x^*$}}

1069: \de\xxvec{\tx{$\xvec^*$}}

1070: \de\xxx{\tx{${*}{*}{*}$}}

1071: \de\xxxx{\tx{${*}{*}{*}{*}$}}

1072: \de\xxxxx{\tx{${*}{*}{*}{*}{*}$}}

1073: \de\xxxxxx{\tx{${*}{*}{*}{*}{*}{*}$}}

1074: \de\yinAs{\tx{$y \in \As$}}

1075: \de\yinAv{\tx{$y \in \Av$}}

1076: \de\yvec{\tx{$\vec{y}$}}

1077: \de\yx{\tx{$y^*$}}

1078: \de\yxvec{\tx{$\yvec\,^*$}}

1079: \de\ytt{{\tx{\tt y}}}

1080: \de\yttx{\ytt\str}

1081: \de\ztt{{\tx{\tt z}}}

1082: \de\zttx{\ztt\str}

1083: \de\zvec{\tx{$\vec{z}$}}

1084: \de\zx{\tx{$z^*$}}

1085: \de\zxvec{\tx{$\zvec\,^*$}}

1086: 

1087: 

1088: % *** MACROS ***

1089: 

1090: \newcount\mmm

1091: \de\itemn#1{\mmm=#1\par\indentn\mmm \advance\mmm by 1

1092:       \hangindent\mmm \parindent \textindent

1093:  }

1094: 

1095: \newcount\nnn

1096: 

1097: \global\de\indentn#1{

1098: \nnn=#1

1099: \loop \ifnum\nnn>0 

1100:       \indent

1101:       \advance\nnn by -1

1102: \repeat

1103: }

1104: 

1105: \de\itemm{\itemn1}

1106: \de\itemmm{\itemn2}

1107: 

1108: \newcount\ppp

1109: 

1110: \global\de\quadn#1{

1111: \n=#1

1112: \loop \ifnum\ppp>0

1113:         \quad

1114:         \advance\ppp by -1

1115: \repeat

1116: }

1117: 

1118: 

1119: \firstfoot{{\ssmallrm ACM Transactions on Computational Logic,

1120:            Vol.\ TBD, No.\ TBD, TBD TBD, Pages TBD.}}

1121: \runningfoot{{\ssmallrm ACM Transactions on Computational Logic,

1122:            Vol.\ TBD, No.\ TBD, TBD TBD.}}

1123: 

1124: \de\AExA{\tx{$\AEE\,_\xtt^A$}}

1125: \de\APhi{\tx{$A^\Phi$}}

1126: \de\APsi{\tx{$A^\Psi$}}

1127: \de\ATTT{\tx{$(A,\TTT)$}}

1128: \de\AbarS{\tx{$\Abar^{\Sbi}$}}

1129: \de\AbsComp{\tx{\bi AbsComp}}

1130: \de\AbsCompA{\tx{$\AbsComp(A)$}}

1131: \de\Ad{\tx{$(A,d)$}}

1132: \de\Af{\tx{$(A,f)$}}

1133: \de\AfalA{\tx{$(A,\,\falA)$}}

1134: \de\AgalAfalA{\tx{$(A,\,\galA,\,\falA)$}}

1135: \de\AinMinNStdAlgSigE{\tx{$A\in\MinNStdAlgSigE$}}

1136: \de\AinMinNStdAlgSigT{\tx{$A\in\MinNStdAlgSigT$}}

1137: \de\AinNStdAlgSigE{\tx{$A\in\NStdAlgSigE$}}

1138: \de\AinNStdAlgSigT{\tx{$A\in\NStdAlgSigT$}}

1139: \de\AlgSigE{\Alg\SigE}

1140: \de\AlgSigF{\Alg\SigF}

1141: \de\AlgSigT{\Alg\SigT}

1142: \de\Approx{\tx{\bi Approx}}

1143: \de\ArrAx{\tx{\sf ArrAx}}

1144: \de\ArrAxSig{\tx{$\ArrAx(\Sig)$}}

1145: \de\Asprod{\tx{$A_{s_1}\times\dots\times A_{s_n}$}}

1146: \de\Aspx{\tx{$A_{s'}^*$}}

1147: \de\Ass{\tx{$A_{s_1},\dots,, A_{s_n}$}}

1148: \de\AtStx{\tx{$\AtSt\,_\xtt$}}

1149: \de\Axbar{\tx{$\ol{\Ax}$}}

1150: \de\AxfalA{\tx{$(\Ax,\,\falA)$}}

1151: \de\AxgalAfalA{\tx{$(\Ax,\,\galA,\,\falA)$}}

1152: \de\Axp{\tx{$\Ax'$}}

1153: \de\angte{\ang{\teasA \mid \sinSortSig}}

1154: \de\BBBbar{\tx{$\ol \BBB$}}

1155: \de\BUEq{\tx{\bi BUEq}}

1156: \de\BUEqSig{\BUEq(\Sig)}

1157: \de\BddAx{\tx{\sf BddAx}}

1158: \de\BddAxSig{\tx{$\BddAx(\Sig)$}}

1159: \de\BddAxo{\tx{$\BddAx^0$}}

1160: \de\BddAxoSig{\tx{$\BddAxo(\Sig)$}}

1161: \de\CA{\tx{$C(A)$}}

1162: \de\CCCInd{\tx{\CCC-\Ind}}

1163: \de\CCCIndSig{\tx{\CCC-\IndSig}}

1164: \de\CCCIndi{\tx{\CCC-\Indi}}

1165: \de\CCCIndiSig{\tx{\CCC-\IndiSig}}

1166: \de\CCCN{\tx{$\CCC^N$}}

1167: \de\CCCNL{\tx{$\CCC^{N,<}$}}

1168: \de\CCCo{\tx{$\CCC_0$}}

1169: \de\CompI{\tx{$\Comp_1$}}

1170: \de\CompIA{\tx{$\Comp_1^A$}}

1171: \de\CompN{\tx{$\Comp(\NN)$}}

1172: \de\CompR{\tx{$\Comp(\RR)$}}

1173: \de\CompalA{\tx{$\Comp_\al(A)$}}

1174: \de\CompxA{\tx{$\Comp\,_\xtt^A$}}

1175: \de\ConcRep{\tx{\bi ConcRep}}

1176: \de\ConcRepA{\tx{$\ConcRep(A)$}}

1177: \de\CondBUEq{\tx{\sf CondBUEq}}

1178: \de\CondBUEqSig{\tx{$\CondBUEq(\Sig)$}}

1179: \de\CondEq{\tx{\sf CondEq}}

1180: \de\CondEqSig{\tx{$\CondEq(\Sig)$}}

1181: \de\CondEqom{\tx{$\CondEq_\omb$}}

1182: \de\CondEqomSig{\tx{$\CondEqom(\Sig)$}}

1183: \de\CondSUEq{\tx{\sf CondSUEq}}

1184: \de\CondSUEqom{\tx{$\CondSUEq_\om$}}

1185: \de\CondSUEqomSig{\tx{$\CondSUEqom(\Sig)$}}

1186: \de\EU{\tx{$E^U$}}

1187: \de\EUus{\tx{$E^U_{u,s}$}}

1188: \de\EVm{\tx{$E^V_m$}}

1189: \de\EVus{\tx{$E^V_{u,s}$}}

1190: \de\EWm{\tx{$E^W_m$}}

1191: \de\EWn{\tx{$E^W_n$}}

1192: \de\Emz{\tx{$E_m(\ztt)$}}

1193: \de\Eal{\tx{$E_\al$}}

1194: \de\Ealx{\tx{$E_\al^*$}}

1195: \de\Ee{\tx{\sf E}}

1196: \de\Ef{\tx{$E^f$}}

1197: \de\Egame{\tx{$E_{\game}$}}

1198: \de\Ekbar{\tx{$E(\kbar)$}}

1199: \de\Enbar{\tx{$E(\nbar)$}}

1200: \de\EqSig{\Eqb(\Sig)}

1201: \de\Eqn{\tx{\bi Eqn}}

1202: \de\EqnA{\tx{$\Eqn(A)$}}

1203: \de\Eqom{\tx{$\Eqb_\omb$}}

1204: \de\EqomSig{\Eqom(\Sig)}

1205: \de\EtilWn{\tx{$\til{E}^W_n$}}

1206: \de\Eval{\tx{\bi Eval}}

1207: \de\EvalAx{\tx{\sf EvalAx}}

1208: \de\EvalAxSig{\tx{$\EvalAx(\Sig)$}}

1209: \de\Ez{\tx{$E(\ztt)$}}

1210: \de\FAP{\tx{\bi FAP}}

1211: \de\FAPA{\tx{$\FAP(A)$}}

1212: \de\FAPC{\tx{\bi FAPC}}

1213: \de\FAPCA{\tx{$\FAPC(A)$}}

1214: \de\FAPCS{\tx{\bi FAPCS}}

1215: \de\FAPCSA{\tx{$\FAPCS(A)$}}

1216: \de\FAPS{\tx{\bi FAPS}}

1217: \de\FAPSA{\tx{$\FAPS(A)$}}

1218: \de\FOL{\tx{\sf FOL}}

1219: \de\FOLSig{\tx{$\FOL(\Sig)$}}

1220: \de\FOLom{\tx{$\FOL_\omb$}}

1221: \de\FOLomSig{\tx{$\FOLom(\Sig)$}}

1222: \de\FOLomo{\tx{$\FOL^{\bk 0}_\omb$}}

1223: \de\FOLomoSig{\tx{$\FOLomo(\Sig)$}}

1224: \de\FPD{\tx{\bi FPD}}

1225: \de\FPDA{\tx{$\FPD(A)$}}

1226: \de\FU{\tx{$F^U$}}

1227: \de\FUus{\tx{$F^U_{u,s}$}}

1228: \de\FVus{\tx{$F^V_{u,s}$}}

1229: \de\Fal{\tx{$F_\al$}}

1230: \de\Falhat{\tx{$F_\alhat$}}

1231: \de\Falx{\tx{$F_\al^*$}}

1232: \de\Falminx{\tx{$F_\almin^*$}}

1233: \de\Ff{\tx{$F^f$}}

1234: \de\Fgame{\tx{$F_{\gamma,e}$}}

1235: \de\Fgamex{\tx{$F_{\gamma,e}^*$}}

1236: \de\Firstx{\tx{$\First\,_\xtt$}}

1237: \de\Fo{{\tx{$F^\circ$}}}

1238: \de\Fotil{\tx{$\til Fo$}}

1239: \de\ForProcSig{\tx{\For\Proc(\Sig)}}

1240: \de\Fmu{\tx{$F_\mu$}}

1241: \de\ForSigbar{\tx{$\For\,(\Sigbar)$}}

1242: \de\ForoAbar{\tx{$\Foro(\Abar)$}}

1243: \de\ForoSigbar{\tx{$\Foro(\Sigbar)$}}

1244: \de\ForolamSigbar{\tx{$\Forolam(\Sigbar)$}}

1245: \de\ForolamAbar{\tx{$\Forolam(\Abar)$}}

1246: \de\Forxp{\tx{$\Forx'$}}

1247: \de\GGGN{\tx{$\GGG^N$}}

1248: \de\GGGo{\tx{$\GGG_0$}}

1249: \de\GamDel{\tx{$\Gam \seqt \Del$}}

1250: \de\GamP{\tx{$\Gam \seqt P$}}

1251: \de\GamQ{\tx{$\Gam \seqt Q$}}

1252: \de\GamR{\tx{$\Gam \seqt R$}}

1253: \de\GL{\tx{\rm GL}}

1254: \de\GLmI{\tx{$\GL_m(I)$}}

1255: \de\GLtmI{\tx{$\GL^{\ssz\tx{\rm T}}_m(I)$}}

1256: \de\HF{\tx{\bi HF}}

1257: \de\HFA{\tx{$\HF(A)$}}

1258: \de\HaltAP{\HaltA{P}}

1259: \de\HaltTestu{\tx{$\HaltTests_u$}}

1260: \de\HaltTestuA{\tx{$\HaltTests_u^A$}}

1261: \de\IIId{\tx{$\III^d$}}

1262: \de\IIIp{\tx{$\III_p$}}

1263: \de\IIItN{\tx{$\III_t^N$}}

1264: \de\INS{\Init\NStdAlg}

1265: \de\INSSigE{\tx{$\INS\SigE$}}

1266: \de\INSSigF{\tx{$\INS\SigF$}}

1267: \de\INSSigT{\tx{$\INS\SigT$}}

1268: \de\IOP{\tx{$\IO_P$}}

1269: \de\ISigE{\Init\SigE}

1270: \de\ISigKK{\Init\SigKK}

1271: \de\ISigT{\Init\SigT}

1272: \de\In{\tx{$I^n$}}

1273: \de\Iq{\tx{$I^q$}}

1274: \de\JSig{\tx{$J\redSig$}}

1275: \de\KKPhi{\tx{$\Bbb K ^{\Phi}$}}

1276: \de\KKbarS{\tx{$\KKbar^{\Sbi}$}}

1277: \de\KKp{\tx{$\Bbb K '$}}

1278: \de\LJ{\tx{\sf LJ}}

1279: \de\LJe{\tx{$\LJ_{\es}$}}

1280: \de\LJeSig{\LJe(\Sig)}

1281: \de\LK{\tx{\sf LK}}

1282: \de\LKe{\tx{$\LK_{\es}$}}

1283: \de\LKeSig{\LKe(\Sig)}

1284: \de\Lomiom{\tx{$L_{\om_1\om}$}}

1285: \de\MA{\tx{$\MMM_A$}}

1286: \de\MCA{\tx{$\MC(A)$}}

1287: \de\MCAbar{\tx{$\MC(\Abar)$}}

1288: \de\NNmin{\tx{$\NN^-$}}

1289: \de\NNp{\tx{$\NN'$}}

1290: \de\NNtoA{\tx{$[\NN\to \!A]$}}

1291: \de\NNtoAs{\tx{$[\NN\to \!A_s]$}}

1292: \de\NNtoBB{\tx{$[\NN\to \BB]$}}

1293: \de\NNtoRR{\tx{$[\NN\to \RR]$}}

1294: \de\NNNbar{\tx{$\ol \NNN$}}

1295: \de\NNNmin{\tx{$\NNN^-$}}

1296: \de\NNNmino{\tx{$\NNN_0^-$}}

1297: \de\NNbar{\tx{$\ol{\NN}$}}

1298: \de\NNtoNN{\tx{$[\NN\to\NN]$}}

1299: \de\MinNStdAlgSigE{\tx{$\MinNStdAlg\SigE$}}

1300: \de\MinNStdAlgSigT{\tx{$\MinNStdAlg\SigT$}}

1301: \de\NStdAlgSigE{\tx{$\NStdAlg\SigE$}}

1302: \de\NStdAlgSigF{\tx{$\NStdAlg\SigF$}}

1303: \de\NStdAlgSigG{\tx{$\NStdAlg\SigG$}}

1304: \de\NStdAlgSigT{\tx{$\NStdAlg\SigT$}}

1305: \de\NStdAlgSigp{\NStdAlg(\Sigp)}

1306: \de\NStdAlgSigx{\NStdAlg(\Sigx)}

1307: \de\NStdAx{\tx{\sf NStdAx}}

1308: \de\NStdAxo{\tx{$\NStdAx_0$}}

1309: \de\NStdAxSig{\tx{$\NStdAx(\Sig)$}}

1310: \de\NStdAxoSig{\tx{$\NStdAxo(\Sig)$}}

1311: \de\NtoN{\tx{$\Ns\to\Ns$}}

1312: \de\Omal{\tx{$\Om_\al$}}

1313: \de\PAN{\tx{$P^{A^N}$}}

1314: \de\PEabc{\tx{$\PE_\abct$}}

1315: \de\PEabcA{\tx{$\PE\,_\abct^A$}}

1316: \de\PRAuv{\tx{$\PRA_{\utov}$}}

1317: \de\PRAbar{\tx{$\PR(\Abar)$}}

1318: \de\PRN{\tx{$\PR(\NNN)$}}

1319: \de\PRNbar{\tx{$\PR(\NNNbar)$}}

1320: \de\PRSchemeSig{\PR\Scheme(\Sig)}

1321: \de\PRSigbar{\tx{$\PR(\Sigbar)$}}

1322: \de\PRSiguv{\tx{$\PRSig_\utov$}}

1323: \de\PRSigbaruv{\tx{$\PR(\Sigbar)_\utov$}}

1324: \de\PRetc{PR (etc.)}

1325: \de\PRminAbar{\tx{$\PRmin(\Abar)$}}

1326: \de\PRminSigbar{\tx{$\PRmin(\Sigbar)$}}

1327: \de\PRoAbar{\tx{$\PRo(\Abar)$}}

1328: \de\PRoSigbar{\tx{$\PRo(\Sigbar)$}}

1329: \de\PRolamAbar{\tx{$\PRolam(\Abar)$}}

1330: \de\PRolamSigbar{\tx{$\PRolam(\Sigbar)$}}

1331: \de\PRor{\PR\ (or \PRx, or \muPR, or \muPRx)}

1332: \de\PRxAbar{\tx{$\PRx(\Abar)$}}

1333: \de\PRxSigbar{\tx{$\PRx(\Sigbar)$}}

1334: \de\PRxSiguv{\tx{$\PRxSig_\utov$}}

1335: \de\Per#1{\tx{\bi Per\,$(#1)$}}

1336: \de\PinProcuv{\tx{$P\in\Procuv$}}

1337: \de\Procabc{\tx{$\Proc\,_\abct$}}

1338: \de\Psrch{\tx{$P_{srch}$}}

1339: \de\QF{\tx{\sf QF}}

1340: \de\RAB{\tx{$R_A(B)$}}

1341: \de\RABs{\tx{$\RAB_s$}}

1342: \de\REC{\tx{REC}}

1343: \de\RRBBfield{\tx{$\RR,\,\BB;\,0,\,1,\,x+y,\,-x,\,x\cdot y,\,x^{-1}$}}

1344: \de\RRRLE{\tx{$\RRR^{<,E}$}}

1345: \de\RRRLEN{\tx{$\RRR^{<,E,N}$}}

1346: \de\RRRLN{\tx{$\RRR^{<,N}$}}

1347: \de\RRRLp{\tx{$\RRR_p^<$}}

1348: \de\RRRNt{\tx{$\RRR^N_t$}}

1349: \de\RRRNIt{\tx{$\RRR^{N,I}_t$}}

1350: \de\RRRd{\tx{$\RRR^d$}}

1351: \de\RRRo{\tx{$\RRR_0$}}

1352: \de\RRRp{\tx{$\RRR_p$}}

1353: \de\RRRtN{\tx{$\RRR_t^N$}}

1354: \de\RRfield{\tx{$\RR;\,0,\,1,\,x+y,\,-x,\,x\cdot y,\,x^{-1}$}}

1355: \de\RRtwon{\tx{$\RR^{2n}$}}

1356: \de\Rc{\tx{$R^\css$}}

1357: \de\Rec{\tx{\bi Rec}}

1358: \de\RemxA{\tx{$\Rem\,_\xtt^A$}}

1359: \de\RepxA{\tx{$\Rep\,_\xtt^A$}}

1360: \de\RepxxA{\tx{$\Rep\,_{\xtt \,*}^A$}}

1361: \de\RestA{\tx{$\Rest\,^A$}}

1362: \de\RestxA{\tx{$\Rest\,_\xtt^A$}}

1363: \de\SAsig{\tx{$\SA\sig$}}

1364: \de\SExA{\tx{$\SE\,_\xtt^A$}}

1365: \de\SR{\tx{\bi SR}}

1366: \de\SRA{\tx{$\SR(A)$}}

1367: \de\SUEq{\tx{\bi SUEq}}

1368: \de\SUEqSig{\SUEq(\Sig)}

1369: \de\SaA#1{\tx{$\ang{#1}^A$}}

1370: \de\San{\cS,a,n}

1371: \de\Sadeln{\cS,\,\adel,\,n}

1372: \de\SasA#1{\tx{$\ang{#1}_s^A$}}

1373: \de\SasAx#1{\tx{$\ang{#1}_s^\Ax$}}

1374: \de\SasnA#1#2{\tx{$\ang{#1}_{s,#2}^A$}}

1375: \de\SauA#1{\tx{$\ang{#1}_u^A$}}

1376: \de\SavA#1{\tx{$\ang{#1}_v^A$}}

1377: \de\Se{\tx{$\Sbi_e$}}

1378: \de\SigE{\tx{$(\Sig,E)$}}

1379: \de\SigF{\tx{$(\Sig,F)$}}

1380: \de\SigG{\tx{$(\Sig,G)$}}

1381: \de\SigIIId{\tx{$\Sig(\IIId)$}}

1382: \de\SigKK{\tx{$(\Sig,\KK)$}}

1383: \de\SigTF{\tx{$(\Sig,T,\FFF)$}}

1384: \de\SigXTF{\tx{$(\Sig,X,T,\FFF)$}}

1385: \de\Sigal{\tx{$\Sig_\al$}}

1386: \de\SigalEal{\tx{$(\Sigal,\Eal)$}}

1387: \de\Sigalx{\tx{$\Sig_\al^*$}}

1388: \de\SigalxEal{\tx{$(\Sigalx,\Eal)$}}

1389: \de\SigalxFal{\tx{$(\Sigalx,\Fal)$}}

1390: \de\SigbarS{\tx{$\Sigbar^{\Sbi}$}}

1391: \de\SigIIIdx{\tx{$\SigIIId^*$}}

1392: \de\Sigf{\tx{$\Sig_\fss$}}

1393: \de\SigpSig{{\tx{$\Sigp/\Sig$}}}

1394: \de\Sigtilxn{\tx{$\til{\Sig}^*_n$}}

1395: \de\Sigxgame{\tx{$\Sig_{\gam,e}^*$}}

1396: \de\Sigxm{\tx{$\Sig_m^*$}}

1397: \de\Sigxn{\tx{$\Sig_n^*$}}

1398: \de\Sigxus{\tx{$\Sig_{u,s}^*$}}

1399: \de\Sigxusp{\tx{$\Sig_{u,s}^{*\prime}$}}

1400: \de\Sigxp{\tx{$\Sigx'$}}

1401: \de\Sigxpm{\tx{$\Sig^{*\prime}_m$}}

1402: \de\Sigxpn{\tx{$\Sig^{*\prime}_n$}}

1403: \de\Sinit{\tx{$S_{init}$}}

1404: \de\SinStmtx{\tx{$S\in\Stmtx$}}

1405: \de\SnapxA{\tx{$\Snap_\xtt^A$}}

1406: \de\Splus{\Stimes}

1407: \de\StateAcupx{\tx{$\StateA\cup\{*\}$}}

1408: \de\StatesA{\tx{$\State_s(A)$}}

1409: \de\StdAlgSigE{\tx{$\StdAlg\SigE$}}

1410: \de\StdAlgSigF{\tx{$\StdAlg\SigF$}}

1411: \de\StdAlgSigT{\tx{$\StdAlg\SigT$}}

1412: \de\StdAlgSigbar{\tx{$\StdAlg(\Sigbar)$}}

1413: \de\StdAlgSigbarS{\tx{$\StdAlg(\SigbarS)$}}

1414: \de\StdAlgSigp{\tx{$\StdAlg(\Sigp)$}}

1415: \de\Stmtx{\tx{$\Stmt_{\,\xtt}$}}

1416: \de\SubAlgStage{\tx{\bi SubAlgStage}}

1417: \de\SubAlgStagexus{\tx{$\SubAlgStage\,_{\xtt,u,s}$}}

1418: \de\Sx{\tx{$[S,\xtt]$}}

1419: \de\TExsA{\tx{$\TE\,_{\xtt,s}^A$}}

1420: \de\TP{\tx{$\TTT(P)$}}

1421: \de\TSigKK{\Tbi\SigKK}

1422: \de\TSigTF{\Tbi\SigTF}

1423: \de\TSigX{\tx{$\Tbi(\Sig,X)$}}

1424: \de\TSigXTF{\Tbi\SigXTF}

1425: \de\TSigx{\tx{$\Tbi(\Sig,\xtt)$}}

1426: \de\TSx{\TTT\Sx}

1427: \de\TermTupa{\tx{$\TermTup_\att$}}

1428: \de\TermTupu{\tx{$\TermTup_u$}}

1429: \de\TermTupx{\tx{$\TermTup_\xtt$}}

1430: \de\TermTupxu{\tx{$\TermTup_{\xtt,u}$}}

1431: \de\TermTupxv{\tx{$\TermTup_{\xtt,v}$}}

1432: \de\TermTupxw{\tx{$\TermTup_{\xtt,w}$}}

1433: \de\Terma{\tx{$\Term_{\,\att}$}}

1434: \de\TermaN{\tx{$\Term_{\,\att}^N$}}

1435: \de\Termabool{\tx{$\Term_{\,\att,\boolss}$}}

1436: \de\Termaboolx{\tx{$\Term_{\,\att,\boolss}^*$}}

1437: \de\Termanatx{\tx{$\Term_{\,\att,\natss}^*$}}

1438: \de\Termas{\tx{$\Term_{\,\att,s}$}}

1439: \de\TermasN{\tx{$\Term_{\,\att,s}^N$}}

1440: \de\TermasuuN{\tx{$\Term_{\,\att,s}^{\uu,N}$}}

1441: \de\Termasx{\tx{$\Term_{\,\att,s}^*$}}

1442: \de\TermauuN{\tx{$\Term_{\,\att}^{\uu,N}$}}

1443: \de\Termaw{\tx{$\Term_{\,\att,w}$}}

1444: \de\Termax{\tx{$\Term_{\,\att}^*$}}

1445: \de\Terms{\tx{$\Term_s$}}

1446: \de\Termx{\tx{$\Term_{\,\xtt}$}}

1447: \de\Termxs{\tx{$\Term_{\,\xtt,s}$}}

1448: \de\TTTp{\tx{$\TTT'$}}

1449: \de\TTTs{\tx{$\TTT_s$}}

1450: \de\TsSigx{\tx{$\Tbi_s(\Sig,\xtt)$}}

1451: \de\VarTupu{\tx{$\VarTup_u$}}

1452: \de\WhileAp{\tx{$\While(A')$}}

1453: \de\WhileProc{\While\Proc}

1454: \de\WhileProcSig{\tx{$\WhileProc(\Sig)$}}

1455: \de\WhileSigp{\tx{$\While(\Sigp)$}}

1456: \de\Whilexp{\tx{$\Whilex'$}}

1457: \de\abct{{\tx{$\att,\btt,\ctt$}}}

1458: \de\ack{\tx{\bi ack}}

1459: \de\aeuA{\tx{$\aee\biu^A$}}

1460: \de\aexA{\tx{$\aee\,_\xtt^A$}}

1461: \de\aexSA{\tx{$\aee\,_{\xtt,S}^A$}}

1462: \de\adel{\tx{$(a,\delbA)$}}

1463: \de\alAbar{\tx{$\al^\Abar$}}

1464: \de\allR{\tx{$\all R$}}

1465: \de\allomR{\tx{$\all_\om R$}}

1466: \de\almin{{\tx{$\alpha^-$}}}

1467: \de\approxKK{\tx{$\approx_\KK$}}

1468: \de\approxTF{\tx{$\approx_{T,\FFF}$}}

1469: \de\bSl{\tx{$\bbb_{S,\lam}$}}

1470: \de\bSlk{\tx{$\bbb_{S,\lam_k}$}}

1471: \de\bbSA{\tx{$\bb{S}^A$}}

1472: \de\bbbk{\tx{$\bbb_k$}}

1473: \de\bbbkx{\tx{$\bbb_k^*$}}

1474: \de\bbtA{\tx{$\bb{t}^A$}}

1475: \de\bSk{\tx{$\bbb_{S,k}$}}

1476: \de\cP{\cnr{P}}

1477: \de\cS{\cnr{S}}

1478: \de\compseqxA#1{\tx{$\comp\seq\,_\xtt^A(#1)$}}

1479: \de\compxA{\tx{$\comp\,_\xtt^A$}}

1480: \de\compxSA{\tx{$\comp\,_{\xtt,S}^A$}}

1481: \de\compuaA{\tx{$\comp\biu\,_\att^A$}}

1482: \de\comptree{\tx{\comp\tree}}

1483: \de\concat{^\smallfrown}

1484: \de\cxtt{\cnr{\xtt}}

1485: \de\dI{\tx{$d_\intvlss$}}

1486: \de\dR{\tx{$d_\realss$}}

1487: \de\delbA{\tx{$\delb_A$}}

1488: \de\delbAu{\tx{$\delb_A^u$}}

1489: \de\delbAv{\tx{$\delb_A^v$}}

1490: \de\delbAw{\tx{$\delb_A^w$}}

1491: \de\delbs{\tx{$\delb^s$}}

1492: \de\delbsA{\tx{$\delb^s_A$}}

1493: \de\delbu{\tx{$\delb^u$}}

1494: \de\delbuA{\tx{$\delb^u_A$}}

1495: \de\en{\tx{\bi en}}

1496: \de\enA{\tx{$\en_A$}}

1497: \de\enAs{\tx{$\en_A^s$}}

1498: \de\enAsx{\tx{$\en_A^{s*}$}}

1499: \de\enAx{\tx{$\en_A\str$}}

1500: \de\endd$1{\tx{{\bi end}\,($$1$)}}

1501: \de\eqNm{\tx{$\eqs_{\natss^-}$}}

1502: \de\eqp{\tx{$\eqs_p$}}

1503: \de\evalAs{\tx{$\evals_s^A$}}

1504: \de\exomL{\tx{$\ex_\om L$}}

1505: \de\exps{\tx{\sf exp}}

1506: \de\expss{\tx{\ssf exp}}

1507: \de\fal{\tx{$\fs_\al$}}

1508: \de\falA{\tx{$\fs_\al^A$}}

1509: \de\falhat{\tx{$\fs_\alhat$}}

1510: \de\falhatA{\tx{$\fs_\alhat^A$}}

1511: \de\fmu{\tx{$f_\mu$}}

1512: \de\fps{\tx{\sf fp}}

1513: \de\fsA{\tx{$\fs^A$}}

1514: \de\fsigA{\tx{$\fs_\sig^A$}}

1515: \de\fupsA{\tx{$\fs_\ups^A$}}

1516: \de\gN{\tx{$g^\natss$}}

1517: \de\gal{\tx{$\gs_\al$}}

1518: \de\galA{\tx{$\gs_\al^A$}}

1519: \de\galhat{\tx{$\gs_\alhat$}}

1520: \de\galhatA{\tx{$\gs_\alhat^A$}}

1521: \de\gbeA{\tx{$\gs_\be^A$}}

1522: \de\galiA{\tx{$\gs_{\al,i}^A$}}

1523: \de\game{\tx{$(\gamma,e)$}}

1524: \de\grps{\tx{\sf grp}}

1525: \de\grpss{{\tx{\ssf grp}}}

1526: \de\haltP{\tx{$\halt_P$}}

1527: \de\haltS{\tx{$\halt_S$}}

1528: \de\hN{\tx{$h^\natss$}}

1529: \de\iI{\tx{$\iota_I$}}

1530: \de\ia{\tx{$\io_\att$}}

1531: \de\invs{\tx{\sf inv}}

1532: \de\invexps{\tx{\sf invexp}}

1533: \de\invexpss{\tx{\ssf invexp}}

1534: \de\isint{\tx{\sf is\_int}}

1535: \de\lamMCA{\tx{$\lamMC(A)$}}

1536: \de\lamMCAbar{\tx{$\lamMC(\Abar)$}}

1537: \de\lamPRA{\tx{$\lamPR(A)$}}

1538: \de\lamPRAbar{\tx{$\lamPR(\Abar)$}}

1539: \de\lamPRN{\tx{$\lamPR(\NNN)$}}

1540: \de\lamPRNbar{\tx{$\lamPR(\NNNbar)$}}

1541: \de\lamPRSigbar{\tx{$\lamPR(\Sigbar)$}}

1542: \de\lamPRminAbar{\tx{$\lamPRmin(\Abar)$}}

1543: \de\lamPRminSigbar{\tx{$\lamPRmin(\Sigbar)$}}

1544: \de\lammuMCA{\tx{$\lammuMC(A)$}}

1545: \de\lammuMCAbar{\tx{$\lammuMC(\Abar)$}}

1546: \de\lammuPRA{\tx{$\lammuPR(A)$}}

1547: \de\lammuPRAbar{\tx{$\lammuPR(\Abar)$}}

1548: \de\lammuPRN{\tx{$\lammuPR(\NNN)$}}

1549: \de\lammuPRNbar{\tx{$\lamPR(\NNNbar)$}}

1550: \de\lammuPRSigbar{\tx{$\lammuPR(\Sigbar)$}}

1551: \de\lammuPRminAbar{\tx{$\lammuPRmin(\Abar)$}}

1552: \de\lammuPRminSigbar{\tx{$\lammuPRmin(\Sigbar)$}}

1553: \de\lammuPRxAbar{\tx{$\lammuPRx(\Abar)$}}

1554: \de\lamz{\lam\ztt}

1555: \de\lastA{\tx{$\last^A$}}

1556: \de\lastxA{\tx{$\last_\xtt^A$}}

1557: \de\lsp{\tx{$\lss_p$}}

1558: \de\lseqs{\tx{\sf lseq}}

1559: \de\lseqp{\tx{$\lseqs_p$}}

1560: \de\maxval{\tx{\bi maxval}}

1561: \de\muPRAbar{\tx{$\muPR(\Abar)$}}

1562: \de\muPRSigbar{\tx{$\muPR(\Sigbar)$}}

1563: \de\muPRminAbar{\tx{$\muPRmin(\Abar)$}}

1564: \de\muPRminSigbar{\tx{$\muPRmin(\Sigbar)$}}

1565: \de\muPRoAbar{\tx{$\muPRo(\Abar)$}}

1566: \de\muPRoSigbar{\tx{$\muPRo(\Sigbar)$}}

1567: \de\muPRolamAbar{\tx{$\muPRolam(\Abar)$}}

1568: \de\muPRolamSigbar{\tx{$\muPRolam(\Sigbar)$}}

1569: \de\muPRxAbar{\tx{$\muPRx(\Abar)$}}

1570: \de\muPRxApprox{\tx{\muPRx-\Approx}}

1571: \de\muPRxApproxA{\tx{$\muPRxApprox(A)$}}

1572: \de\muPRxApproxAutos{\tx{$\muPRxApproxA_\utos$}}

1573: \de\muPRxSigbar{\tx{$\muPRx(\Sigbar)$}}

1574: \de\natmin{\tx{$\nats^-$}}

1575: \de\natsp{\tx{$\nats'$}}

1576: \de\natsx{\tx{$\nats^*$}}

1577: \de\natssmin{\tx{$\natss^-$}}

1578: \de\natxutos{\tx{$\nats\times u\to s$}}

1579: \de\natxutoss{{\tx{$\natss\times u\to s$}}}

1580: \de\notoverxA{\tx{$\notover\,_\xtt^A$}}

1581: \de\notoverxSA{\tx{$\notover\,_{\xtt,S}^A$}}

1582: \de\notoveruaA{\tx{$\notover\biu\,_\att^A$}}

1583: \de\ord{\tx{\bi ord}}

1584: \de\ordusA{\tx{$\ord_{u,s}^A$}}

1585: \de\orb#1{\tx{\bi orb\,$(#1)$}}

1586: \de\peabcA{\tx{$\pe\,_\abct^A$}}

1587: \de\peabcPA{\tx{$\pe\,_{\abct,P}^A$}}

1588: \de\per#1{\tx{\bi per\,$(#1)$}}

1589: \de\phipsi{\tx{$(\phi,\psi)$}}

1590: \de\phihat{\tx{$\hat{\phi}$}}

1591: \de\phis{{\tx{$\phi_s$}}}

1592: \de\pib{\tx{$\pi_\btt$}}

1593: \de\plustwo{\tx{\bi plus2}}

1594: \de\posreals{\tx{\sf pos-real}}

1595: \de\prodtt{\tx{\tt prod}}

1596: \de\preds{\tx{\sf pred}}

1597: \de\psis{{\tx{$\psi_s$}}}

1598: \de\qinNN{\tx{$q\in\NN$}}

1599: \de\redSigf{\,|\,_{\Sigf}}

1600: \de\redSigp{\,|\,_{\Sigp}}

1601: \de\remseqxA#1{\tx{$\rem\seq\,_\xtt^A(#1)$}}

1602: \de\remxA{\tx{$\rem\,_\xtt^A$}}

1603: \de\remxSA{\tx{$\rem\,_{\xtt,S}^A$}}

1604: \de\remuaA{\tx{$\rem\biu\,_\att^A$}}

1605: \de\restA{\tx{$\restt^A$}}

1606: \de\restuaA{\tx{$\restt\biu\,_\att^A$}}

1607: \de\restxA{\tx{$\restt\,_\xtt^A$}}

1608: \de\restxSA{\tx{$\restt\,_{\xtt,S}^A$}}

1609: \de\ruu{\tx{$r^\uu$}}

1610: \de\sE{{\tx{\ssf E}}}

1611: \de\saA#1{\tx{$\ang{#1}^A$}}

1612: \de\siginStateA{\tx{$\sig\in\StateA$}}

1613: \de\sigs{\tx{$\sig_s$}}

1614: \de\sinS{\tx{$s \in \Sbi$}}

1615: \de\sinSe{\tx{$s \in \Se$}}

1616: \de\sexA{\tx{$\se\,_\xtt^A$}}

1617: \de\sexSA{\tx{$\se\,_{\xtt,S}^A$}}

1618: \de\sigxtt{\tx{$\sig[\xtt]$}}

1619: \de\snapseqxA#1{\tx{$\snap\seq\,_\xtt^A(#1)$}}

1620: \de\snapuaA{\tx{$\snap\biu\,_{\att}^A$}}

1621: \de\snapxA{\tx{$\snap\,_\xtt^A$}}

1622: \de\snapxSA{\tx{$\snap\,_{\xtt,S}^A$}}

1623: \de\sprod{\tx{${s_1}\times\dots\times {s_n}$}}

1624: \de\stateuaA{\tx{$\state\biu\,_\att^A$}}

1625: \de\statexA{\tx{$\state\,_\xtt^A$}}

1626: \de\statexSA{\tx{$\state\,_{\xtt,S}^A$}}

1627: \de\subalg{\preceq}

1628: \de\tAsig{\tx{$\tA\sig$}}

1629: \de\teA{\tx{$\te^A$}}

1630: \de\teaBA{\tx{$\te\,_{\att,\boolss}^A$}}

1631: \de\teasA{\tx{$\te\,_{\att,s}^A$}}

1632: \de\teauA{\tx{$\te\,_{\att,u}^A$}}

1633: \de\teavA{\tx{$\te\,_{\att,v}^A$}}

1634: \de\teawA{\tx{$\te\,_{\att,w}^A$}}

1635: \de\texsA{\tx{$\te\,_{\xtt,s}^A$}}

1636: \de\texsAx{\tx{$\te\,_{\xtt,s}^\Ax$}}

1637: \de\texstA{\tx{$\te\,_{\xtt,s,t}^A$}}

1638: \de\texuA{\tx{$\te\,_{\xtt,u}^A$}}

1639: \de\texvA{\tx{$\te\,_{\xtt,v}^A$}}

1640: \de\texwA{\tx{$\te\,_{\xtt,w}^A$}}

1641: \de\texyxboolAx{\tx{$\te\,_{\xtt,\yttx,\boolss}^\Ax$}}

1642: \de\tinTSig{\tx{$t\in\TSig$}}

1643: \de\tinTSigx{\tx{$t\in\TSigx$}}

1644: \de\ttil{\tx{$\til t$}}

1645: \de\tree{\tx{\bi tree}}

1646: \de\truncs{\tx{\sf trunc}}

1647: \de\tupprimetimes#1#2#3{\tx{$#1_{#2}' \times \dots \times #1_{#3}'}}

1648: \de\tuu{\tx{$t^\uu$}}

1649: \de\uneqs{\tx{\sf uneq}}

1650: \de\uneqp{\tx{$\uneqs_p$}}

1651: \de\updates{\tx{\sf update}}

1652: \de\us{{\tx{$u,s$}}}

1653: \de\whicht{\tx{\tt which}}

1654: \de\vN{\tx{$v^\natss$}} 

1655: \de\valxaA{\tx{$\val_\xa^A$}}

1656: \de\valxaAt{\tx{$\val_\xa^A(t)$}}

1657: \de\ws{{\tx{$w(s)$}}}

1658: \de\wxx{{\tx{$w^{**}$}}}

1659: \de\xS{\cxtt,\,\cS}

1660: \de\xSan{\xS,\,a,\,n}

1661: \de\xa{{\ang{\xtt/a}}}

1662: \de\xiinNNtoA{\tx{$\xi \in \NNtoA$}} 

1663: \de\xixi{\xi_1,\xi_2}

1664: \de\xttt{\tx{$\ang{\xtt/t}$}}

1665: \de\zxx{\tx{$z^{**}$}}

1666: 

1667: 

1668: 

1669: % *** END OF HEADER ***

1670: 

1671: 

1672: \cbb{Abstract Computability and Algebraic Specifications}

1673: \bigskip

1674: \bigskip

1675: \ce{{\bf J.V. Tucker}

1676: }

1677: \sn

1678: \ce{\it Department of Computer Science,}

1679: \ce{\it University of Wales, Swansea  SA2 8PP, Wales}

1680: \ce{\tt J.V.Tucker@swansea.ac.uk}

1681: \bn

1682: \ce{{\bf J.I. Zucker}\footnotemark"*"}

1683: \footnotetext"*\ "{

1684: \n The research of the second author was supported 

1685: by a grant from the Natural Sciences

1686: and Engineering Research Council (Canada),

1687: and by a Visiting Fellowship from the Engineering

1688: and Physical Sciences Research Council (U.K.)

1689: }

1690: 

1691: \ce{\it Department of Computing and Software,}

1692: \ce{\it McMaster University, Hamilton, Ont\.  L8S 4L7, Canada}

1693: \ce{\tt zucker@mcmaster.ca}

1694: \bigskip

1695: \bigskip

1696: \bigskip

1697: \bigskip

1698: \bigskip

1699: \cbb{Abstract}

1700: 

1701: \bn

1702: {

1703: %\narrower\narrower

1704: \smallrm 

1705: Abstract computable functions are defined by

1706: abstract finite deterministic algorithms on many-sorted algebras.

1707: We show that there exist finite universal algebraic specifications

1708: that specify uniquely 

1709: (up to isomorphism) 

1710: \,($i$) all absract computable functions on any many-sorted algebra;

1711: \,and

1712: \,($ii$) all functions effectively approximable by 

1713: abstract computable functions on any metric algebra.

1714: We show that there exist universal algebraic specifications

1715: for all the classically computable functions on the

1716: set \,\RR\ \,of real numbers.

1717: The algebraic specifications used are mainly bounded universal

1718: equations and conditional equations.

1719: We investigate the initial algebra semantics of

1720: these specifications, and derive situations

1721: where algebraic specifications define precisely the computable functions.

1722: 

1723: \mn

1724: {\smallbf Categories and Subject Descriptors:}

1725: F.1.1 ({\smallbf Computation by Abstract Devices}):

1726: Models of Computation --- {\smallit computability theory\/};

1727: %F.3.1 ({\smallbf Logics and Meanings of Programs}):

1728: %\,Specifying and Verifying and Reasoning about Programs;

1729: F.4.1 ({\smallbf Mathematical Logic and Formal Languages}):

1730: \,Mathematical Logic --- {\smallit computability theory; \,proof theory\/}

1731: 

1732: \mn

1733: {\smallbf General Terms:}

1734: \,Abstract Computability,

1735: Algebraic Specification,

1736: Computable Analysis,

1737: Conditional Equations, Equational Logic, Metric Algebras

1738: 

1739: \mn

1740: {\smallbf Additional Key Words and Phrases:}

1741: \,Birkhoff's theorem, Grzegorczyk-Lacombe computability,

1742: initial  algebras, term models,

1743: many-sorted algebras, Mal'cev's theorem, topological algebras

1744: 

1745: }

1746: 

1747: \newpage

1748: 

1749: \Shead0{Introduction}

1750: Abstract computability theory is the theory of computable

1751: functions and relations over many-sorted algebras. It is a

1752: generalisation of classical recursion theory on the natural

1753: numbers, based on notions of finite deterministic

1754: computation on an arbitrary many-sorted algebra. An

1755: important feature of the theory is its analysis of

1756: computations that are {\it uniform\/} over classes of algebras, and

1757: a natural application of the theory is to analyse the scope and

1758: limits of models of computation and specification over

1759: abstract data types and their implementations. Since the 1960s, many

1760: abstract models of computation have been defined and

1761: classified, starting with the models of 

1762: E. Engeler, Y. Moschovakis, H. Friedman and J.C. Shepherdson,

1763: and generalised Church-Turing Theses for

1764: computation and specification have been formulated and

1765: defended \cite{tz:book,tz:jlp}.

1766: Here we will use the model of computation \,\muPRx\

1767: (a generalised form of Kleene schemes), which involves

1768: {\it simultaneous primitive recursion\/} and {\it least number search\/}

1769: over a many-sorted algebra augmented by the booleans,

1770: natural numbers and finite sequences of every sort. In

1771: \cite{tz:book} the model \,\muPRx\  is shown to be

1772: equivalent to \qwhiles-array programs over these algebras, the

1773: primary mathematical model of imperative programming.

1774: 

1775: Working with finite computation on any algebra enables us

1776: to develop a number of special computability theories

1777: for algebras, such as {\it rings and fields of real numbers\/} 

1778: \cite{jvt:cas,eng93,bss,bcss-manifesto,bcss}

1779: and {\it topological and metric algebras\/} \cite{tz:top}.

1780: For a comprehensive

1781: introduction to abstract computability, including a survey

1782: of its origins in the 1950s and principal literature, see

1783: our survey \cite{tz:hb}.

1784: 

1785: In this paper we prove theorems that show that functions

1786: that are abstractly computable over many-sorted algebras,

1787: or have abstractly computable {\it approximations\/} on topological

1788: algebras, can be specified by purely algebraic methods,

1789: but that the converse does not hold in the absence of

1790: certain topological conditions.

1791: 

1792: Algebraic specification methods characterise functions as

1793: the solutions of systems of algebraic formulae; 

1794: normally, the solutions are

1795: unique. By algebraic formulae, we mean

1796: equations

1797: $$

1798: t(\xtt) \ = \ t'(\xtt)

1799: $$

1800: or conditional equations

1801: $$

1802: t_1(\xtt) = t_1'(\xtt) \,\con \,\dots \,\con \,t_k(\xtt) = t'_k(\xtt) 

1803: \ \ \longto\ \ t(\xtt) = t'(\xtt),

1804: \tagx

1805: $$

1806: or, more generally, conditional formulae 

1807: $$

1808: R_1\con \dots \con R_k \ \longto\ R

1809: \tagxx

1810: $$

1811: where the formulae $R_i$ and $R$ are

1812: generalisations of equations, 

1813: making use of the distinguished sorts \nats\ of naturals and \reals\

1814: of reals

1815: (as we will see below).

1816: %The conditional formulae ($**$)

1817: %may enjoy certain

1818: %algebraic properties (\eg, having initial algebras).

1819: To define a unique solution for a system of equations,  in

1820: logic one often thinks of {\it definability up to isomorphism\/},

1821: and in computing one often thinks in terms of {\it initial algebra

1822: semantics\/} (or possibly {\it final algebra semantics\/}). However,

1823: notice that there are many more equational methods, \eg,

1824: for specifying concurrent processes using metric space

1825: methods to solve equations \cite{db-rutten,db-devink}, or

1826: for computing  solutions of differential or integral

1827: equations.

1828: 

1829: 

1830: In computation over a many-sorted algebra $A$ we use the

1831: booleans, natural numbers and finite sequences over $A$. 

1832: With regard to algebraic specifications over such structures,

1833: generalising conditional equations

1834: leads to the concept of

1835: {\it conditional bounded universal (BU) equations\/},

1836: in which the formulae $R_i$ and $R$ of ($**$)

1837: may have the form 

1838: $$

1839: t_1 = t_2 \qquad \tx{or} \qquad 

1840: \all \ztt<t \,[\,t_1= t_2\,]

1841: $$

1842: where the variable \ztt\ and term $t$ 

1843: are of sort \nats.

1844:  

1845: Conditional BU equations are new and provide us

1846: with more appropriate axiomatisations for some properties

1847: using the natural number sort; we show they are equivalent

1848: with conditional equations. The main theorems are first

1849: proved for conditional BU equations and the reduction

1850: method applied to obtain conditional equational

1851: specifications.

1852: 

1853: In the first part of the paper, we begin with the ``simple"

1854: situation where there is a system $E$ of conditional equations

1855: over a signature \Sig, and a \Sig-algebra $A$ such that $E$ has one and only

1856: one solution $f$ on $A$. We call this method of

1857: characterising functions {\it conditional equation definability\/}

1858: on $A$. We address the obvious general question: 

1859: 

1860: {\displaytext

1861: Does

1862: abstract computability imply conditional equation

1863: definability?

1864: 

1865: }\n

1866: The answer is yes, and we show that there exist 

1867: {\it universal specifications\/}

1868: that specify {\it all\/} computable functions, as follows

1869: (Section 5, Theorem 4).

1870: 

1871: \Thmn{A \,(Algebra)}

1872: Given a signature \Sig\ and function type \ta\ over \Sig,

1873: there

1874: exists a finite set of conditional equations \,\Ez\

1875: \,(with a distinguished natural number variable \,\ztt)

1876: over a finite expansion \Sigp\ of \Sig,

1877: \,such that for any

1878: abstract program \al\ over \Sig, if  $A$ is any  \Sig-algebra and

1879: $f$  a total function on $A$ of type \ta\ computed by \al, then $f$ is

1880: defined uniquely on $A$ by \,\Ekbar, \,where \kbar\ is

1881: a numeral instantiating \,\ztt\ which is effectively calculable

1882: from \al. The system \,\Ez\ \,is uniformly computable from \Sig\ and

1883: \ta.

1884: \endpr

1885: 

1886: Applying our abstract computability theory to {\it metric algebras\/},

1887: we can obtain an important, strictly broader, class of functions:

1888: namely, those {\it uniformly approximable by 

1889: abstractly computable functions\/}.

1890: In metric algebras, approximation is elegantly formulated 

1891: in terms of the distance function,

1892: which uses the sort \reals.

1893: This gives rise to 

1894: a broader class of conditional formulae than ($*$),

1895: called

1896: {\it conditional equations and inequalities\/},

1897: namely formulae ($**$) 

1898: in which the formulae $R_i$ and $R$ 

1899: may have the form

1900: $$

1901: t_1 = t_2 \qquad \tx{or} \qquad 

1902: t_1< t_2

1903: $$

1904: where, in the case of inequality ($t_1<t_2$),

1905: $t_1$ and $t_2$ are of sort \reals.

1906: %may be replaced by inequalities ($t_i<t_i'$) between terms of sort \reals,

1907: %thus:

1908: %$$

1909: %\aligned

1910: %&t_1(\xtt) = t_1'(\xtt) \ \con \ \dots \ \con \ t_j(\xtt) = t'_j(\xtt)

1911: %\ \con\\

1912: %&\qquad \con \ t_{j+1}(\xtt) < t_{j+1}'(\xtt) \ \con \ \dots 

1913: %\ \con \ t_k(\xtt)<t'_k(\xtt)\\

1914: %&\qqqquad\qqqquad \longto\ \ t(\xtt) = t'(\xtt)

1915: %\endaligned

1916: %\tagxx

1917: %$$

1918: %$$

1919: %\multline

1920: %t_1(\xtt) = t_1'(\xtt) \ \con \ \dots \ \con \ t_l(\xtt) = t'_l(\xtt)

1921: %\ \con

1922: % \ t_{j+1}(\xtt) < t_{j+1}'(\xtt) \ \con \ \dots 

1923: %\ \con \ t_k(\xtt)<t'_k(\xtt)\\

1924: %\longto\ \ t(\xtt) = t'(\xtt)

1925: %\endmultline

1926: %\tagxx

1927: %$$

1928: %for some $j$ ($1\le j \le k$),

1929: %where for $i>j$, \,$t_i$ and $t_i'$ are of sort \reals.

1930: 

1931: From Theorem A we then prove (Section 6, Theorem 2):

1932: 

1933: \Thmn{B \,(Metric algebra)}

1934: Given a signature \Sig\ and function type \ta\ over \Sig,

1935: there exists a finite

1936: set of conditional equations and inequalities 

1937: \,\Ez\ \,(with a distinguished natural number variable \ztt)

1938: over a finite expansion \Sigp\ of \Sig,

1939: \,such that for any

1940: abstract program \al\ over

1941: \Sig, if  $A$ is any  metric \Sig-algebra and $f$  a total function

1942: on $A$ of type \ta, approximable by \al\ in the following sense: 

1943: for all

1944: $a\in A$ and all \,$n$

1945: $$

1946: d(f(a), \,\bb{\al}(n, a)) < 2^{-n},

1947: $$

1948: then $f$ is

1949: defined uniquely on $A$ by \,\Ekbar,

1950: \, where \kbar\ is

1951: a numeral instantiating \,\ztt\ which is effectively calculable

1952: from \al. 

1953: The system \,\Ez\ \,is uniformly computable from

1954: \Sig\ and \ta.

1955: \endpr

1956: 

1957: Thus, {\sl there is a bound $B(\Sig,\tau)$ on the number of 

1958: conditional equations and inequalities

1959: needed to define all computable or computably

1960: approximable functions, that depends only on the signature \Sig\

1961: and the function type \ta.}

1962: 

1963: Using Theorem B, we show that all the classically computable functions

1964: of real analysis are unique solutions of finite sets of

1965: conditional equations and inequalities.  

1966: These classically computable functions have several characterisations,

1967: starting with those of Grzegorczyk \cite{grzeg55,grzeg57} 

1968: and Lacombe \cite{lacombe55},

1969: and hence are often called {\it GL-computable\/}.

1970: The familiar functions of analysis, such as 

1971: \,$\sin x$, $e^x$, $\log x$, etc.,

1972: are all GL-computable.

1973: Now there exists a certain simple total metric algebra

1974: \,\IIId\ \,over the real unit interval $I = [0,1]$, such that the 

1975: total functions on $I$ which are uniformly approximable 

1976: by abstractly computable functions on \,\IIId\

1977: \,are precisely the GL-computable functions on $I$

1978: \cite{tz:top,tz:hb}.

1979: We prove the following (Section 6, Theorem 3).

1980: 

1981: \Thmn{C \,(Metric algebra over a real interval)}

1982: For each positive integer $m$ 

1983: there is a signature \Sigxm\ 

1984: which is an expansion of the signature of \,\IIId\

1985: \,by finitely many function symbols,

1986: \,and a finite system of conditional equations and inequalities

1987: \,\Emz\ 

1988: \,(with a distinguished natural number variable \ztt)

1989: \,over \Sigxm, 

1990: \,such that any total function

1991: \,$f\:[0,1]^m\to\RR$ 

1992: \,that is GL-computable,

1993: is the unique solution of \,\Ekbar\ \,for some substitution of

1994: a numeral \,\kbar\ \,for \,\ztt.  

1995: The specification \,$(\Sigxm,\,\Emz)$ \,is uniformly computable from $m$.

1996: \endpr

1997: 

1998: 

1999: Thus

2000: {\sl there is a bound $B(m)$ on the

2001: number of conditional equations and inequalities 

2002: needed to define all $m$-ary GL-computable functions on $[0,1]$.\/}

2003: 

2004: 

2005: The signature \Sigxm\ consists of

2006: the sorts of booleans \BB\ and naturals \NN, with their standard operations;

2007: the sort of reals \RR, with its ring operations,

2008: together with division of reals by naturals;

2009: the sort of the unit interval $I$,

2010: with its embedding into \RR;

2011: the sort of finite arrays on \RR\

2012: with their standard operations;

2013: the standard metrics on all these sorts;

2014: a ``universal function" which approximably abstractly computes

2015: all $m$-ary GL-computable total functions on $I$,

2016: together with the auxiliary functions

2017: used in its computation;

2018: the function $2^{-n}$ used for expressing approximations;

2019: and a function for computing bounded quantification over \NN.

2020: 

2021: This theorem has some interesting consequences,

2022: one of which we illustrate (Section 6, Theorem 4):

2023: 

2024: \Cor

2025: For each $n>0$,

2026: there is a finite universal algebraic specification,

2027: consisting of conditional equations and inequalities,

2028: for all 

2029: computable finite dimensional dynamical systems

2030: on the unit $n$-cube and over the unit time interval.

2031: \endpr

2032: 

2033: 

2034: Next we consider the converse problem:

2035: 

2036: \Problem\sl

2037: Find (reasonable) conditions under which

2038: algebraic definability implies abstract computability.

2039: \endpr

2040: 

2041: From Theorem C it follows that

2042: the converse to Theorem A is false,

2043: at least for specifications consisting of

2044: conditional equations and inequalities;

2045: for example, for the sine and cosine functions on the unit interval.

2046: 

2047: It is an open problem

2048: whether the 

2049: converse of the approximation

2050: result (Theorems B and C) holds.

2051: It seems that some extra {\it topological condition\/}

2052: such as continuity is required for a converse result.

2053: This suggests an

2054: interesting research area;

2055: see the example and discussion in Section 6.3.

2056: %Little seems to be known about this subject; for related work,

2057: %see \cite{gaertner-hotz}.

2058: 

2059: In the second part of the paper, we show how the

2060: conditional equational theories, and conditional BU equational

2061: theories, can be used with standard

2062: algebraic specification methods associated with proof

2063: systems, term rewriting and initial algebra semantics.

2064: 

2065: Now, when using the booleans, natural numbers and finite

2066: sequences, the algebraic specifications and their initial

2067: algebra semantics must define the corresponding standard

2068: models of the booleans, natural numbers and finite arrays. We

2069: develop extensions of the Birkhoff-Mal'cev Completeness

2070: Theorems that underlie the algebraic specification methods,

2071: designed to ensure that these sorts have standard models.

2072: Then we prove (Section 8, Theorems 4 and 5):

2073: 

2074: \Thmn{D \,(Initial algebra)}

2075: Given a signature \Sig\ and function type \ta\

2076: over \Sig, there exists a finite set of conditional equations

2077: \,\Ez\ \,(with a distinguished natural number variable \ztt)

2078: over a finite expansion \Sigp\ of \Sig,

2079: such that for any abstract program \al\ over \Sig, if \al\

2080: computes a total function $f$ on $A$ 

2081: of type \,\ta, 

2082: \,and $A$ has an initial

2083: algebra specification by a set $E$ of either conditional

2084: equations or conditional BU equations

2085: (with hidden sorts and functions), 

2086: then

2087: \,\Af\ has an initial algebra specification by a set \,$E\cup \Ekbar$,

2088: where \,\Ez\ \,consists of conditional equations and 

2089: \,\kbar\ \,is a numeral instantiating \,\ztt\ \,which is effectively

2090: calculable from \al.  

2091: The system \,\Ez\ \,is uniformly computable

2092: from \Sig\ and \ta.

2093: Furthermore, if the specification $E$ of $A$ has $e$ axioms,

2094: then the specification of \Af\ is finite, with $e + e'$

2095: axioms, where $e'$ is a constant 

2096: computed uniformly from \Sig\ and \ta.

2097: \endpr

2098: 

2099: This paper is part of our series on abstract computability

2100: theory on many-sorted algebras and its applications,

2101: starting in  \cite{tz:book} and most recently surveyed in \cite{tz:hb}.

2102: Knowledge of computation and our studies of computation

2103: versus specifications 

2104: \cite{tz:jlp,tz:ijfocs}

2105: and verification  

2106: \cite{tz:leeds}

2107: will be helpful, but only our work on topological data types

2108: \cite{tz:top}

2109: is necessary.

2110: 

2111: The subject of this paper is also a generalisation of the

2112: theory of algebraic specifications for {\it computable,

2113: semicomputable and co-semicomputable} algebras developed by

2114: one of us (JVT) with J.A. Bergstra: see

2115: \cite{bt80a,bt80b,bt82a,bt83b,bt87,bt95}

2116: and the surveys \cite{meseguer-goguen,stolt-jvt95}.

2117: However, at least

2118: initially, the generalised computability raises new

2119: questions concerning topological data types, uniformity and

2120: parameterisation, and standard models. Knowledge of the

2121: theory for computable algebras is not required for this

2122: paper.

2123: 

2124: In Section 1 we define how to augment structures with the

2125: {\it standard sorts} of the booleans and naturals, and

2126: finite sequences or {\it arrays} over all sorts, together

2127: with the corresponding operations. For the rest of the

2128: paper we consider, without loss of generality, only

2129: N-standard signatures and structures with the booleans and

2130: naturals.

2131: 

2132: In Section 2 we introduce a number of proof systems, all

2133: based in the calculus of sequents over a many-sorted signature \Sig. These

2134: are systems for ($i$) first order logic over \Sig\ with

2135: equality, ($ii$) conditional equational logic, ($iii$)

2136: conditional bounded universal (BU) equational logic, and

2137: ($iv$) conditional standard universal (SU) equational

2138: logic. The systems ($ii$) and ($iii$) are subsystems of the

2139: classical predicate calculus ($i$), and are used in the

2140: following sections, while ($iv$) is an infinitary system

2141: introduced for interest.

2142: 

2143: In Section 3 we define the basic technical notion of a

2144: theory uniquely specifying a function on an arbitrary algebra with

2145: hidden sorts and functions. This

2146: leads to a simple notion of specifiable parameterisation

2147: which we illustrate by showing how a conditional

2148: equational  (or conditional BU equational) specification of

2149: a standard structure $A$ can be extended to a similar

2150: specification of the array structure \Ax. We also show how

2151: to ``reduce" a conditional BU equational specification  over

2152: \Sig\ to a conditional equational specification over an

2153: expansion of \Sig.

2154: 

2155: In Section 4 we recall the basic notions of 

2156: {\it computability} of functions, 

2157: including universality of the \,\muPRx\ functions.

2158: 

2159: In Section 5 we prove Theorem A above,

2160: concerning the conditional equational definability of

2161: computable functions.

2162: 

2163: In Section 6 we prove Theorems B and C, 

2164: concerning the definability, by conditional equations and inequalities,

2165: of computably approximable functions on metric algebras.

2166: 

2167: In Section 7 we describe the construction of {\it initial

2168: standard models} for conditional equational and conditional

2169: BU equational theories, and work out the completeness

2170: theorems for the corresponding proof systems in Section 2.

2171: The reduction of a conditional BU equational specification

2172: over \Sig\ to a conditional equational specification over

2173: an expansion of \Sig\ is proved for initial models.

2174: 

2175: In Section  8  we investigate the relationship between {\it

2176: computability} and {\it algebraic specifiability} of

2177: functions on initial N-standard algebras, and prove Theorem D.

2178: Finally, in Section 9, we consider the converse problem of

2179: finding sufficient conditions for algebraic

2180: specifiability to imply computability on classes of

2181: standard structures. Two equivalence theorems are proved.

2182: 

2183: We wish to thank an anonymous referee for some

2184: very helpful comments.

2185: 

2186: 

2187: \Shead{1}{Many-sorted signatures and algebras}

2188: In this section

2189: we briefly review concepts defined and discussed in

2190: \cite[\S1]{tz:hb},

2191: where more detailed information can be found.

2192: Background information on universal algebra can be found in

2193: \cite{meinke-jvt,ehrig-mahr,wechler}.

2194: 

2195: \shead{1.1}{Basic definitions}

2196: A {\it signature} \Sig\ (for a many-sorted algebra)

2197: is a pair consisting of

2198: ($i$) a finite set \SortSig\ of {\it sorts},

2199: and ($ii$) a finite set \FuncSig\ of {\it (primitive)

2200: function symbols},

2201: each symbol $F$ having a {\it type}  \ $\tuptimes{s}{1}{m}\to s$,

2202: where $s_1 ,\ldots, s_m, s \in \SortSig$;

2203: in that case we write

2204: \ $F:\ \tuptimes{s}{1}{m}\to s$,

2205: with \,$\dom{F} \eqdf \tuptimes{s}1m$.

2206: \ (The case $m = 0$ corresponds to {\it constant symbols}.)

2207: 

2208: A 

2209: \Sig-{\it product type\/}

2210: has the form

2211: \,$u = \tuptimes{s}{1}{m}$ \,($m\ge0$),

2212: where \,\tup{s}{1}{m} \,are \Sig-sorts.

2213: We use the notation

2214: \,$u,v,w,\dots$ \,for \Sig-product types.

2215: 

2216: A \Sig-{\it algebra} $A$ has, for each sort $s$ of \Sig,

2217: a non-empty {\it carrier set} \,\As\

2218: \,of sort $s$,

2219: and for each \Sig-function symbol

2220: \,$F:\utos$,

2221: \,a function

2222: \,$\FA : \Au \to \As$

2223: \,(where, for the \Sig-product type \ $u  = \tuptimes{s}{1}{m}$,

2224: \,we write

2225: \,$\Au \ \eqdf \ A_{s_1} \times\dots\times A_{s_m}$).

2226: 

2227: Given an algebra $A$, we sometimes write \,$\Sig(A)$ \,for its signature.

2228: 

2229: The algebra $A$ is {\it total} if 

2230: \,\FA\ is total for each \Sig-function symbol $F$.

2231: Without such a totality assumption,

2232: $A$ is called {\it partial}.

2233: 

2234: In this paper we deal with total algebras,

2235: except in \S8.4.

2236: 

2237: We will also consider classes \KK\ of \Sig-algebras.

2238: A {\it \Sig-adt (abstract data type\/})

2239: \,is defined to be any such class, closed under \Sig-isomorphism.

2240: In particular, \AlgSig\ denotes the class of all \Sig-algebras.

2241: 

2242: \Examples

2243: ($a$)

2244: \,The algebra of {\it booleans\/}

2245: has the carrier \ $\BB = \{\ttt,\,\fff\}$

2246: \ of sort \,\bools. It can be displayed as follows:

2247: $$

2248: \boxed{

2249: \matrix \format\l&\quad\l\\

2250: \algebras &\BBB\\

2251: \carrierss &\BB\\

2252: \functionss&\ttt, \fff:\ \ \to\BB,\\

2253: &\ands^\BBB, \ors^\BBB:\BB^2\to\BB\\

2254: &\nots^\BBB: \BB\to\BB\\

2255: \ends&

2256: \endmatrix

2257: }

2258: \quad

2259: \tx{with signature}

2260: \quad

2261: \boxed{

2262: \matrix \format\l&\quad\l\\

2263: \signatures &\Sig(\BBB)\\

2264: \sortss &\bools\\

2265: \functionss&\trues, \falses:\ \ \to\bools,\\

2266: &\ands, \ors:\bools^2\to\bools\\

2267: &\nots: \bools\to\bools\\

2268: \ends&

2269: \endmatrix

2270: }

2271: $$

2272: For notational simplicity, we will usually not distinguish between

2273: function names in the signature (\trues, etc.)

2274: and their intended interpretations ($\trues^\BBB = \ttt$, etc.)

2275: \sn

2276: ($b$) \,The algebra \,\NNNo\ \,of naturals

2277: has a carrier \,\NN\ \,of sort \,\nats,

2278: \,together with the zero constant and successor function:

2279: $$

2280: \boxed{

2281: \matrix \format\l&\quad\l\\

2282: \algebras &\NNNo \\

2283: \carrierss &\NN\\

2284: \functionss&0:\ \ \to\NN,\\

2285: &\Ss:\NN\to\NN\\

2286: \ends&

2287: \endmatrix

2288: }

2289: $$

2290: 

2291: ($c$) \,The ring \,\RRRo\ \,of reals

2292: has a carrier \RR\ of sort \reals:

2293: $$

2294: \boxed{

2295: \matrix \format\l&\quad\l\\

2296: \algebras &\RRRo \\

2297: \carrierss &\RR\\

2298: \functionss&0,1:\ \ \to\RR,\\

2299: &+,\times:\RR^2\to\RR,\\

2300: &-:\RR\to\RR\\

2301: \ends&

2302: \endmatrix

2303: }

2304: $$

2305: \endpr

2306: 

2307: We make the following assumption about 

2308: the signatures \Sig.

2309: 

2310: \pr{Instantiation Assumption}

2311: {\sl For every sort $s$ of \Sig, there is a closed term of that sort,

2312: called the default term \,\delbs\ \,of that sort.}

2313: \endpr

2314: 

2315: This guarantees the presence of {\it default values} \,\delbsA\ \, 

2316: in a \Sig-algebra $A$ at all sorts $s$, 

2317: and {\it default tuples} \,\delbuA\ \,at all product types $u$.

2318: 

2319: \sheads{1.2}{Some definitions}

2320: \Defn{1 \,(Subalgebra)}

2321: Given \Sig-algebras $A$ and $B$,

2322: we say that

2323: $B$ is a \Sig-{\it subalgebra\/} of $A$

2324: \,(written $B\subalg A$)

2325: \,iff \,($i$) for all \Sig-sorts $s$, \,$B_s\sseq A_s$, 

2326: \,and ($ii$) for every \Sig-function symbol $F$,

2327: \,$F^B = \FA\rest B$.

2328: \endpr

2329: 

2330: 

2331: \Defn{2 \,(Expansions and reducts)}

2332: Let \Sig\ and \Sigp\ be signatures with $\Sig \subset \Sig'$.

2333: \sn

2334: \ ($a$) If $A'$ is a \Sigp-algebra, \,then the \Sig-{\it reduct of} $A'$,

2335: \ $A' \redSig$,

2336: \ is the algebra

2337: of signature \Sig, consisting of the carriers of $A'$

2338: named by the sorts of \Sig\ and equipped with the functions

2339: of $A'$ named by the function symbols of \Sig.

2340: \sn

2341: \ ($b$) If \,$A$ is a \,\Sig-algebra and $A'$ is a \Sigp-algebra,

2342: \,then $A'$ is a {\it \Sigp-expansion} of $A$ iff $A$ is the

2343: \Sig-reduct of $A'$.

2344: \sn

2345: ($c$) If \,\KKp\ \,is a \,\Sigp-adt,

2346: \,then \,$\KKp\redSig$ \,is the class of \Sig-reducts of algebras in \KKp.

2347: \endpr

2348: 

2349: \shead{1.3}{Adding booleans: \,Standard signatures and algebras}

2350: Recall the algebra \BBB\ of booleans (Example ($a$) in \S1.1).

2351: 

2352: A signature \Sig\ is called {\it standard} if

2353: %\itemm{($i$)}

2354: \,($i$) \,$\Sig(\BBB) \ \sseq \Sig$;

2355: %\itemm{($ii$)}

2356: \,($ii$) \,the \Sig-function symbols include

2357: a {\it conditional}

2358: $$

2359: \ifs_s :\bools\times s^2 \to s

2360: $$

2361: for all sorts $s$ of \Sig\ other than \bools; \,and

2362: %\itemm{($iii$)}

2363: \,($iii$) \,the \Sig-function symbols include

2364: an {\it equality  operation\/}

2365: $$

2366: \eqs_s: s^2 \to \bools

2367: $$

2368: for all $s\in\EqSortSig$, 

2369: \,where \,$\EqSortSig\sseq\SortSig$

2370: \,is the set of \Sig-{\it equality sorts\/}.

2371: %(We assume there is a ``computable equality" on these sorts.)

2372: 

2373: Given a standard signature \Sig,

2374: a \Sig-algebra $A$ is {\it standard\/} if

2375: %\itemm{($i$)}

2376: \,($i$) \,it is an expansion of \BBB;

2377: %\itemm{($ii$)}

2378: \,($ii$) \,the conditionals 

2379: have their standard interpretation in $A$,

2380: \ie, for $b\in\BB$ and $x,y \in \As$,

2381: $$

2382: \ifs_s(b,x,y) \ = \

2383: \cases

2384:    x \tif{$b = \ttt$}\\

2385:    y \tif{$b = \fff$};

2386: \endcases

2387: $$

2388: %\itemm{($iii$)}

2389: and \,($iii$)

2390: \,the equality operator $\eqs_s$ is interpreted as {\it identity} on each

2391: \Sig-equality sort $s$.

2392: 

2393: 

2394: Note that any many-sorted signature \Sig\

2395: can be {\it standardised} to a signature \,\SigBBB\

2396: \,by adjoining the sort \,\bools\ \,together

2397: with the standard boolean operations;

2398: and, correspondingly, any algebra $A$ can be standardised

2399: to an algebra \ABBB\

2400: \,by adjoining the algebra \BBB\

2401: and the conditional \,$\ifs_s$ \,at all \Sig-sorts $s$,

2402: and the equality operator \,$\eqs_s$

2403: \,at the specified equality sorts:

2404: $$

2405: \boxed{

2406: \matrix \format\l&\quad\l&\quad\l\\

2407: \algebras &\ABBB &\\

2408: \imports &A,\,\BBB&\\

2409: &\ifs_s:\BB\times A_s^2\to\As &(\sinSortSig),\\

2410: &\eqs_s: A_s^2\to\BB &(s\in\EqSortSig)\\

2411: \ends&&

2412: \endmatrix

2413: }

2414: $$

2415: Thus the standardisation of a \Sig-algebra $A$ depends on 

2416: the specification of \,\EqSortSig.

2417: These will be the sorts

2418: for which an equality test is considered to be ``computable"

2419: in some sense.

2420: 

2421: 

2422: \Examples

2423: ($a$) \,The simplest standard algebra is the algebra \BBB\ of the booleans.

2424: \smskipn

2425: ($b$) \,The standard algebra of naturals \NNN\

2426: is formed by standardising the algebra \NNNo\ 

2427: (Example ($b$) in \S1.1)

2428: with \,\nats\ \,as an equality sort,

2429: and, further, adjoining the order relation

2430: \lsnat\ as a boolean-valued operation on \NN:

2431: $$

2432: \boxed{

2433: \matrix \format\l&\quad\l\\

2434: \algebras &\NNN \\

2435: \imports &\NNNo, \,\BBB\\

2436: \functionss

2437: &\ifnat:\BB\times\NN^2\to\NN,\\

2438: &\eqnat,\,\lsnat:\NN^2\to\BB\\

2439: \ends&

2440: \endmatrix

2441: }

2442: $$

2443: ($c$) \,The standard algebra \,\RRR\ \,of reals

2444: is formed similarly by standardising the ring \RRRo\

2445: (Example ($c$) in \S1.1), with \,\reals\ \,{\it not\/} an equality sort.

2446: In fact, neither the equality nor the order relation on \RR\

2447: is included as an operation 

2448: on \reals. 

2449: (The significance of this is discussed later; \cf\ Remark 3 in \S5.3.)

2450: 

2451: 

2452: \StdAlgSig\ \,denotes the class of all standard \Sig-algebras.

2453: 

2454: \shead{1.4}{Adding counters: \ N-standard signatures and algebras}

2455: A standard signature \Sig\ is called {\it N-standard\/}

2456: if it includes (as well as \bools)

2457: the {\it numerical sort\/} \nats,

2458: and also function symbols for the

2459: {\it standard operations\/} of {\it zero\/}

2460: and {\it successor\/}, 

2461: as well as

2462: the {\it conditional\/} and {\it equality\/} and {\it order\/}

2463: on the naturals:

2464: $$

2465: \align

2466: 0\: \ \ &\to\nats\\

2467: \Ss\:\nats&\to\nats\\

2468: \ifnat\:\bools\times\nats&\to\nats\\

2469: \eqnat\:\nats^2&\to\bools\\

2470: \lsnat\:\nats^2&\to\bools.

2471: \endalign

2472: $$

2473: The corresponding \Sig-algebra $A$ is {\it N-standard} if the carrier

2474: $A_\natss$ is the set of natural numbers

2475: \ \NN = \{0,1,2,\dots\},

2476: and the standard operations (listed above) have their

2477: {\it standard interpretations\/} on \NN.

2478: \endpr

2479: 

2480: Note that any standard signature \Sig\ can be

2481: N-{\it standardised} to a signature \SigN\ 

2482: by adjoining the sort \nats\ and the operations 0, \Ss,

2483: \eqnat, \lsnat\ and \ifnat.

2484: Correspondingly,

2485: any standard \Sig-algebra $A$ can be N-{\it standardised\/} 

2486: to an algebra \AN\ 

2487: by adjoining the carrier \NN\

2488: together with the corresponding standard functions:

2489: $$

2490: \boxed{

2491: \matrix \format\l&\quad\l\\

2492: \algebras &\AN \\

2493: \imports &A,\,\NNN\\

2494: \ends&

2495: \endmatrix

2496: }

2497: $$

2498: 

2499: 

2500: \Examples

2501: ($a$) \,The simplest N-standard algebra is \NNN\ (Example ($b$) in \S1.3).

2502: \sn

2503: ($b$) \,The N-standard algebra \,\RRRN\ \,of reals is

2504: formed by N-standardising the standard real algebra \,\RRR\

2505: \,(Example ($c$) in \S1.3).

2506: \endpr

2507: 

2508: \NStdAlgSig\ \,denotes the class of all N-standard \Sig-algebras.

2509: 

2510: \pr{N-standardness Assumption}\sl

2511: We will assume throughout this paper

2512: that the signatures and algebras are N-standard,

2513: except where stated otherwise.

2514: \endpr

2515: 

2516: \mn

2517: We also consider a notion stricter than N-standardness.

2518: 

2519: \shead{1.5}{Strictly N-standard signatures and algebras}

2520: An N-standard signature \Sig\ is

2521: {\it strictly N-standard} 

2522: if the {\it only operations of \Sig\/}

2523: with range sort \nats\ or \bools\

2524: are the {\it standard numerical operations\/}

2525: \,$0,\,\Ss, \,\ifnat\,\eqnat, \,\lsnat$ (\S1.4)

2526: \,and the {\it boolean operations\/}

2527: \,$\trues,\,\falses, \,\ands,\,\ors,\,\nots$

2528: \,(\S1.1).

2529: An algebra is {\it strictly N-standard} if its signature is.

2530: 

2531: \Remarks

2532: \itemm{(1)}

2533: Any N-standardised signature and algebra

2534: are automatically strictly N-standard.

2535: \itemm{(2)}

2536: A strictly N-standard signature

2537: has no equality sorts other than \nats.

2538: \itemm{(3)}

2539: Any subterm of a term of sort \,\nats\ \,or \,\bools\

2540: of a strictly standard signature is itself of sort \,\nats\ \,or \,\bools.

2541: (Proved by structural induction on the term.)

2542: \endpr

2543: 

2544: 

2545: The notion of strict N-standardness will be used in Section 9.

2546: 

2547: \shead{1.6}{Adding arrays: \ Algebras \Ax\ of signature \Sigx}

2548: The significance of arrays for computation

2549: is that they provide

2550: {\it finite but unbounded memory}.

2551: 

2552: 

2553: Given a standard signature \Sig,

2554: and standard \Sig-algebra $A$, we expand \Sig\ and $A$

2555: in two stages:

2556: 

2557: \nin($1^\circ$) N-standardise these to form \SigN\ and \AN, as in \S1.3.

2558: 

2559: \nin($2^\circ$)

2560: Define, for each sort $s$ of \Sig, the carrier \Asx\

2561: to be the set of {\it finite sequences\/} or {\it arrays\/} \ax\

2562: over \As, of ``starred sort" \sx.

2563: 

2564: The reason for introducing starred sorts

2565: is the lack of effective coding

2566: of finite sequences within abstract algebras in general.

2567: (Note that, for simplicity, our definition excludes a starred sort \,\natsx,

2568: \,which would be redundant.)

2569: 

2570: 

2571: The resulting algebras \Ax\ have signature \Sigx,

2572: which expands \SigN\

2573: by including,

2574: for each sort $s$ of \Sig, the new starred sort \sx,

2575: and also the following new function symbols:

2576: 

2577: \nin($i$) the operator 

2578: \ $\Lgths_s: \,s^*\to \nats$, \ 

2579: where \,$\Lgths(\ax)$ \,is the length of the array \ax;

2580: 

2581: \smskipn

2582: ($ii$)  the application operator \ $\Aps_s: \,\sx\times\nats \to s$,

2583: \ where

2584: $$

2585: \Aps_s^A  (\ax, k ) \

2586: \ \cases

2587:      \ax  [ k ] \tif{$k<\Lgths(\ax)$}\\

2588:      \delbs \ow

2589:   \endcases

2590: $$

2591: where \,\delbs\ \,is the default value at sort $s$ 

2592: guaranteed by the Instantiation Assumption (\S1.1)\fn

2593: {We assume that $\ax[k]$ is undefined for $k\ge\tx{\ssf Lgth}(\ax)$.};

2594: 

2595: \smskipn

2596: ($iii$)  the null array \ $\Nulls_s: \sx$ \ of zero length;

2597: 

2598: \smskipn

2599: ($iv$)  the operator \ $\Updates_s: \,\sx\times\nats\times s \to s^*$, 

2600: \ where

2601: \ $\Updates_s^A  (\ax,n ,x )$ \ is the array \ $\bx\in\Asx$  

2602: of length $\Lgths(\bx) = \Lgths(\ax)$, \,such that

2603: for all $k< \Lgths(\ax)$

2604: $$

2605:    \bx  [k] \ = \

2606:       \cases

2607:          \ax [k]        \tif{$k \neq n$} \\

2608:             x           \tif{$k = n$}

2609:       \endcases

2610: $$

2611: ($v$)  the operator \ $\Newlengths_s: \ \sx\times\nats \to s^*$,

2612: \ where  \ $\Newlengths_s^A   (\ax, m )$  \ is the array \bx\ of length $m$ 

2613: such that for all $k<m$,

2614: $$

2615: \bx  [k] \ = \

2616:     \cases

2617:        \ax [k]        \ift{$k < \Lgths(\ax)$} \\

2618:        \delbs         \ift{$\Lgths(\ax) \le k <m$}

2619:     \endcases

2620: $$

2621: ($vi$) the {\it conditional\/} on \Asx\

2622: for each sort $s$; \ and

2623: \smskipn

2624: ($vii$) the {\it equality} operator on \Asx\ for each equality sort $s$.

2625: 

2626: \nin

2627: Note that

2628: \Ax\ is an N-standard \Sigx-{\it expansion} of $A$.

2629: 

2630: The justification for ($vii$) is that if a sort $s$ has

2631: ``computable" equality, then clearly so has the sort \sx,

2632: since it amounts to testing equality of finitely

2633: many pairs of objects of sort $s$, up to a computable length.

2634: 

2635: 

2636: 

2637: \Shead2{Proof systems and theories for \Sig-algebras}

2638: To reason about computations,

2639: we choose a first-order language based on \Sig\ as a specification

2640: language.

2641: 

2642: Note, in this connection, that the {\it operations\/} in \Sig\

2643: are used for  computation.

2644: In particular, boolean-valued operations

2645: are used for tests in computations.

2646: By contrast, for specification and reasoning about these algebras,

2647: we may add {\it predicates\/} to the language, which are

2648: not, in general, computable or testable.

2649: For example, our specification language 

2650: will include the equality predicate

2651: at all sorts (as we will see),

2652: whereas only the {\it equality sorts\/} 

2653: $s$ have the ``computable" equality operator \,$\eqs_s$

2654: (\S1.3).

2655: In writing specifications on the reals we may also add the `$<$' predicate

2656: (again, not computable, at least if defined totally),

2657: as we will do later (\S5.3) for the specification of

2658: approximable computability.  Note that 

2659: these predicates added to the language

2660: do not form part of the signature.

2661: Intuitively, think of the equality operation as

2662: a ``computable" boolean test, but the equality predicate

2663: as a ``provable" assertion of equality between two terms.

2664: \endpr

2665: 

2666: So let $\LangSig$ be the first order language

2667: over the signature \Sig, with the {\it equality predicate\/} at all sorts.

2668: The syntax of \LangSig\

2669: is generated as follows.

2670: For each \Sig-sort $s$

2671: there are countably many {\it variables\/}

2672: of sort $s$, denoted 

2673: \,$\att,\,\btt, \,\dots, \,\xtt,\,\ytt,\,\dots$\,.

2674: Next, for each \Sig-sort $s$,

2675: there are {\it terms\/} of sort $s$,

2676: generated from variables and the function

2677: symbols of \Sig\ according to

2678: the standard typing rules.

2679: %These are generated by the rules

2680: %$$

2681: %t \ ::= \ \xtt^s \mid F(\tup{t}{1}{n})

2682: %$$

2683: %where \,$s,\tup{s}{1}{n}$ \,are \Sig-sorts,

2684: %\ $F: \tup{s}{1}{n}\to s$ \ is a \Sig-function symbol

2685: %and \ $t_i\in \Term_{s_i}$ \,for \,$i=1,\dots,n$ \,($n\ge0$).

2686: We write \,$t^s$ \,or \,$t:s$ \,if $t$ is a term of sort $s$,

2687: and, for a product type $u = \tuptimes{s}1m$,

2688: we write \,$t:u$ \,if $t$ is a $u$-tuple of terms,

2689: \ie, an $m$-tuple of terms of type \,\tuptimes{s}1m.

2690: 

2691: The {\it atomic formulae} of \,\LangSig, \,then, are

2692: equations 

2693: \,($t_1^s = t_2^s$)

2694: \,between terms of sort $s$,

2695: for all \Sig-sorts $s$ (whether equality sorts or not),

2696: and the propositional constants \,\trues\ \,and \,\falses.

2697: {\it Formulae\/} of \LangSig\ are built up from these

2698: by the logical connectives

2699: \,$\land,\;\lor,\;\to,\;\neg$,

2700: \,and the quantifiers \,$\all_s$ and $\ex_s$\,

2701: \ for all sorts $s$ of \Sig.

2702: 

2703: \endpr

2704: 

2705: We will consider (in the following four subsections)

2706: four formal systems in \LangSig, 

2707: conveniently formulated as {\it sequent calculi\/}.

2708: The first is our basic system \FOLSig,

2709: full first order logic with equality over \Sig.

2710: The next two are subsystems of this,

2711: which will be used in Section 7.

2712: The final system is a subsystem of \FOLSig,

2713: extended by an infinitary proof rule.

2714: 

2715: Background information on sequent calculus proof systems

2716: can be found in \cite{tak}.

2717: 

2718: Note that we do {\it not\/} assume (N-)standardness of \Sig\ 

2719: in subsections 2.1 and 2.2 (only) below.

2720: 

2721: \shead{2.1}{\FOLSig: \ Full first order logic with equality over \Sig}

2722: This can be formulated in a system

2723: \,\LKeSig,

2724: \,which is an adaptation to the many-sorted signature \Sig\

2725: of the systems \LK\ and \LKe\ of \cite{gentzen,tak}.

2726: The atomic formulae are equations at all \Sig-sorts.

2727: 

2728: A {\it sequent} of \LKeSig\ is a construct of the form \ \GamDel,

2729: \ where \Gam\ and \Del\ are each finite sequences of formulae of \LangSig.

2730: 

2731: {\it Derivations\/} (of sequents) are then constructed

2732: from certain specified {\it initial sequents\/} (``axioms")

2733: by means of specified {\it inference rules\/}.

2734: 

2735: The system \LKe\ can be augmented in two ways:

2736: \sn

2737: ($a$) Adding {\it axioms of a theory},

2738: or rather all substitution instances of these,

2739: as initial sequents;

2740: \sn

2741: ($b$) Adding {\it induction\/} for a class \CCC\ of formulae

2742: (in case \Sig\ is N-standard),

2743: in the form of the inference rule

2744: $$

2745: \CCCIndSig:

2746: \quad{

2747: \GamDel, F(0)\qquad F(\att), \Pi \seqt \Lam, F(\Ss \att)

2748: \over

2749: \Gam,\Pi \seqt \Del,\Lam, F(t)}

2750: $$

2751: where the induction variable \att\ has sort \nats,

2752: and the induction formula $F(\att)$ belongs to the class \CCC.

2753: We write \,\IndSig\ \,for full \Sig-induction,

2754: \ie, where \CCC\ is the set of all first-order \Sig-formulae.

2755: 

2756: We will also be interested in the ``intuitionistic"

2757: version \,\CCCIndi\ \,of \,\CCCInd,

2758: in which the sequences \Del\ and \Lam\ above are empty.

2759: 

2760: Analogous augmentations can be made for the other systems considered below.

2761: 

2762: In the next three subsections we will consider 

2763: three further systems, the first two of which are subsystems of \FOLSig\

2764: and the third of which is a subsystem of \FOLSig\

2765: augmented by an infinitary \om-rule.

2766: These subsystems are,

2767: in fact, also subsystems of \LJeSig,

2768: which is an adaptation to \Sig\ of the ``intuitionistic" system \LJe\

2769: (\loccit),

2770: in which the sequents have only one formula on the rhs.

2771: (When we are working with these subsystems,

2772: the scheme \,\CCCInd\ \,will 

2773: consist of intuitionistic sequents, so that

2774: the sequences \Del\ and \Lam\ above are empty.)

2775: 

2776: 

2777: \shead{2.2}{\CondEqSig: \ conditional equational logic over \Sig}

2778: A {\it conditional equation} is a formula of the form

2779: $$

2780: P_1 \con \dots \con P_n \to P

2781: \tagx

2782: $$

2783: where $n \ge 0$ and $P_i$ and $P$ are equations.

2784: A {\it conditional equational theory}

2785: is a set of such formulae (or their universal closures).

2786: An {\it equational sequent} is a sequent of the form

2787: $$

2788: P_1, \,\dots, \,P_n \ \seqt \ P

2789: $$

2790: where $n \ge 0$ and $P_i$ and $P$ are equations.

2791: This sequent {\it corresponds to} the conditional equation ($*$).

2792: 

2793: The {\it initial sequents\/} are all substitution instances of 

2794: the \Sig-equality axioms

2795: (expressing that equality is a congruence relation with respect to \Sig),

2796: and the {\it inferences\/} are

2797: {\it structural inferences\/},

2798: {\it atomic cuts\/}

2799: and 

2800: {\it substitution\/} of terms for free variables in sequents.

2801: 

2802: \shead{2.3}{\CondBUEqSig: \ Conditional BU equational logic over \Sig}

2803: A {\it BU (bounded universal) quantifier\/} is a quantifier

2804: of the form `$\all \ztt < t$',

2805: where \,$\ztt:\nats$ \,and \,$t:\nats$.

2806: (The most elegant approach is to think of this 

2807: as a primitive construct, with its own

2808: introduction rule: see below.)

2809: A (\Sig-){\it BU equation} is formed by prefixing

2810: an equation by a string of 0 or more

2811: bounded universal quantifiers.

2812: A {\it conditional BU equation} is a formula of the form

2813: $$

2814: Q_1 \con \dots \con Q_n \to Q

2815: \tagxx

2816: $$

2817: \ where $n \ge 0$ and $Q_i$ and $Q$ are BU equations.

2818: A {\it conditional BU equational theory}

2819: is a set of such formulae (or their universal closures).

2820: A {\it BU equational sequent} is a sequent of the form

2821: $$

2822: Q_1, \,\dots, \,Q_n \ \seqt \ Q

2823: $$

2824: where $n \ge 0$ and $Q_i$ and $Q$ are BU equations.

2825: This sequent {\it corresponds to} the conditional BU equation

2826: ($**$).

2827: 

2828: The system \,\CondBUEqSig\

2829: consists of {\it BU equational sequents\/}.

2830: The initial sequents are the \Sig-equality axioms,

2831: as before, plus the {\it boundedness axioms}

2832: \TOL

2833: $$

2834: P(0),\dots,P(\ol{n - 1}) \ \seqt \ \all \ztt < \kbar P(\ztt)

2835: \tag"\BddAxSig:"

2836: $$

2837: \TOR

2838: for all \Sig-equations $P$

2839: and all \,\ninNN,

2840: \,where \nbar\ is the {\it numeral\/} for $n$,

2841: \ie, the term \,$\Ss\dots\Ss 0$ \,($n$ times `\Ss').

2842: The only inferences are {\it structural inferences},

2843: {\it cut\/}, {\it substitution\/},

2844: and the rules for the BU quantifiers:

2845: $$

2846: \all_b L :

2847: \ {{\Gam \seqt s < t \qquad Q(s),\Del \seqt Q}

2848:   \over

2849:   {\all \ztt<t Q(\ztt), \Gam, \Del \seqt Q}}

2850: \qqquad

2851: \all_b R:

2852: \ {{\att<t, \Gam \seqt Q(\att)}

2853:   \over

2854:   {\Gam \seqt \all \ztt<t Q(\ztt)}}

2855: $$

2856: where $s$ and $t$ are terms of sort \nats,

2857: \ `$s<t$' \ stands for \ `$\lsnat(s,t) = \trues$',

2858: \ and the variable \,$\att:\nats$

2859: \,is the `{\it eigenvariable}' of the inference $\all_b R$,

2860: which does not occur in the conclusion of that inference.

2861: 

2862: \Remarkn{\,(Boundedness axioms)}

2863: The boundedness axioms \,\BddAxSig\ \,hold (of course)

2864: in N-standard models.

2865: We remark here that they are derivable in \,\FOLSig\ 

2866: \,from the {\it N-standardness axioms\/}

2867: \,\NStdAxoSig\

2868: (a set of conditional equations

2869: defined in \S7.2),

2870: plus the single formula

2871: $$

2872: \ \ztt_1<\Ss \ztt_2 \ \longimp\  \ztt_1 < \ztt_2 \,\dis \,\ztt_1 = \ztt_2

2873: $$

2874: which is, however, {\it not a conditional

2875: BU equation\/}.

2876: This formula

2877: is derivable, in turn, in

2878: \,$\FOLSig +\QF\tx{-}\IndSig$

2879: \,(induction for quantifier-free formulae),

2880: from \,\NStdAxoSig.

2881: It is not clear whether the boundedness axioms

2882: are derivable in conditional BU equational logic alone

2883: from \,\NStdAxoSig,

2884: \,which is why we are adding them as axioms.

2885: \endpr

2886: 

2887: \shead{2.4}{\CondSUEqomSig: \ Conditional SU equational logic over \Sig}

2888: The final two systems that interest us,

2889: in this and the next subsection, 

2890: are not subsystems of \LKe,

2891: but {\it infinitary} systems.

2892: They will be used

2893: for another illustration of a Malcev-type theorem

2894: for N-standard algebras

2895: (see Section 6, Theorem 4).

2896: However they will not be used in the investigation

2897: of the relationship between computability 

2898: and algebraic specifiability in Section 8.

2899: 

2900: A (\Sig-){\it SU (standard universal) equation\/} is

2901: formed by prefixing

2902: an equation by a string of 0 or more

2903: universal quantifiers of sort \nats.

2904: A {\it conditional SU equation\/} is a formula of the form

2905: $$

2906: R_1 \con \dots \con R_n \to R

2907: \tagxxx

2908: $$

2909: where $n \ge 0$ and $R_i$ and $R$ are SU equations.

2910: A {\it conditional SU equational theory\/}

2911: is a set of such formulae (or their universal closures).

2912: An {\it SU equational sequent\/} is a sequent of the form

2913: $$

2914: R_1, \,\dots, \,R_n \ \seqt \ R

2915: $$

2916: where $n \ge 0$ and $R_i$ and $R$ are SU equations.

2917: This sequent {\it corresponds to} the conditional SU equation

2918: (\xxx).

2919: 

2920: The system \,\CondSUEqomSig\

2921: contains SU equational sequents.

2922: It contains the equality axioms and the following inferences:

2923: the structural inferences, cut, and the following

2924: rules for the universal number quantifier

2925: (where \,$t:\nats$):

2926: $$

2927: \all L :

2928: \ {{R(t),\Gam \seqt R}

2929:   \over

2930:   {\all \ztt R(\ztt), \Gam \seqt R}}

2931: \qqquad

2932: \allomR:

2933: \ {{\dots \ \Gam \seqt R(\bar n) \ \dots \quad (\tx{all} \ n \in \NN)}

2934:   \over

2935:   {\Gam \seqt \all \ztt R(\ztt)}}

2936: \tagx

2937: $$

2938: Note that the rule \,\allomR\ \,is actually an {\it infinitary \om-rule}.

2939: 

2940: 

2941: \shead{2.5}{\FOLomSig: \ full first-order logic with equality 

2942: and an \om-rule over \Sig}

2943: This modifies the system \,\FOLSig\ \,(\S2.1)

2944: by replacing the usual universal number quantifier rule \,\allR\

2945: \,by the infinitary rule \,\allomR\ \,(\S2.4),

2946: with also the corresponding rule \,\exomL\

2947: \,dually.

2948: We omit details, except to point out that

2949: \,$\FOL + \IndSig$

2950: can easily be interpreted in it.

2951: 

2952: 

2953: We write \,\EqSig, \,\BUEqSig\ \,and \,\SUEqSig\

2954: \,for the classes of equations, BU equations and SU equations (respectively)

2955: over \Sig.

2956: 

2957: 

2958: \shead{2.6}{Conservativity lemmas}

2959: One reason for the importance of

2960: (finite or infinite) conditional equational logic lies in the following

2961: lemmas.

2962: First we need a definition

2963: which will be 

2964: given again in context in Section 7.

2965: Let \,\FFF\ \,be a formal system

2966: (typically \,\CondEqSig\ \,or \,\CondEqomSig),

2967: and let $T$ be a theory over \Sig\

2968: (typically a conditional equational or

2969: \om-conditional equational theory). 

2970: We say that

2971: $T$ {\it determines \nats} in \FFF\

2972: if every closed term of sort \nats\ is, provably in \,\FFF\ \,from $T$,

2973: equal to a numeral.

2974: 

2975: \nin

2976: (1) \,(\FOL\ \,over \,\CondEq.)

2977: \ Let $E$ be a \Sig-conditional equational theory,

2978: \,and let \,\GamP\ \,be a \Sig-equational sequent.

2979: Then \,\GamP\ \,is provable from $E$ in \,\FOLSig\

2980: \,if, and only if, it is provable from $E$ in \,\CondEqSig.

2981: \endpr

2982: 

2983: \nin

2984: (2) \,($\FOL+\Ind$ \,over \,\CondEq.)

2985: \ Let $E$ be a \Sig-conditional equational theory

2986: \,which determines \,\nats\ \,in \,\CondEqSig,

2987: \,and let \,\GamP\ \,be a {\it closed\/} \Sig-equational sequent.

2988: Then \,\GamP\ \,is provable from $E$ in \,$\FOLSig+\IndSig$

2989: \,if, and only if, it is provable from $E$ in \,\CondEqSig.

2990: \endpr

2991: 

2992: \nin

2993: (3) \,($\FOL+\Ind$ \,over \,\CondBUEq.)

2994: \ Let $F$ be a \Sig-conditional BU equational theory

2995: \,which determines \,\nats\ \,in \,\CondBUEqSig,

2996: \,and let \,\GamQ\ \,be a {\it closed\/} \Sig-BU equational sequent.

2997: Then \,\GamQ\ \,is provable from $F$ in \,$\FOLSig+\IndSig$

2998: \,if, and only if, it is provable from $F$ in \,\CondBUEqSig.

2999: \endpr

3000: 

3001: \nin

3002: (4) \,(\FOLom\ \,over \,\CondSUEqom.)

3003: \ Let $G$ be a \Sig-conditional SU equational theory over \Sig\

3004: \,which determines \,\nats\ \,in \,\CondSUEqomSig,

3005: \,and let \,\GamR\ \,be a {\it closed\/} \Sig-conditional SU

3006: equational sequent.

3007: Then \,\GamR\ \,is provable from $G$ in \,\FOLomSig\

3008: \,if, and only if, it is provable from $G$ in \,\CondSUEqomSig.

3009: 

3010: 

3011: All four lemmas can be proved

3012: by cut elimination.

3013: We omit proofs, except to note briefly that the

3014: two conditions, that $E$ determines \,\nats\ \,

3015: and that \,\GamP\ \,is closed,

3016: are used in (2) and (3) to eliminate induction inferences,

3017: and in (4) to eliminate cuts of formulae universally or existentially

3018: quantified over \,\nats.

3019: 

3020: \Remarks

3021: (1) \,These conservativity lemmas (at least for simple equations)

3022: also follow from the Birkhoff-Mal'cev-type completeness theorems 1--4

3023: in Section 7.

3024: \sn

3025: (2) \,Infinitary systems come into their own

3026: when reasoning about infinite objects such as 

3027: infinite streams of data.

3028: Some applications in this direction,

3029: using a related infinitary system (\CondEqom),

3030: are given in \cite{tz:fef}.

3031: \endpr

3032: 

3033: 

3034: \Shead3{Specifiability of functions by theories}

3035: \sshead{3.1}{Specifiability over algebras and over classes of algebras}

3036: Recall from Section 2 that \,\LangSig\ \,is the first order 

3037: language over \Sig, with equality as the only predicate at all sorts.

3038: 

3039: A \Sig-{\it theory} is just a set $T$ of formulae in \LangSig.

3040: The {\it axioms} of $T$ are the formulae in $T$.

3041: We will be particularly interested in theories $T$

3042: satisfying certain syntactic conditions;

3043: for example, $T$ might be a set of conditional equations.

3044: This is considered more carefully in Section 7.

3045: 

3046: We are also interested (when \,\Sig\ \,is N-standard)

3047: in classes \KK\

3048: of the {\it N-standard models} of such \Sig-theories:

3049: \,$\KK = \NStdAlgSigT \subseteq \NStdAlgSig$.

3050: In this case we say also that \,\SigT\ \,is an

3051: (N-standard) {\it specification} for 

3052: the adt \,\KK.

3053: 

3054: {\it Assume, for the rest of this section, that \,\Sig,

3055: \Sigp\ and \Sigpp\ are N-standard signatures with 

3056: $\Sig \subset \Sig' \subset \Sig''$.

3057: Also, $A$ is an N-standard \Sig-algebra and 

3058: $A'$ is an N-standard \Sigp-algebra.

3059: Also, $T$ is a \Sig-theory,

3060: $T'$ is a \Sigp-theory and $T''$ is a \Sigpp-theory.}

3061: 

3062: Note that

3063: any expansion of a standard algebra

3064: is also standard, and

3065: any expansion of an N-standard algebra

3066: is also N-standard.

3067: 

3068: \Defn1

3069: Let $A_1'$ and $A_2'$ be two \Sigp-algebras with

3070: \,$A_1'\redSig = A_2'\redSig$.

3071: Then $A_1'$ and $A_2'$ are 

3072: {\it \Sigp/\Sig-isomorphic},

3073: written 

3074: \,$A_1' \cong_{\Sigp/\Sig} A_2'$,

3075: \ if there is a \Sigp-isomorphism from $A_1'$ to $A_2'$

3076: whose restriction to \Sig\ is the identity on $A_1'\redSig$.

3077: \endpr

3078: 

3079: \Defn2

3080: Suppose $A'$ is a \Sigp-expansion of $A$.

3081: We say that \,\SigpTp\ \,{\it specifies $A'$ over $A$\/} \,iff 

3082: $A'$ is the unique (up to \Sigp/\Sig-isomorphism) 

3083: \Sigp-expansion of $A$

3084: satisfying $T'$;

3085: \ in other words:

3086: \smskipn 

3087: ($i$) $A' \ttstile T'$; \ and 

3088: \smskipn 

3089: ($ii$) for all \Sigp-expansions $B'$ of $A$,

3090: \ if \,$B' \ttstile T'$ \, then \,$B' \cong_{\Sigp/\Sig} A'$.

3091: \endpr

3092: 

3093: We will occasionally write: \,``$T'$ specifies $A'$ over $A$"

3094: \,instead of \,``\SigpTp\ \,specifies $A'$ over $A$".

3095: 

3096: An important special case of Definition 2 is the following.

3097: 

3098: \Defn{${\bk 2}^{\bk f}$}

3099: Suppose \,$\Sig' = \Sig\cup\curly{\fs}$.

3100: We say that \,$\SigpTp$ \,{\it specifies $f$ over $A$\/} \,iff

3101: $f$ is the unique 

3102: (up to \Sigp/\Sig-isomorphism) 

3103: function on $A$ (of the type of \fs\,)

3104: such that \,$\Af\ttstile T'$.

3105: \endpr

3106: 

3107: 

3108: \Defn3

3109: Suppose $A'$ is a \Sigp-expansion of $A$.

3110: We say that \,\SigppTpp\ \,{\it specifies $A'$ over $A$} 

3111: {\it with  hidden sorts and/or functions} 

3112: \,iff $A'$ is the unique (up to \Sigp/\Sig-isomorphism) 

3113: \Sigp-expansion of $A$

3114: such that some \Sigpp-expansion of $A'$ satisfies $T''$;

3115: in other words:

3116: \itemm{($i$)}

3117: $A'$ is a \Sigp-reduct of a \Sigpp-model of $T''$; \ and

3118: \itemm{($ii$)}

3119: for all \Sigp-expansions $B'$ of $A$,

3120: \ if $B'$ is a \Sigp-reduct of a standard \Sigpp-model of $T''$,

3121: then \,$B' \cong_{\Sigp/\Sig} A'$.

3122: \endpr

3123: 

3124: Again, an

3125: important special case:

3126: 

3127: \Defn{${\bk 3}^{\bk f}$}

3128: Suppose \,$\Sig' = \Sig\cup\curly{\fs}$.

3129: We say that \,$\SigppTpp$ \,{\it specifies $f$ over $A$\/}

3130: {\it with  hidden sorts and/or functions} \,iff

3131: $f$ is the unique function on $A$ (of the type of \fs\,)

3132: such that some \Sigpp-expansion of \Af\ satisfies $T''$.

3133: \endpr

3134: 

3135: 

3136: \Defn4

3137: An operator \ $\Phi : \NStdAlgSig \to \NStdAlg(\Sig')$ 

3138: \ is {\it expanding (over \Sig)}

3139: iff for all N-standard \Sig-algebras $A$, 

3140: \,$\Phi(A)$ \,is a \Sigp-expansion of $A$,

3141: \ie, $\Phi(A)\redSig = A$.

3142: \endpr

3143: 

3144: \Example

3145: The array construction \,$A \mapsto \Ax$ 

3146: \,is an expanding operator.

3147: \endpr

3148: 

3149: {\it Assume further, for the rest of this section, that

3150: \ $\Phi : \NStdAlgSig \to \NStdAlg(\Sig')$ 

3151: \, is an expanding operator over \Sig,

3152: and that \,$\KK\sseq\NStdAlgSig$.}

3153: 

3154: \Notation

3155: (1) We will write \,\APhi\ \,for \,$\Phi(A)$.

3156: 

3157: \itemm{(2)}

3158: We write \KKPhi\ for

3159: (the closure w.r.t\. \Sigp-isomorphism of) the class

3160: $\{ \APhi \mid A \in \KK \} \subseteq \NStdAlg(\Sig')$.

3161: \endpr

3162: 

3163: \Defn5

3164: \itemm{($a$)\ }

3165: \SigpTp\ {\it specifies \Ph\ uniformly over} \KK\ \,iff 

3166: for all \AinKK,

3167: \SigpTp\ specifies \APhi\ over $A$.

3168: \itemm{($b$)\ }

3169: \SigpTp\ {\it specifies \Ph\ uniformly over} \Sig\ \,iff 

3170: \,\SigpTp\ specifies \Ph\ uniformly over 

3171: \nl

3172: \NStdAlgSig.

3173: \endpr

3174: 

3175: 

3176: 

3177: \Propn1

3178: Suppose \SigpTp\ specifies \Ph\ uniformly over \KK.

3179: \sn

3180: ($i$) \ \ For \,\AinKK, 

3181: \ $A \ttstile T \ \ \llongtofrom \ \ \APhi \ttstile T + T'$.

3182: \sn

3183: ($ii$) \ If \ $\KK = \NStdAlgSigT$, \,then 

3184: \ $\KKPhi = \NStdAlg(\Sig', \,T + T')$.

3185: \endpr

3186: 

3187: 

3188: \newpage

3189: 

3190: \Defn6

3191: \itemm{($a$)\ }

3192: \SigppTpp\ {\it specifies \Ph\ uniformly over \KK\ with 

3193: hidden sorts and/or functions}

3194: \,iff \,for all \AinKK,

3195: \,\SigppTpp\ \,specifies \APhi\ over $A$ with hidden sorts and/or functions.

3196: \itemm{($b$)\ }

3197: \SigppTpp\ {\it specifies \Ph\ uniformly over \Sig\ 

3198: with hidden sorts and/or functions} \,iff

3199: \,\SigppTpp\ \,specifies \Ph\ uniformly over \NStdAlgSig\ 

3200: with hidden sorts and/or 

3201: functions.

3202: \endpr

3203: 

3204: \Propn2

3205: Suppose \,\SigppTpp\ \,specifies \Ph\ uniformly over \Sig\

3206: with hidden sorts and/or functions.

3207: \sn

3208: ($i$) \ \ \ $A \ttstile T \ \llongtofrom 

3209: \ \tx{\APhi\ is a \Sigp-retract of a \Sigpp-model of \,$T+T''$}$.

3210: \sn

3211: ($ii$) \ If \ $\KK = \NStdAlgSigT$, \,then 

3212: \ $\KKPhi = \bigl(\NStdAlg(\Sig'', \,T + T'')\bigr) \,|\,_{\Sig'}$.

3213: \endpr

3214: 

3215: Interesting special cases of the above notions,

3216: in which the theories

3217: $T$, $T'$ and $T''$ are subject to certain

3218: syntactic conditions, are considered below (\S3.3) and in Section 7.

3219: First we give an important example of a specification

3220: of an expanding operator.

3221: 

3222: We write {\it conditional equational specification\/}

3223: and {\it conditional BU equational specification\/} for specifications

3224: in which the formulae are all conditional equations

3225: and conditional BU equations, respectively.

3226: 

3227: 

3228: \shead{3.2}{Conditional BU equational specification of the array construction}

3229: Let \,\ArrAxSig\,\  be the following set of axioms in $A$

3230: (dropping sort subscripts):

3231: $$

3232: \boxed{

3233: \gathered

3234: \Lgths(\Nulls) = 0, \\

3235: \lsnat(\ztt,\Lgths(\att)) = \falses \ \ \to\ \ \Aps(\att,\ztt) = \,\delb, \\

3236: \Lgths(\Updates(\att,\ztt,\xtt)) = \Lgths(\att), \\

3237: \eqnat(\ztt,\ztt_0) = \falses \ \ \to\ \ \Aps(\Updates(\att,\ztt_0,\xtt),\ztt) 

3238:   = \Aps(\att,\ztt), \\

3239: \lsnat(\ztt, \Lgths(\att)) = \trues \ 

3240:   \ \to \ \ \Aps(\Updates(\att,\ztt,\xtt),\ztt) = \xtt, \\

3241: \Lgths(\Newlengths(\att,\ztt)) = \ztt, \\

3242: \lsnat(\ztt,\ztt_1) = \trues \ \ \to\ \ \Aps(\Newlengths(\att,\ztt_1),\ztt) 

3243:   = \Aps(\att,\ztt),\\

3244: \Lgths(\att_1) = \Lgths(\att_2) \con 

3245: \all \ztt<\Lgths(\att_1) \bigl[\Aps(\att_1,\ztt) 

3246: = \Aps(\att_2,\ztt)\bigr]

3247: \ \ \to\ \ \att_1 = \att_2.

3248: \endgathered

3249: }

3250: $$

3251: %where `\defaults' is the default term of the appropriate sort.

3252: The last axiom relates equality on \sx\

3253: to equality on $s$, for all equality sorts $s$

3254: except \nats\ \,(since there is no starred sort \,\natsx, \,as 

3255: explained in \S1.6).

3256: 

3257: Note that all the axioms of \,\ArrAxSig\,\ are conditional equations,

3258: except for the last one, which is a conditional BU equation!

3259: 

3260: \Thmn1

3261: The specification \,$(\Sigx,\,\ArrAxSig)$ \,specifies the array construction

3262: \ $A \mapsto \Ax$ \ uniformly over \Sig.

3263: \endpr

3264: 

3265: \n{\bf{Proof \,(outline):}}

3266: %\,The proof, in outline, proceeds as follows.

3267: \,Given an N-standard \Sig-algebra $A$, and a \Sig-sort $s$,

3268: the axioms for `\Nulls', `\Newlengths' and `\Updates'

3269: guarantee that {\it at least\/} all the ``standard" arrays over \As\

3270: are present (or can be ``constructed").

3271: On the other hand, the axiom for array equality

3272: guarantees that there are no ``non-standard" arrays,

3273: \ie, no elements of \Axs\ other than these.

3274: \endpf

3275: 

3276: This array specification will be considered again, 

3277: from the viewpoints of specification of \muPRx\ computations

3278: (\S5.2),

3279: and initial algebra specifications (\S8.2).

3280: 

3281: \sheads{3.3}{Reducing conditional BU to 

3282: conditional equational specifications}

3283: \Thmn{2 \,(BU elimination)}

3284: Let \,$\Sig\sset\Sigp$,

3285: \,let $A'$ be a \Sigp-expansion of $A$, 

3286: \,and let $F$ be a conditional BU equational \Sigp-theory

3287: which specifies $A'$ over $A$.

3288: Then there is an expansion \Sigpp\ of \Sigp\ by function symbols,

3289: and a conditional equational \Sigpp-theory $E$ which specifies 

3290: $A'$ over $A$, with hidden functions.

3291: If $F$ contains $q$ occurrences of BU quantifiers,

3292: then \Sigpp\ expands \Sigp\ by $q$ new function symbols.

3293: Moreover, if

3294: $F$ is finite, with $e$ axioms (say),

3295: then so is $E$, with $e+4q$ axioms.

3296: \endpr

3297: 

3298: \Pf

3299: The idea is to incorporate in the signature,

3300: for each BU quantifier occurring in $F$,

3301: a {\it characteristic function}

3302: for that quantifier,

3303: or (expressed differently) a function which computes that quantifier.

3304: Consider (for notational simplicity)

3305: the case of an equation with a single BU quantifier

3306: $$

3307: \all \ztt < s(\xtt) \,\bigl[ t_1(\ztt,\xtt) = t_2(\ztt,\xtt)\bigr].

3308: \tagx

3309: $$

3310: with \,$\xtt:u$.

3311: (In the general case, we ``eliminate" the quantifiers

3312: successively, from the inside out.)

3313: We adjoin, for each such BU quantifier ($*$) occurring in $F$,

3314: a boolean-valued function symbol 

3315: $$

3316: \fs:\ \nats\times u \ \to \ \bools

3317: $$

3318: intended to satisfy in $A$ 

3319: $$

3320: \fs(n,x) = \trues\ \ \ \llongtofrom\ 

3321: \ \all z<n \bigl[t_1(z,x)=t_2(z,x)\bigr].

3322: $$

3323: for all \,\ninNN, \,\xinAu.

3324: This interpretation is imposed on \fs\ by

3325: {\it adjoining\/} to $F$ the following axioms giving the

3326: inductive definition for \fs:

3327: $$

3328: \gathered

3329: \fs(0,\xtt) = \trues\\

3330: \fs(\ztt,\xtt) = \trues\ \con t_1(\ztt,\xtt) = t_2(\ztt,\xtt) \ \ \to

3331:   \ \ \fs(\Ss \ztt,\xtt) = \trues\\

3332: \fs(\Ss \ztt,\xtt) = \trues \ \ \to\ \ \fs(\ztt,\xtt) = \trues\\

3333: \fs(\Ss \ztt,\xtt) = \trues \ \ \to\ \ t_1(\ztt,\xtt) = t_2(\ztt,x)

3334: \endgathered

3335: \tagxxx

3336: $$

3337: and {\it replacing} ($*$) in $F$ by 

3338: $$

3339: \fs(s(\xtt),\xtt) = \trues.

3340: \tagxxxx

3341: $$

3342: In this way we replace $F$ by a conditional

3343: equational \Sigpp-theory $E$, with the stated properties.

3344: \endpf

3345: 

3346: %\Remarks

3347: %(1) \,The number of axioms in $E$, 

3348: %may be much greater than that for $F$.

3349: %Thus we see here a tradeoff between the {\it cardinality}

3350: %of a specification and its {\it syntactic complexity}.

3351: %\sn

3352: Note that if $F$ contains infinitely many occurences of BU quantifiers,

3353: then \Sigpp\ contains, correspondingly, infinitely many new function symbols, 

3354: which is (strictly speaking) not allowed by our definition of signature,

3355: although it is harmless enough here.

3356: 

3357: We will return to this topic in the context

3358: of initial algebra specifications in \S7.7.

3359: \endpr

3360: 

3361: 

3362: 

3363: \Shead4{Computable functions}

3364: In this section we consider various notions of 

3365: computability over abstract algebras.

3366: (An equivalent approach, using an 

3367: imperative model of programming featuring the \qwhiles\ construct,

3368: was developed in \cite{tz:book,tz:hb}

3369: where the equivalence of these two approaches are explained.)

3370: In \S4.1 two computability classes are introduced.

3371: In \S4.2 two more classes are formed by adjoining 

3372: the \muu\ operator to these.

3373: 

3374: \shead{4.1}{\PRSigb\ and \PRxSigb\ computable functions}

3375: Given an N-standard signature \Sig,

3376: we define \PR\  {\it schemes\/} over \Sig\

3377: which generalise the schemes for primitive recursive functions 

3378: over \NN\ in \cite{kleene:im}.

3379: They

3380: define (total) functions $f$ either outright

3381: (as in the base cases ($i$)---($ii$) below)

3382: or from other functions ($g,\dots,$ $h,\dots$) 

3383: (as in the inductive cases ($iii$)---($v$)) as follows:

3384: 

3385: 

3386: \itemm{\bi(a)}

3387: {\bi Basic schemes: \,Initial functions\/}

3388: \itemm{($i$)}

3389: {\it Primitive \Sig-functions:}

3390: %corresponding to the primitive function and constant symbols of \Sig:

3391: $$

3392: \align

3393: f(x) \ &= \ F(x)\\

3394: f(x) \ &= \ c

3395: \endalign

3396: $$

3397: of type \,\utos,

3398: \,for all the primitive function symbols \,$F\:\utos$ 

3399: \,and constant symbols $c$ of \Sig, 

3400: where $x:u$. 

3401: %and $y:s$, for any \Sig-sort $s$.

3402: \itemm{($ii$)}

3403: {\it Projection:}

3404: $$

3405:       f (x) \ = \ x_i

3406: $$

3407: of type \,$u\to s_i$,  \,where \,$x = (\tup{x}1m)$ \,is of type

3408: \,$u = \tuptimes{s}1m$.

3409: 

3410: \itemm{\bi(b)}

3411: {\bi Inductive schemes:\/}

3412: \itemm{($iii$)}

3413: {\it Composition:}

3414: $$

3415:       f (x) \ = \ h ( g_1 (x) ,\ldots, g_m (x) )

3416: $$

3417: of type \utos,

3418: where \,$g_i\:u\to s_i$ ($i=1,\dots,m$) \,and 

3419: \,$h\:\tuptimes{s}1m\to s$.

3420: 

3421: \itemm{($iv$)}

3422: {\it Definition by cases:}

3423: $$

3424: f (b, x, y) \ = \

3425: \cases

3426: x \tif{b = \ttt} \\

3427: y \tif{b = \fff}

3428: \endcases

3429: $$

3430: of type \,$\bools\times s^2 \to s$.

3431: 

3432: \itemm{($v$)}

3433: {\it Simultaneous primitive recursion on \NN\/}:

3434: \,This defines, on each \AinNStdAlgSig,

3435: for fixed $m > 0$ (the degree of simultaneity),

3436: $n \geq 0$ (the number of parameters),

3437: and product types $u$

3438: and $v = \tuptimes{s}1m$,

3439: an $m$-tuple of functions

3440: \,$f = ( f_1 ,\ldots, f_m )$ \,with

3441: \,$f_i:\nats \times u \to s_i$, \,such that

3442: for all  \,\xinAu\

3443: \,and  $i = 1 ,\ldots, m$,

3444: $$

3445: \align

3446: f_i (0,x ) & \ = \ g_i (x) 	\\

3447: f_i ( z+1, x ) & \ = \

3448: h_i ( z,x, f_1 (z,x) ,\ldots, f_m (z,x) )

3449: \endalign

3450: $$

3451: where \,$g_i\:u\to s_i$ \,and \,$h_i\:\nats\times u \times v \to s_1$

3452: \,($i=1,\dots,m$). 

3453: 

3454: \n

3455: Note that

3456: the last scheme 

3457: uses the N-standardness of the algebras,

3458: \ie\ the carrier \NN.

3459: 

3460: For details of the syntax and semantics of PR schemes,

3461: see \cite[\S4.1.5]{tz:book},

3462: from which it can be seen that

3463: a scheme for a function

3464: contains (hereditarily) the schemes for

3465: all the auxiliary functions used to define it.

3466: 

3467: In the context of algebraic specification theory,

3468: it often turns out to be more convenient to work with

3469: PR {\it derivations\/} instead of PR schemes.

3470: A PR derivation is, roughly, a ``linear version"

3471: of a PR scheme, in which all the auxiliary functions

3472: are displayed in a list.  More precisely:

3473: 

3474: 

3475: \Defn{\,(PR derivation)}

3476: A {\it \PRSig\ derivation\/} \al\

3477: is a list of pairs 

3478: $$

3479: \al\ = \ ((f_0,\sig_0),\ (f_1,\sig_1),\ \dots \ (f_n,\sig_n))

3480: \tagx

3481: $$

3482: of functions (actually function symbols) $f_i$ and PR schemes $\sig_i$

3483: ($i=1,\dots,n$)

3484: where for each $i$, either $f_i$ is an initial function,

3485: or $f_i$ is defined by $\al_i$ from

3486: functions $f_j$, for certain $j<i$.

3487: The derivation \al\ is said to be a {\it PR derivation of\/} $f_n$,

3488: with {\it auxiliary functions\/} \tup{f}0{n-1}.

3489: The {\it type\/} of \al\ is the type of $f_n$.

3490: \endpr

3491: 

3492: \Notation

3493: A \,\PRSigutos\ \,scheme (or derivation)

3494: is a \,\PRSig\ \,scheme (or derivation) of type \,\utos.

3495: \endpr

3496: 

3497: \Remarks

3498: (1) \,The formalism of \PRSig\ derivations

3499: is equivalent to that of \PRSig\ schemes:

3500: from a PR scheme we can derive an  equivalent PR derivation

3501: by ``linearising" the subschemes, and conversely,

3502: given the derivation ($*$), the scheme $\sig_n$

3503: is equivalent to it.

3504: Below, we will usually work with derivations.

3505: \sn

3506: (2) \,A \,\PRSigutos\ 

3507: \,derivation \,$\al\:\utos$

3508: \,defines, or rather computes, a

3509: function \,$\falA\:\Au\to\As$, or, more generally,

3510: a family of functions  

3511: \curly{\falA \mid \AinNStdAlgSig}

3512: uniformly over \NStdAlgSig.

3513: \sn

3514: (3) \,We assume a standard G\"odel numbering of \PRSig\ derivations,

3515: writing \cnr{\al} for the G\"odel number of derivation \al.

3516: \endpr

3517: 

3518: It turns out that a broader class

3519: of functions provides a better generalisation of the

3520: notion of primitive recursiveness, namely

3521: {\it \PRx\ computability}.

3522: A function on $A$ is \PRxSig\ computable

3523: if it is defined by a PR derivation over \Sigx,

3524: interpreted on \Ax\

3525: (\ie, using starred sorts for the auxiliary functions 

3526: used in its definition).

3527:  

3528: \shead{4.2}{\muPRSigb\ and \muPRxSigb\ \,computable functions}

3529: The \,\muPR\ {\it schemes\/} over \Sig\

3530: are formed by adding to the PR schemes of \S4.1

3531: the inductive scheme: 

3532: 

3533: \itemm{($vi$)}

3534: {\it Least number {\rm or} \muu\ operator:}

3535: $$

3536: f (x) \ \simeq \ \mu  z [ g ( x , z ) = \ttt ]

3537: $$

3538: of type \,$u\to\nats$,  \,where

3539: \,$g:u\times\nats\to\bools$

3540: \,is \muPR.

3541: Here \,$f (x) \da z$ \,if, and only if, \,$g (x , y ) \da \fff$ 

3542: \,for each $y < z$ and $g (x ,z ) \da \ttt$.

3543: 

3544: \n

3545: Note that this scheme also uses the N-standardness of the algebra.

3546: Also, \muPR\ computable functions are, in general, {\it partial}.

3547: We use the notation \,$f(x)\da y$ \,to mean that $f(x)$

3548: is defined and equal to $y$.

3549: The notation \,`$\simeq$'

3550: \,means that the two sides are either 

3551: both defined and equal,

3552: or both undefined. The schemes for composition and simultaneous

3553: primitive recursion are correspondingly re-interpreted

3554: to allow for partial functions.

3555: 

3556: These schemes generalise the schemes given in \cite{kleene:im}

3557: for partial recursive functions

3558: over \NN.

3559: 

3560: As before, we can define

3561: the concepts of

3562: \muPRSig\ {\it derivations\/} and \muPRSig\ {\it computability\/}.

3563: 

3564: Again, a broader class turns out to be more useful, namely

3565: {\it \muPRx\ computability}.

3566: This is just \PRx\ computability with \,\muu.

3567:  

3568: \Notation

3569: \PRA\ \,is the class of functions \,\PR\ \,computable

3570: on $A$,

3571: \,and

3572: \,\PRAutos\ \,is the subclass of \,\PRA\ \,of functions of type \,\utos.

3573: Similarly for \,\PRxA, \,\muPRA, \,etc.

3574: \endpr

3575: 

3576: There are many other models of computability, due to 

3577: Moschovakis, Friedman, Shepherdson and others, 

3578: which turn out to be equivalent to \muPRx\ computability:

3579: see \cite[\S7]{tz:hb}.

3580: All these equivalences have led to the postulation of a 

3581: {\it generalised Church-Turing Thesis

3582: for deterministic computation of functions},

3583: which 

3584: can be roughly formulated as follows:

3585: 

3586: {\displaytext

3587: Computability of functions on many-sorted algebras 

3588: by deterministic algorithms

3589: can be formalised by \,\muPRx\ computability.

3590: 

3591: }

3592: 

3593: \shead{4.3}{Equivalent imperative programming models of computation}

3594: In \cite{tz:hb} we investigate computation on many-sorted \Sig-algebras,

3595: using imperative programming models:

3596: \,\WhileSig, \,based on the \,\qwhiles\ \,loop construct over \Sig,

3597: \,\ForSig, \,based similarly on the \,\qfors\ \,loop,

3598: and \WhilexSig\ \,and \,\ForxSig,

3599: \,which use arrays, \ie, auxiliary variables of starred sort over \Sig.

3600: 

3601: Writing \WhileA\ \,for the class of functions \While-computable on $A$, etc.,

3602: we can list the equivalences

3603: between the ``schematic" 

3604: and ``imperative" computational models as follows.

3605: 

3606: \Thm

3607: \itemm{($i$)}

3608: $\PRA \ = \ \ForA$

3609: \itemm{($ii$)}

3610: $\PRxA \ = \ \ForxA$

3611: \itemm{($iii$)}

3612: $\muPRA \ = \ \WhileA$

3613: \itemm{($iv$)}

3614: $\muPRxA \ = \ \WhilexA$,

3615: 

3616: \n

3617: in all cases, uniformly for \,\AinNStdAlgSig.

3618: \endpr

3619: 

3620: These results are all stated in \cite{tz:hb}, and 

3621: can be proved by the methods of \cite{tz:book}.

3622: 

3623: 

3624: \newpage

3625: 

3626: \shead{4.4}{Universal Function Theorem for \muPRxb}

3627: The following is a uniform version of a theorem proved in

3628: \cite[\S4.9]{tz:hb}

3629: (using the equivalent formalism of \,\Whilex\ \,programs):

3630: 

3631: \Thm

3632: For any \Sig-function type \utos,

3633: there is a \,\muPRxSig\ \,derivation \,$\ups:\natxutos$

3634: \,which is

3635: universal for \,\muPRxSig\ \,derivations of type \,\utos.

3636: \endpr

3637: 

3638: In other words, we can enumerate all the \muPRx\ derivations 

3639: of type \utos:

3640: $$

3641: \al_0,\ \al_1,\ \al_2,\ \dots

3642: $$

3643: so that, putting

3644: $$

3645: \varphi^A_i \ \eqdf \ \fs^A_{\al_i}\: \Au\ \to\ \As

3646: $$

3647: and

3648: $$

3649: \UnivusA \ = \ \fs^A_\ups: \NN\times\Au\to\As

3650: $$

3651: we have

3652: $$

3653: \UnivusA(i,a) \ = \ \varphi^A_i(a)

3654: $$

3655: for all \,\AinNStdAlgSig\ \,and \,$i=0,1,2,\dots$.

3656: %which is \muPRx\ computable, uniformly in $A$ and

3657: %in the types $u,s$.

3658: 

3659: \Remarksn{ \,(Canonical forms of \,\muPRxb\ \,derivations)}

3660: (1) \,From the construction of the universal \,\muPRxSigutos\ \,derivation

3661: \ups\

3662: \cite[\S4]{tz:hb},

3663: it can be seen that

3664: \ups\ uses the \muu-operator exactly once.

3665: \sn

3666: (2) \,For any \,\muPRxSigutos\ \,derivation \al,

3667: the universal derivation \,$\ups\:\natxutos$

3668: \,provides an equivalent

3669: {\it canonical  {\rm or} normal form derivation\/} \,\alhat, such that

3670: \,$\falhatA \,= \,\falA$

3671: \,for all N-standard \Sig-algebras $A$.

3672: This canonical derivation is formed 

3673: in a simple way from \,\ups, \,essentially

3674: by substituting the G\"odel number \cnr{\al} of \al\ 

3675: for the distinguished \,\nats\ \,variable of \ups,

3676: \,so that for all N-standard $A$,

3677: $$

3678: \fs_{\alhat}^A \ = \ \varphi_{\cnr{\al}}^A \ = \ \fs_\al^A.

3679: $$

3680: This is, in fact, a generalisation to \,\NStdAlgSig\

3681: \,of Kleene's Normal Form Theorem for (essentially) \,\muPR(\NNN)

3682: \,\cite{kleene:im}.

3683: \sn

3684: (3) \,From the constructions in (1) and (2) it follows that \,\alhat\

3685: \,also uses the \muu-operator exactly once, 

3686: and in such a way that for any N-standard $A$,

3687: 

3688: {\displaytext

3689: \falhatA\ \,is total if, and only if,

3690: this application of the \muu-operator is total on $A$.

3691: 

3692: }

3693: 

3694: \Shead5

3695: {Algebraic specifications for computable functions}

3696: We will consider functions $f$ {\it computable\/} on a \Sig-algebra,

3697: by \,\PR\ \,and \,\muPRx\ \,derivations,

3698: and show that they are 

3699: algebraically {\it specifiable\/} by

3700: conditional equational, and conditional BU equational, theories.

3701: 

3702: We will also consider,

3703: in the context of {\it metric algebras\/}

3704: (\ie, algebras with metrics  such that 

3705: the functions in the signature are continuous)

3706: a broader class of functions than \muPRx\ computable,

3707: namely those functions

3708: {\it uniformly approximable\/} by \muPRx\ computable functions,

3709: and show that such functions

3710: are specifiable by {\it conditional equations and inequalities\/},

3711: which are conditional formulae built up from inequalities ($t_1<t_2$)

3712: on the reals

3713: as well as equations ($t_1=t_2$) between terms of the same sort.

3714: 

3715: \shead{5.1}{Algebraic specifications for \,\PR\

3716: \,computable functions} 

3717: Let \Sig\ be an N-standard signature.  

3718: For each \PRSig\ derivation \al, there is a finite

3719: set \Eal\ of ``specifying equations" for the function $f$,

3720: as well as the auxiliary functions

3721: \,$g=(\tup{g}1{k_\al})$,

3722: \,defined by \al.

3723: 

3724: The set \Eal\ consists of {\it equations\/}

3725: in an expanded signature \,$\Sigal = \Sig \cup \{ \gal, \fal \}$,

3726: \,where \,$\gal \ident \gs_{\al,1},\dots,\gs_{\al,k_\al}$.

3727: It is defined by course of values induction on the length of the 

3728: derivation \al,

3729: with cases ($i$)---($v$) (\S4.1)

3730: according to the last scheme in \al.

3731: In fact, \Eal\ is formed by adjoining, in each case,

3732: specifying equation(s) like those shown for that case in \S4.1.

3733: These are simple (\ie, not conditional) equations;

3734: for example, in the case ($iv$) {\it definition by cases\/},

3735: there are two equations:

3736: $$

3737: \align

3738: \fs(\trues,\,\xtt,\,\ytt) \ &= \ \xtt \\

3739: \fs(\falses,\,\xtt,\,\ytt) \ &= \ \ytt

3740: \endalign

3741: $$

3742: and in the case ($v$) {\it simultaneous primitive recursion\/},

3743: there are \,$2m$ \,equations (where $m$ is the degree of simultaneity):

3744: $$

3745: \align

3746: \fs_i (0,\,\xtt ) & \ = \ \gs_i (\xtt)      \\

3747: \fs_i ( \ztt+1, \,\xtt ) & \ = \

3748:         \hs_i ( \ztt,\,\xtt, \,\fs_1 (\ztt,\xtt) ,\ldots, \fs_m (\ztt,\xtt) )

3749: \endalign

3750: $$

3751: for \,$i=1,\dots,m$.

3752: 

3753: Thus we have an effective map 

3754: \ $\al \mapsto (\Sigal, \Eal)$

3755: \ from \PRSig\ derivations to (simple) equational specifications.

3756: 

3757: Now for each \PR\ derivation \al\ and N-standard \Sig-algebra $A$, let

3758: \,\falA\ \,be the function on $A$ computed by \al,

3759: and let \,\galA\ \,be the corresponding auxiliary functions on $A$.

3760: Consider the operators

3761: $$

3762: A \longmapsto \AfalA

3763: \tagx

3764: $$

3765: and

3766: $$

3767: A \ \longmapsto \ \AgalAfalA.

3768: \tagxx

3769: $$

3770: 

3771: Recall the definition of uniform specification of an operator 

3772: over a class of \Sig-algebras (\S3.1, Definitons 5 and 6).

3773: 

3774: \Thmn{1 \,(Equational specification of PR functions)}

3775: For each \PRSig\ derivation \al,

3776: the equational specification \,$(\Sigal, \Eal)$ 

3777: \, specifies the operator ($**$)

3778: uniformly over 

3779: $A \in \NStdAlgSig$.

3780: Hence it specifies the operator

3781: ($*$) uniformly over all N-standard \Sig-algebras $A$,

3782: with hidden functions.

3783: \endpr

3784: 

3785: \Pf

3786: By course of values induction on the length of \al.

3787: \endpf

3788: 

3789: In other words, the equations \Eal\ specify not only \falA,

3790: but also the auxiliary functions \galA, uniformly over 

3791: all N-standard \Sig-algebras $A$.

3792: 

3793: Similarly with

3794: \PRx\ computability:

3795: for a \PRxSig\ derivation \al, 

3796: let \Eal\ be the set of

3797: specifying equations for the function \,\ \fal\

3798: and the auxiliary functions \gal\

3799: defined by \al, 

3800: in the signature \,$\Sigalx = \Sigx \cup \{ \gal, \fal \}$.

3801: 

3802: \Cor

3803: For each \PRxSig\ derivation \al,

3804: the equational specification \,\SigalxEal\

3805: specifies the operator

3806: ($*$) uniformly over \Sig,

3807: with hidden sorts and functions.

3808: \endpr

3809: 

3810: 

3811: \shead{5.2}{Algebraic specifications for \muPRxb\ computable functions}

3812: We now consider \,\muPRxSig\ \,derivations \al.

3813: For each such derivation there is again

3814: a finite set \Fal\ of ``specifying equations"

3815: for the function $f$ defined by \al\

3816: and its auxiliary functions $g$.

3817: This set is constructed like \Eal\ (\S5.1),

3818: by structural induction on \al.

3819: Now, however,

3820: \Fal\ consists of 

3821: {\it conditional BU equations}

3822: in a signature \ $\Sigalx = \Sigx \cup \{ \gal, \fal \}$,

3823: \ because of the new case, \ie,

3824: scheme ($vi$) for the \muu-operator

3825: (\S4.2),

3826: which results in the addition to \Fal\ of the conditional BU equation

3827: \TOL

3828: $$

3829: \all \ztt < \ytt \,(\gs_0(\xtt,\,\ztt) \,= \,\falses) \ 

3830: \con \ (\gs_0(\xtt,\,\ytt) \,= \,\trues) 

3831: \ \ \longimp \ \ \fs(\xtt) = \ytt .

3832: \tag{\Fmu}

3833: $$

3834: \TOR

3835: 

3836: Again we have an effective map 

3837: \ $\al \mapsto \SigalxFal$

3838: \ from \,\muPRxSig\ \,derivations to conditional BU equational specifications.

3839: 

3840: Now there are complications in the theory, 

3841: since \muPRx\ computable functions are, in general, partial.

3842: We intend to study specification theory for partial algebras

3843: systematically in a future paper.

3844: Here we limit ourselves to 

3845: the case where the \muPRx\ computable function

3846: is, in fact, total.

3847: 

3848: As before, for a \muPRx\ derivation \al\

3849: and an N-standard \Sig-algebra $A$, let

3850: \,\falA\ \,be the function on $A$ defined by \al,

3851: and let \,\galA\ \,be corresponding auxiliary functions on \Ax.

3852: A further problem is that, even if \,\falA\ \,is total,

3853: the functions \,\galA\ \,might not be.

3854: We will now show that we can, without loss of generality, 

3855: restrict attention to the case that the

3856: \,\galA\ \,are also total.

3857: We accomplish this

3858: by the use of the uniform derivations

3859: provided by the Universal Function Theorem for \muPRx\ (\S4.4),

3860: as we now explain.

3861: 

3862: \Def

3863: A \muPRx\ derivation \al\ is {\it total on\/} $A$ iff 

3864: the auxiliary functions \galA, as well as \falA, are all total on \Ax.

3865: \endpr

3866: 

3867: 

3868: \pr{Totality Lemma}

3869: Given any \muPRxSig\ derivation \,$\al\:\utos$,

3870: \,we can effectively find a \muPRxSig\ derivation \,$\alhat\:\utos$

3871: \,such that for any N-standard \Sig-algebra $A$,

3872: \itemm{$(i)$}

3873: $\falhatA \,= \,\falA$;

3874: \itemm{$(ii)$}

3875: if \,\falA\ \,is total, then \alhat\ is total on $A$.

3876: \endpr

3877: 

3878: \Pf

3879: This follows from the Universal Function Theorem

3880: and the three remarks following it (\S4.4).

3881: \endpf

3882: 

3883: Now consider the operators ($*$) (\S5.1 above)

3884: and

3885: $$

3886: A \ \longmapsto \ (\Ax,\,\galhatA,\,\falhatA)

3887: \tagxxx

3888: $$

3889: where \alhat\ is constructed from \al\ as in the totality lemma.

3890: %(Recall that \,$\falhatA = \falA$ \,on all N-standard \Sig-structures $A$.)

3891: Let \,$\Sigalx = \Sigx \cup \curly{\galhat,\falhat}$.

3892: Recall the definition of 

3893: the array specification \,\ArrAxSig\ \,in \S3.2,

3894: and the definition of 

3895: the conditional BU specification

3896: \,\Falhat\ \,of \,\falhatA\ ($= \falA$).

3897: 

3898: 

3899: \Thmn{2 \,(Conditional BU equational specification of \muPRxb\ functions)}

3900: \nl

3901: For each 

3902: \muPRxSig\ derivation \al, let

3903: $$

3904: \Falx \ \eqdf \ \ArrAx(\Sig) \,+ \,\Falhat

3905: $$

3906: where \alhat\ is constructed from \al\ 

3907: as in the totality lemma.

3908: Then the conditional BU equational specification

3909: \,$(\Sigalx,\,\Falx)$\,\ specifies the operator (\xxx)

3910: in the following sense:

3911: for any $A$ on which \falA\ is total,

3912: $$

3913: (\Ax,\,\galhatA,\,\falhatA)\ \ttstile \ \Falx.

3914: $$

3915: Hence

3916: \,$(\Sigalx,\,\Falx)$

3917: \,specifies the operator $(*)$ 

3918: uniformly over all N-standard \Sig-algebras $A$

3919: \,on which \,\falA\ \,is total,

3920: with hidden sorts and functions.

3921: \endpr

3922: 

3923: \Pf

3924: As with Theorem 1, by course of values induction on the length of \al.

3925: \endpf

3926: 

3927: 

3928: Note that the specification given in Theorem 2 is uniform

3929: over all N-standard \Sig-algebras $A$

3930: on which \,\al\ \,is total.

3931: In fact, there is a stronger form of uniformity

3932: for \muPRx\ computability,

3933: following from the Universal Function Theorem for \muPRx.

3934: (Actually, this is already implicit in the construction of

3935: the derivation \alhat\ in the

3936: totality lemma, which is really a normal form lemma for \muPRx\ derivations.)

3937: 

3938: 

3939: \Thmn{3 \,(Universal conditional BU equational specification)}

3940: For each \Sig-function type \,\utos\

3941: \,we can effectively find a 

3942: signature \,\Sigxus\ \,which expands \Sigx\ 

3943: by function symbols, and a

3944: finite conditional BU equational specification

3945: \,$(\Sigxus,\,\FUus(\ztt))$

3946: \,which is 

3947: universal for specifications of total \,\muPRxSig-computable functions

3948: of that type,

3949: in the following sense:

3950: it contains a distinguished number variable \,\ztt\

3951: \,such that 

3952: for each \,\muPRxSig\ \,derivation \,$\al:\utos$,

3953: \,and each N-standard \Sig-algebra $A$,

3954: if \,\falA\ \,is total on $A$,

3955: then 

3956: \,$(\Sigxus,\,\FUus(\kbar))$,

3957: \,where \,$k=\cnr{\al}$,

3958: \,specifies \,\falA\ \,on $A$, with hidden sorts and functions.

3959: \endpr

3960: 

3961: (Here \,$\FUus(\kbar)$

3962: \,is the result of substituting the numeral \kbar\ for \ztt\ in 

3963: \,$\FUus(\ztt)$.)

3964: 

3965: Next, by the BU Reduction Theorem (Theorem 2 in Section 3),

3966: we derive as a corollary to Theorem 3:

3967: 

3968: \Thmn{4 \,(Universal conditional equational specification)}

3969: For each \Sig-function type \,\utos\

3970: \,we can effectively find a 

3971: signature \,\Sigxusp\ \,which expands \,\Sigxus\  

3972: \,(of Theorem 3)

3973: by function symbols, and a

3974: finite conditional specification

3975: \,$(\Sigxusp,\,\EUus(\ztt))$

3976: \,which is 

3977: universal for specifications of total \,\muPRxSig-computable functions

3978: of that type,

3979: in the following sense:

3980: it contains a distinguished number variable \,\ztt\

3981: \,such that 

3982: for each \,\muPRxSig\ \,derivation \,$\al:\utos$,

3983: \,and each N-standard \Sig-algebra $A$, 

3984: if \falA\ is total on $A$,

3985: \,then 

3986: \,$(\Sigxusp,\,\EUus(\kbar))$,

3987: \,where \,$k=\cnr{\al}$,

3988: \,specifies \,\falA\ \,on $A$, with hidden sorts and functions.

3989: \endpr

3990: 

3991: From the above uniformity theorems it follows trivially that 

3992: for a given \Sig-function type \,\utos\

3993: \,there is a uniform bound to the lengths of conditional

3994: BU \Sigx-specifications,

3995: {\it or\/} conditional

3996: equational \Sigx-specifications respectively,

3997: for total \,\muPRx-computable functions 

3998: on N-standard \Sig-algebras.

3999: \endpr

4000: 

4001: 

4002: %\Remarkn{3 \,(Herbrand-G\"odel computability on \NN)}

4003: %As a particular case of the above, with $A \NNN$

4004: %\,(\S1.3; 

4005: %and recall that \,\NNN\str\ can be coded in \NNN\ \cite{tz:hb}),

4006: %we derive the classical result that any

4007: %total \,\muPR\ \,computable function on \NNN\

4008: %can be specified by a finite set of conditional equations over \NN\

4009: %--- actually, simple equations in this special case ---

4010: %with hidden functions \cite[\S57]{kleene:im}.

4011: %(This can be strengthened to an equivalence:

4012: %\,see the Remark at the end of Section 8).

4013: %\endpr

4014: 

4015: 

4016: \Shead6{Algebraic specifications for computably approximable functions}

4017: We have shown that 

4018: $$

4019: \tx{\sl computability $\implies$ algebraic specifiability}

4020: $$

4021: where (for example) if ``computability" means \,\muPRx\

4022: \,(or, equivalently, \,\Whilex) \,compu\-tability,

4023: then ``algebraic specifiability" means specifiability

4024: by conditional BU equations.

4025: 

4026: It is natural to ask in what sense a converse holds.

4027: We will see (below) that a full converse to the above cannot be expected,

4028: since algebraic specifiability is more powerful,

4029: in some sense, than \muPRx-computability.

4030: (In Section 7 we will investigate partial converses.)

4031: We show here in fact that,

4032: on metric algebras,

4033: $$

4034: \tx{\sl computable approximability\ $\implies$ \ algebraic specifiability.}

4035: $$

4036: ``Computable approximability", to be defined shortly,

4037: is a strong extension of the notion

4038: of computability;

4039: while ``algebraic specifiability" will be (re-)defined 

4040: so as to permit the order relation (as well as equality) 

4041: between pairs of terms of sort \reals.

4042: 

4043: \shead{6.1}{Metric algebras}

4044: We refer to \cite{tz:top} and \cite[\S6]{tz:hb} for definitions of

4045: (total) {\it metric algebra\/} and related concepts.

4046: We review some definitions and results from these references.

4047: (Note that in these references the subject is discussed

4048: in the broader context of {\it partial algebras\/}.)

4049: 

4050: A {\it metric \Sig-algebra \Ad, based on a \Sig-algebra\/} $A$,

4051: is an algebra of the form

4052: $$

4053: \boxed{

4054: \matrix \format\l&\quad\l\\

4055: \algebras &\Ad \\

4056: \imports &A \\

4057: \carrierss & \RR \\

4058: \functionss&d_s:A_s^2\to\RR \quad (\sinSortSig)\\

4059: \ends&

4060: \endmatrix

4061: }

4062: $$

4063: where $d$ is a family \,\ang{d_s\mid \sinSortSig}

4064: \,of metrics $d_s$ on the carriers \As, where

4065: (in the case that $A$ is standard or N-standard)

4066: \,$d_\boolss$ \,and \,$d_\natss$ \,are the discrete metrics

4067: on \,\BB\ \,and \,\NN\ \,

4068: respectively,

4069: and such that the primitive functions on $A$

4070: are all continuous under these metrics.

4071: 

4072: We will often write `$d$' for the metric $d_s$, and

4073: `$A$' for the metric algebra \Ad.

4074: 

4075: \Examples

4076: ($a$)

4077: \,The metric algebra \,\RRRd\

4078: \,on the reals (``$d$" for ``distance") is defined by

4079: $$

4080: \boxed{

4081: \matrix \format\l&\quad\l\\

4082: \algebras &\RRRd \\

4083: \imports &\RRRN\\

4084: \functionss&\divN:\RR\times\NN\to\RR,\\

4085: &d_\realss:\RR^2\to\RR,\\

4086: &d_\natss:\NN^2\to\RR,\\

4087: &d_\boolss:\BB^2\to\RR\\

4088: \ends&

4089: \endmatrix

4090: }

4091: $$

4092: where \,\RRRN\ \,is the N-standard algebra of reals (\S1.4, Example ($b$)),

4093: \divN\ is division of reals by naturals

4094: (where division by zero is defined as zero),

4095: the metric on \,\RR\ \,is the standard one,

4096: and the metrics on \,\NN\ \,and \,\BB\ \,are discrete.

4097: 

4098: Note that \,\RRRd\ \,does not contain the (total) boolean-valued

4099: functions \eqreal\ or \lsreal,

4100: since they are not continuous with respect

4101: to this metric.

4102: %(\cf\ the partial functions \eqp\ and \lsp\ of \RRRp).

4103: %It is therefore not an expansion

4104: %of the standard algebra \RRR\ of reals (\S2.2, Example ($c$))

4105: %which contains \eqreal.

4106: %(Compare the N-standardisation \RRRN\ of \RRR\ (\S2.3, Example ($b$))

4107: %which does contain \eqreal).

4108: 

4109: \smskipn ($b$)

4110: \,The interval metric algebra \,\IIId:

4111: \ Here the unit interval

4112: \,$I = [0,1]$

4113: \,is included as a separate carrier of

4114: sort `\intvls', again with the usual metric.

4115: This is useful for studying real continuous functions

4116: with compact domain.

4117: (We could also choose \,$I = [-1,1]$, \,etc.)

4118: The algebra \,\IIId\ \,is defined by

4119: $$

4120: \boxed{

4121: \matrix \format\l&\quad\l\\

4122: \algebras &\IIId \\

4123: \imports &\RRRd\\

4124: \carrierss & I\\

4125: \functionss &\iI:I\to\RR,\\

4126: &d_\intvlss:I^2\to\RR\\

4127: \ends&

4128: \endmatrix

4129: }

4130: $$

4131: where \iI\ is the embedding of $I$ into \,\RR.

4132: Because of the importance of the metric algebra \IIId\ as

4133: in our computation theory, let us review 

4134: its construction.

4135: It contains \RR\ with its standard ring operations,

4136: \NN\ and \BB\ with their standard operations,

4137: functions for definition by cases on \RR, \NN\ and \BB,

4138: division of reals by naturals,

4139: the unit line interval \II\ and its embedding in \RR,

4140: and the standard metrics on all four carriers.

4141: \endpr

4142: 

4143: 

4144: \shead{6.2}{Definitions and theorems}

4145: Now let $A$ be an N-standard metric \Sig-algebra with metric $d$.

4146: 

4147: \Defn{1 \,(\muPRxb\ computably approximable functions)}

4148: A total function \,$f:\Au \to \As$ \,on $A$

4149: is \,\muPRx\ {\it computably approximable,

4150: uniformly on\/} $A$,

4151: if there is a total \,\muPRx\ computable function

4152: $$

4153: G:\NN\times \Au\ \ \to \ \As

4154: $$

4155: and a total computable function \,$g:\NN\to\NN$

4156: on $A$ such that, putting

4157: \,$G_n \eqdf G(n,\,\cdot\,)$,

4158: \,the sequence $G_n$ {\it approximates $f$ uniformly\/} on \Au\

4159: with modulus of approximation $g$,

4160: \ie, for all $n$, $k$ and all $x \in \Au$,

4161: $$

4162: k\ge g(n) \ \implies\ d(G_k(x),f(x)) \ < \ 2^{-n}.

4163: $$

4164: \endpr

4165: 

4166: 

4167: \Defn{2 \,(Fast \,\muPRxb\ computably approximable functions)}

4168: A total function \,$f:\Au \to \As$ \,on $A$

4169: is {\it fast \,\muPRx\ computably approximable,

4170: uniformly on\/} $A$,

4171: if there is a total \,\muPRx\ computable function

4172: \ $G:\NN\times \Au\ \,\to \,\As$

4173: \ on $A$ such that, putting

4174: \,$G_n \eqdf G(n,\,\cdot\,)$,

4175: \,the sequence $G_n$ {\it approximates $f$ uniformly fast\/} on \Au,

4176: \ie, for all $n$ and all $x \in \Au$,

4177: $$

4178: d(G_n(x),f(x)) \ < \ 2^{-n}.

4179: \tagx

4180: $$

4181: \Remarkn1

4182: It is easy to see that Definitions 1 and 2 are equivalent;

4183: for given a (computable) approximating sequence $G_n$

4184: with modulus of approximation $g$,

4185: we can effectively replace it by the fast (computable)

4186: approximating sequence \,$G_n' \eqdf G_n\circ g$.

4187: We will therefore usually tacitly assume w.l.o.g. that

4188: our approximating sequences are fast,

4189: and work with the (simpler) Definition 2.

4190: \endpr

4191: 

4192: \Defn{3 \,(Fast \muPRxb\ approximating derivations)}

4193: Let $A$ be a metric \Sig-algebra.

4194: A derivation 

4195: \,$\gam\:\natxutos$

4196: \,is an {\it approximating derivation for\/} 

4197: a total function

4198: \,$f\:\Au\to\As$ 

4199: \,if 

4200: \,($i$) \,the function \,$G:\NN\times\Au\to\As$ \,computed by \,\gam\ \,on $A$

4201: is total on $A$; and ($ii$) \,$G$ and $f$

4202: satisfy ($*$) above.

4203: \endpr

4204: Note that {\it at most one\/} function is

4205: \,\muPRx\ \,approximable by a given derivation on any metric algebra.

4206: 

4207: 

4208: \Defn{4 \,(Conditional equation or inequality)}

4209: \nl

4210: ($a$) \,A {\it conditional equation or inequality \/} is defined

4211: like a conditional equation, except that the atomic

4212: statements may be {\it either\/} equations ($t_1 = t_2$)

4213: between terms of the same sort,

4214: {\it or\/}

4215: order 

4216: ($t_1<t_2$)

4217: between terms of sort \reals.

4218: \sn

4219: ($b$) A {\it conditional BU equation or inequality\/} 

4220: is defined like a conditional equation, except that the atomic

4221: statements may be {\it either\/} equations ($t_1=t_2$)

4222: or BU equations 

4223: \linebreak

4224: ($\all \ztt<t\,[\,t_1=t_2\,]$)

4225: between terms of the same sort,

4226: %(where \ztt\ and $t$ have sort \nats),

4227: or inequalities

4228: ($t_1<t_2$)

4229: between terms of sort \reals.

4230: \endpr

4231: 

4232: \Remarkn2

4233: Here we are treating the order relation on the reals

4234: as a new atomic predicate of \,\LangSigx\

4235: \,(like equality), {\it not\/} as a boolean-valued operation

4236: $$

4237: \lsreal:\reals^2\to\bools.

4238: $$

4239: This predicate (unlike such an operation)

4240: does not form part of the signature \,\Sig.

4241: (See the analogous Remark concerning equality

4242: at the beginning of Section 2.)

4243: \endpr

4244: 

4245: Note also that ($*$) is a conditional inequality

4246: (actually a simple inequality, without an antecedent).

4247: 

4248: \Notation

4249: We write \,\muPRxApproxA\ \,for the class of \muPRx\ computably

4250: approximable functions on $A$, and

4251: \,\muPRxApproxAutos\ \,for those of type \,\utos.

4252: \endpr

4253: 

4254: In preparation for the next theorem,

4255: we note that 

4256: a ``Universal Function Theorem"

4257: holds for \,\muPRxApproxA,

4258: \,in the following sense. 

4259: For any \Sig-function type \utos, let

4260: $$

4261: H^\us\ \eqdf\ \Univ^A_\natxutoss:

4262: \ \NN\times\NN\times \Au\ \to\ \As

4263: $$

4264: be the universal function for

4265: \,$\muPRxA_\natxutoss$

4266: \,given by the Universal Function Theorem (\S4.4).

4267: Then for each \,$f\in\muPRxApproxAutos$,

4268: there is a number $k$ such that

4269: (writing \,$H^\us_{k,n} = H^\us(k,n,\,\cdot\,)$)

4270: \,the sequence of functions

4271: \,$H^\us_{k,0},\,H^\us_{k,1},\,H^\us_{k,2},\,\dots$

4272: \,uniformly approximates $f$.

4273: The number $k$ can be chosen as the G\"odel number of an

4274: approximating derivation for $f$,

4275: \ie, a derivation \,$\gam\:\natxutos$

4276: \,of the function \,$H^\us(k,\,\cdot\,)$. 

4277: Combining this with Theorem 3 of Section 5, we obtain:

4278: 

4279: 

4280: \Thmn{1 \,(Universal conditional BU specification 

4281: of \,\muPRxb\ approximable functions)}

4282: For each \Sig-function type \,\utos\

4283: \,we can effectively find a

4284: signature \,\Sigxus\ \,which expands \Sigx\

4285: by function symbols, and a

4286: finite conditional 

4287: BU specification \,$(\Sigxus,$ $\FVus(\ztt))$

4288: \,consisting of conditional BU equations and inequalities,

4289: \,which is

4290: universal for specifications of 

4291: \,\muPRxSig-computably approximable functions of that type,

4292: in the following sense:

4293: it contains a distinguished number variable \,\ztt\

4294: \,such that

4295: for each \,\muPRxSig\ \,derivation 

4296: \,$\gam\:\natxutos$

4297: \,and each metric \Sig-algebra $A$

4298: and total function \,$f\:\Au\to\As$,

4299: \,if \,\gam\ \,is an approximating derivation for $f$ on $A$,

4300: then

4301: \,$(\Sigxus,\,\FVus(\kbar))$,

4302: \,where \,$k=\cnr{\gam}$,

4303: \,specifies \,$f$ \,on $A$, with hidden sorts and functions.

4304: \endpr

4305: 

4306: 

4307: \Pf

4308: Define

4309: $$

4310: \FVus(\ztt) \ \eqdf\ \FUus(\ztt) + E_{\invexpss} + E_*(\ztt)

4311: $$

4312: where \,$\FUus(\ztt)$ \,is the conditional BU equational specification

4313: constructed as in Theorem 3 in Section 5 for 

4314: the universal function $H$ for \,\muPRx\ \,computable functions

4315: of type 

4316: \linebreak

4317: \,$\nats \times u \to s$,

4318: \,$E_{\invexpss}$ \,is the set of specifying equations for

4319: the computable real-valued function

4320: \,$\invexps(n) = 2^{-n}$,

4321: \,\ie, its recursive definition: 

4322: $$

4323: \invexps(0) \ = \ 1,\qquad \invexps(\Ss \ntt) \ = \ \divN(\invexps(\ntt),\,2),

4324: $$

4325: and \,$E_*(\ztt)$ \,is the inequality ($*$) above ---

4326: or rather, its formal version

4327: $$

4328: \ds(\Hs(\ztt, \,\ntt,\,\xtt), \,\fs(\xtt)) \ < \ \invexps(\ntt).

4329: \tagxx

4330: $$

4331: (Note that every \,\muPRx\ \,function \,$G\:\NN\times\Au\to\As$

4332: \,is obtainable from $H$ by substituting the G\"odel number

4333: of its derivation for the first argument \,\ztt\ \,of $H$.)

4334: Let \,\Sigxus\ \,be the signature formed by

4335: expanding \,\Sigx\ \,by symbols for \,\Hs\ \,and 

4336: \,\invexps, \,as well as the auxiliary functions 

4337: used in their computations.

4338: Then for any \muPRx\ derivation \,$\gam\:\natxutos$,

4339: \,metric \Sig-algebra $A$

4340: and function \,$f\:\NN\times\Au\to\As$,

4341: \,if \gam\ is an

4342: approximating derivation for $f$

4343: then \,$(\Sigxus,\,\FVus(\kbar))$

4344: \,(where \,$k=\cnr{\gam}$)

4345: \,is a conditional BU specification of $f$ on $A$,

4346: with hidden sorts and functions,

4347: consisting of conditional BU equations and inequalities

4348: \endpf

4349: 

4350: Now, by adapting the BU Reduction Theorem (\S3.3)

4351: to specifications with inequalities,

4352: we derive as a corollary to Theorem 1:

4353: 

4354: \newpage

4355: 

4356: \Thmn{2 \,(Universal conditional specification 

4357: of \,\muPRxb\ approximable functions)}

4358: For each \Sig-function type \,\utos\

4359: \,we can effectively find a

4360: signature \,\Sigxusp\ \,which expands \,\Sigxus\  

4361: \,(of Theorem 1)

4362: by function symbols, and a

4363: finite conditional 

4364: specification \,$(\Sigxusp,\,\EVus(\ztt))$,

4365: \,consisting of conditional equations and inequalities,

4366: \,which is

4367: universal for specifications of approximably

4368: \,\muPRxSig-computable functions of that type,

4369: in the following sense:

4370: it contains a distinguished number variable \,\ztt\

4371: \,such that

4372: for each \,\muPRxSig\ \,derivation 

4373: \,$\gam\:\natxutos$

4374: \,and each metric \Sig-algebra $A$

4375: and total function 

4376: \,$f\:\Au\to\As$,

4377: \,if \,\gam\ \,is an approximating derivation for $f$ on $A$,

4378: then

4379: \,$(\Sigxusp,\,\EVus(\kbar))$,

4380: \,where \,$k=\cnr{\gam}$,

4381: \,specifies \,$f$ \,on $A$, with hidden sorts and functions.

4382: \endpr

4383: 

4384: 

4385: \Remarkn{3 \,(Replacing  the order predicate by a boolean-valued operation)}

4386: The order relation in the above specification is used in one place only: 

4387: in the (conditional) relation ($*$) (or ($**$)).

4388: In fact (Remark 2 above notwithstanding)

4389: ($*$) {\it could\/} be interpreted as a conditional {\it equation\/}

4390: (so that $f$ is {\it conditionally equationally definable\/}

4391: with hidden sorts and functions)

4392: by interpreting \,`$t_1<t_2$' \,as \,`$\lsreal(t_1,t_2)=\trues$',

4393: \,where the boolean-valued operator

4394: $$

4395: \lsreal:\reals^2\to\bools

4396: $$

4397: is included in the signature of the metric algebra over \RR.

4398: The problem here is that (as discussed in \cite{tz:top})

4399: whereas all functions in the signature of metric algebras

4400: (and hence all functions computable over these)

4401: are {\it continuous\/}, the (total) function \,\lsreal\ \,is discontinuous.

4402: The only way to restore continuity is to consider

4403: a {\it partial\/} continuous \,\lsreal\ \,operator,

4404: which leads to a study of {\it topological partial algebras\/}.

4405: This can be done, and the whole of the present theory

4406: could be re-cast in such a context,

4407: but that would take us too far afield from the present study.

4408: \endpr

4409: 

4410: 

4411: Let us apply Theorem 2 to 

4412: the classical notion of

4413: {\it Gzegorczyk-Lacombe (GL) computability\/}

4414: on the  unit interval $I=[0,1]$.

4415: This includes all the well-know functions of real analysis

4416: (sin, exp, log, etc.) restricted to $I$.

4417: 

4418: \Notation

4419: We write

4420: \,\GLtmI\ for the class of GL-computable total functions 

4421: \,$f\:I^m\to \RR$.

4422: \endpr

4423: 

4424: Many concrete models of computation on \,\II\ 

4425: \,are equivalent to this class \cite{pourel-richards,weih:book}.

4426: It has been shown that (recall the definition of \IIId\

4427: in Example ($b$)):

4428: $$

4429: \GLtmI \ = \ \muPRxApprox(\IIId)_{\intvlss^m\to\realss}.

4430: $$

4431: %\GLtmI\ is precisely the class of total functions

4432: %uniformly \muPRx\ approximable on 

4433: %the metric algebra \IIId\

4434: (For details, see 

4435: \cite[\S9]{tz:top}, \cite[\S5.9]{tz:hb}.)

4436: Hence, again, a kind of

4437: ``Universal Function Theorem"

4438: holds for \GLtmI,

4439: in the following sense. 

4440: For $m=1,2,\dots$, \,let

4441: $$

4442: H^m\ \eqdf\ \Univ^{\IIId}_{\natss\times\intvlss^m\to\realss}:

4443: \ \NN\times\NN\times I^m\ \to\ \RR

4444: $$

4445: be the universal function for

4446: \,$\muPRx(\IIId)_{\natss\times\intvlss^m\to\realss}$

4447: \,given by the Universal Function Theorem (\S4.4).

4448: Then for each \,$f\in\GLtmI$,

4449: there is a number $k$, 

4450: effectively obtainable from the GL-code for $f$,

4451: such that (writing \,$H^m_{k,n} = H^m(k,n,\,\cdot\,)$)

4452: \,the sequence of functions

4453: \,$H^m_{k,0},\,H^m_{k,1},\,H^m_{k,2},\,\dots$

4454: \,uniformly approximates $f$ on $I$.

4455: 

4456: So by

4457: Theorem 2 applied to \IIId:

4458: \endpr

4459: 

4460: 

4461: \Thmn{3 \,(Universal specification of GL computable functions)}

4462: For each $m>0$

4463: there is a signature

4464: \,\Sigxm\ \,which is an expansion of \,\SigIIIdx\ \,by finitely many

4465: function symbols,

4466: and a finite conditional specification 

4467: \,$(\Sigxm,\,\EWm(\ztt))$

4468: \,consisting of conditional equations and inequalities,

4469: \,which is 

4470: universal for specifications of \GLtmI,

4471: in the following sense:

4472: it contains a distinguished natural number variable \ztt\

4473: such that 

4474: each function

4475: $f\in\GLtmI$ 

4476: is specified (with hidden sorts and functions)

4477: by a suitable substitution instance

4478: \,$(\Sigxm,\,\EWm(\kbar))$,

4479: \,where $k$ can be found effectively from a GL-code for $f$.

4480: \endpr

4481: 

4482: \Remarkn{4 \,(Description of the signature \,\Sigxm\ \,of Theorem 3)}

4483: The signature \Sigxm\ is an expansion of \,\SigIIId\

4484: \,(for a description of which see the remark at the end of \S6.1)

4485: by the following sorts and functions:

4486: 

4487: \itemm{($i$)}

4488: the sorts and functions of the array structure over \IIId\ (\S1.6);

4489: 

4490: \itemm{($ii$)}

4491: the \muPRx\ ``universal function" $F^m$ for \GLtmI\

4492: (as described in the above discussion)

4493: together with the auxiliary functions in its derivation;

4494: 

4495: \itemm{($iii$)}

4496: the function $2^{-n}$, used for assertions about 

4497: computable approximations, as explained in the proof of 

4498: Theorem 1;

4499: 

4500: \itemm{($iv$)}

4501: the characteristic function for BU quantification,

4502: as described in the proof of the

4503: BU elimination theorem (\S3.3).

4504: 

4505: \sn

4506: Note that there is only one function of type ($iv$) in \Sigxm,

4507: namely that obtained by eliminating

4508: the conditional BU equation \Fmu\ (\S5.2) specifying the (single!)

4509: \muu-operator occurrence in the 

4510: \muPRx\ derivation for $H^m$ in ($ii$)

4511: (see Remark 1 in \S4.4).

4512: (There are no conditional BU equality axioms for arrays (\S3.2)

4513: to eliminate here, since \,\reals\ \,and \,\intvls\ 

4514: \,are not equality sorts.)

4515: \endpr

4516: 

4517: 

4518: 

4519: \shead{6.3}{Illustration: \ Specification of dynamical system}

4520: We illustrate the connection between algebraic specification methods

4521: and models of physical systems.

4522: 

4523: A {\it deterministic dynamical system\/} with finite dimensional state space 

4524: \,$S\sseq\RR^n$ \,and \,time \,$T\sseq\RR$

4525: \,is represented in a model by a function

4526: $$

4527: \phi: \ T\times S\ \to\ S

4528: $$

4529: where for \,$t\in T$, $s\in S$, 

4530: \,$\phi(t,s)$ \,is the state of the system at time $t$ with 

4531: initial state $s$.

4532: For example, the state of a particle in motion

4533: is represented by position and velocity.

4534: Thus, for a system of $n$ particles in 3-dimensional space,

4535: the state space has $6n$ dimensions.

4536: 

4537: In practice, the model is specified by

4538: ordinary differential equations (ODEs) whose complete solution is \,$\phi$.

4539: Specifically, in the modern qualitative theory of ODEs

4540: \cite{arnold}, \,$\phi$ \,is differentiable,  and the function

4541: \,$\phi_t:S\to S$ \,defined by

4542: $$

4543: \phi_t(s) \ = \ \phi(t,s) \quad \tx{for} \ \ t\in T, s\in S,

4544: $$

4545: is a 1-parameter group of diffeomorphisms of $S$;

4546: the action of this group on $S$ is 

4547: called the {\it flow\/} on the {\it phase space\/} $S$.

4548: This flow can be specified by a vector field on $S$.

4549: 

4550: In modelling a physical system, one aim is to compute

4551: values of the function \,$\phi$ \,on some time interval

4552: and subspace of the space of initial conditions.

4553: Many methods exist to derive algorithms for $\phi$

4554: \,from the equations that define it.

4555: Indeed, various fields of applied mathematics

4556: exist in order to design such equations,

4557: and the field of numerical analysis exists

4558: to design such solution methods.

4559: 

4560: Conversely, we suppose that \,$\phi$ can be simulated on a

4561: digital computer, \ie, \,$\phi$ \,is a classically 

4562: computable (\eg, GL-computable) function.

4563: {\it Assume also that the state space $S$ is the unit $n$-cube $I^n$,

4564: and the time dimension $T$ is the unit interval $I$.\/}

4565: Thus

4566: $$

4567: \phi\: I\times I^n\ \to \ I.

4568: $$

4569: We can now apply Theorem 3 to show that 

4570: the dynamical system has a finite algebraic specification.

4571: 

4572: 

4573: \Thmn{4 \,(Universal specification of computable dynamical systems)}

4574: For each $n>0$

4575: there is a signature \,\Sigtilxn\ \,which

4576: extends \,\SigIIIdx\ \,by function symbols,

4577: and a finite conditional specification

4578: \,$(\Sigtilxn,\,\EtilWn(\ztt))$

4579: \,consisting of conditional equations and inequalities,

4580: which is universal for all 

4581: classically computable dynamical systems

4582: on the unit $n$-cube $I^n$ over the unit time interval $I$.

4583: \endpr

4584: Note that \Sigtilxn\ is essentially the signature \Sigxm\ 

4585: of Theorem 3, with $n=m+1$.

4586: 

4587: 

4588: We have shown above how powerful algebraic specifications are,

4589: even for topological data types.

4590: More research needs to be done to determine the extent of its power,

4591: especially on metric algebras.

4592: Here topological notions such as continuity can play a part,

4593: as we see from the following example.

4594: 

4595: \shead{6.4}{Example: \,Specification of function assuming continuity}

4596: Consider the two equations for a (total) function $f$ on the real line:

4597: $$

4598: f(x+y) \ = \ f(x)\times f(y), \qqquad f(1) \ = \ c

4599: \tagx

4600: $$

4601: for some constant $c>0$.

4602: If we assume that $f$ is continuous, even at one point,

4603: then it is easy to see that these equations are satisfied uniquely

4604: by the function 

4605: $$

4606: f(x) \ = \ c^x.

4607: $$

4608: However, in the absence of any such

4609: continuity assumption,

4610: it can be shown that (for any $c>0$) there are

4611: $2^{2^{\aleph_0}}$ non-constructive solutions to ($*$).

4612: Here ``non-constructive" means both that these solutions 

4613: are non-computable, and that their existence is proved 

4614: by non-constructive means, using Zorn's Lemma

4615: to show the existence of a Hamel basis on \,\RR, \,\ie,

4616: a maximal linear independent subset of \,\RR\ \,over \,\QQ.

4617: 

4618: Note that any solution $f$ of ($*$) is a homomorphism 

4619: from the additive group of reals to the multiplicative group

4620: of positive reals.

4621: 

4622: This example suggests the following 

4623: 

4624: \Question

4625: On metric algebras, does conditional equational

4626: specifiability,

4627: together with a topological condition

4628: such as continuity,

4629: imply computability?

4630: \endpr

4631: 

4632: Specifically, is there a continuous function on $I$ which

4633: is definable by equations but not approximably computable?

4634: 

4635: Note, in this connection, that there are other ``equational specifications"

4636: for the exponential function \,$e^x$:

4637: \itemm{(1)}

4638: the differential equation \ $f'(x) = f(x)$ \ with 

4639: initial condition \ $f(0) = 1$;

4640: \itemm{(2)}

4641: from the polynomial approximations

4642: given by the partial sums of the Maclaurin expansion 

4643: \ $\sum_{i=0}^\infty {x^i}\slash{i!}$,

4644: \ a specification consisting of conditional equations and inequalities

4645: can be derived 

4646: by the methods of this section for approximating computations;

4647: \itemm{(3)}

4648: similarly, from the polynomial approximations,\ $(1+x\slash n)^n$,

4649: \,a specification consisting of

4650: conditional equations and inequalities

4651: can be derived.

4652: 

4653: Note that in the first of these specifications, differentiability

4654: of $f$ is (of course) implicitly assumed,

4655: and uniqueness of the solution follows by

4656: the Lipschitz condition;

4657: however no assumptions of continuity are needed in (2) or (3).

4658: 

4659: The above question points to an open field of research.

4660: The investigation of computable solutions of recursive equations

4661: in \cite{gaertner-hotz} would be relevant here.

4662: 

4663: \bn

4664: 

4665: \itemm{\bbf 7 \ \ }

4666: {\bbf Initial algebra specifications with

4667: conditional equations and conditional BU equations}

4668: 

4669: \mn

4670: In this section

4671: we will consider theories $T$, which we assume to be 

4672: formalised in logical formalisms \FFF\

4673: of the kind described in Section 2;

4674: for example, \,$\FFF= \CondBUEqSig$.

4675: \shead{7.1}{Pre-initial and initial models}

4676: In this subsection (only), we make no assumptions 

4677: concerning the (N-)standardness of

4678: signatures or algebras.

4679: Let \Sig\ be a signature and

4680: let \KK\ be a \Sig-adt.

4681: 

4682: A formalism \FFF\ is said to be {\it valid for\/} \KK\

4683: if the axioms and inference rules of \FFF\

4684: hold for all algebras in \KK.

4685: Note, for example, that \CondBUEqSig\

4686: is valid for \NStdAlgSig,

4687: but not, in general, for \AlgSig.

4688: 

4689: 

4690: A \Sig-algebra $A$ is {\it pre-initial for} \KK\

4691: if there is a unique \Sig-homomorphism from $A$ to

4692: every algebra in \KK;

4693: \,{\it pre}-initial in that

4694: it might not itself belong to \KK.

4695: (The notion of \Sig-{\it homomorphism\/} 

4696: between \Sig-algebras is defined as usual \cite{meinke-jvt}.)

4697: 

4698: Note that the {\it closed term algebra} \ \TSig\

4699: \ is pre-initial for \KK.

4700: 

4701: An {\it initial algebra of\/} \KK\

4702: is a pre-initial algebra 

4703: which belongs to \KK.

4704: As is easily seen,

4705: any two initial algebras of \KK\

4706: must be \Sig-isomorphic.

4707: We denote any initial algebra of \KK\ by \ISigKK.

4708: 

4709: We will be interested in the case that 

4710: $$

4711: \KK \ = \ \AlgSigT,

4712: $$

4713: the class of models of a first-order \Sig-theory $T$,

4714: where $T$ may have certain syntactic restrictions.

4715: We will assume:

4716: 

4717: \bull 

4718: in this subsection that $T$ is a conditional

4719: equational theory; 

4720: 

4721: \bull

4722: in \S7.2 

4723: likewise, but restrict attention to N-standard models of $T$;

4724: 

4725: \bull

4726: in \S7.3 that $T$ is a conditional BU equational theory

4727: (again with N-standard models);

4728: 

4729: \bull

4730: and in \S7.4 that it is a conditional SU equational theory

4731: (ditto).

4732: 

4733: \n

4734: (Recall the formal systems defined in Section 2.)

4735: Finally in \S7.5 we will show how conditional BU equational initial algebra

4736: specifications can be ``reduced" to conditional equational initial algebra

4737: specifications.

4738: 

4739: Let $T$ be a \Sig-theory.

4740: We write \,\ISigT \,for the initial algebra \,

4741: $\Init\bigl(\Sig,\AlgSigT\bigr)$

4742: \,(if it exists), and call it the {\it initial model of $T$}.

4743: 

4744: Consider the {\it closed term algebra\/} \,\TSigTF\

4745: \,formed from \,\TSig\ \,by identifying closed terms

4746: provably equal from $T$, in some formalism \FFF, \ie, 

4747: $$

4748: \TSigTF \ \eqdf\ \TSig/\approxTF

4749: $$

4750: where

4751: $$

4752: t_1 \approxTF t_2 \ \ \ \llongtofrom_{df}\ \ \ 

4753:   \tx{$t_1=t_2$ is provable from $T$ in \FFF}.

4754: $$

4755: 

4756: \Lemma

4757: If \FFF\ is valid for \AlgSigT,

4758: then \TSigTF\ is pre-initial for \AlgSigT.

4759: \endpr

4760: 

4761: We will investigate whether \,\TSigTF\

4762: is, further, {\it initial\/} for \,\AlgSigT, \,\ie, whether

4763: $$

4764: \TSigTF \ = \ \ISigT.

4765: $$

4766: 

4767: 

4768: \pr{Initiality Lemma}\sl

4769: Suppose \FFF\ is valid for \AlgSigT.

4770: If $\TSigTF \in \AlgSigT$,

4771: \,then it is (\Sig-isomorphic to) \ISigT.

4772: \endpr

4773: 

4774: 

4775: \Defs

4776: Let $A$ be a \Sig-algebra.

4777: \itemm{(1)}

4778: $A$ has an {\it initial algebra specification} \SigT\

4779: if \,$A \,\cong\, \ISigT$.

4780: \itemm{(2)}

4781: $A$ has an {\it initial algebra specification with

4782: hidden sorts and/or functions} \,$(\Sigp,$ $ T')$

4783: \,if \Sigp\ is an expansion of \Sig\ by sorts and/or functions, 

4784: $T'$ is a \Sigp-theory and 

4785: $$

4786: A \ \cong \ \Init\bigl(\Sig,\,\Alg(\Sigp,T')\redSig\,\bigr).

4787: $$

4788: \endpr

4789: 

4790: \newpage

4791: 

4792: \Thmn{1 \,\cite{malcev}}

4793: Let $E$ be a conditional equational theory over \Sig.

4794: Let \,$I \eqdf \Tbi(\Sig,E,\CondEqSig)$.

4795: Then $I$ is an initial model of $E$.

4796: \,Furthermore, if

4797: \,$t_1, t_2$ \,are two closed \Sig-terms of the same sort,

4798: then the following are equivalent:

4799: \itemm{($i$)}

4800: $t_1$ and $t_2$ have the same value in $I$,

4801: \itemm{($ii$)}

4802: $t_1$ and $t_2$ have the same value in all models of $E$,

4803: \itemm{($iii$)}

4804: $t_1 = t_2$ \ is provable from $E$ in \,\CondEqSig,

4805: \itemm{($iv$)}

4806: $t_1 = t_2$ \ is provable from $E$ in \,\FOLSig.

4807: \endpr

4808: 

4809: \Pf

4810: The main thing here is to show that 

4811: \,$I\ttstile E$,

4812: from which \,$(ii)\tto(i)$ \,will follow.

4813: %({\it Exercise.\/}

4814: %This follows from the closure of \,\CondEqSig\

4815: %\,under substitutions and under atomic cuts.)

4816: Since $I$ is a (closed) term model,

4817: it is sufficient 

4818: to show that $I$ satisfies all {\it closed substitution instances}

4819: of the axioms of $E$.

4820: So consider any closed instance

4821: \ $P_1 \con \dots \con P_n \to P$ 

4822: \ of an axiom of $E$,

4823: where $P_i$ and $P$ are closed equations.

4824: Note that the corresponding sequent

4825: $$

4826: P_1, \dots, P_n \seqt P

4827: \tagx

4828: $$

4829: is derivable from $E$ in \,\CondEqSig,

4830: \,by the substitution rule.

4831: Suppose \ $I \ttstile P_i$ \ for $i = 1,\dots,n$.

4832: Then, by the definition of $I$, $P_i$ is provable from $E$ in \CondEqSig.

4833: But then $P$ is also provable, 

4834: by repeated (atomic) cuts

4835: of the sequent ($*$)

4836: with the sequents \ $\seqt P_i$,

4837: and so $I \ttstile P$.

4838: 

4839: Hence \,$I\ttstile E$.

4840: It follows, by the Initiality Lemma,

4841: that $I$ is an {\it initial\/} model of $E$.

4842: Hence also \,$(ii)\tto(i)$.

4843: The further implications

4844: \,$(i)\tto(iii)\tto(iv)\tto(ii)$ 

4845: \,are all trivial.

4846: \endpf

4847: 

4848: \Remarkn{\,(Completeness and conservativity)}

4849: Mal'cev's Theorem \cite{malcev}, in the form given above, can be viewed

4850: as expressing both

4851: \ ($a$) {\it completeness} of \,\CondEqSig, \,given by the implication

4852: \,$(ii)\impp(iii)$, \,and

4853: \ ($b$) {\it conservativity\/} of first order logic with equality

4854: over \,\CondEqSig, \,given by the implication \,$(iv)\impp(iii)$.

4855: (\Cf\ conservativity lemma (1) and the remark in \S2.6.)

4856: \endpr

4857: 

4858: Necessary and sufficient conditions for the existence of initial models

4859: of theories are given in \cite{mahr-makowsky}.

4860: 

4861: \shead{7.2}{Initial N-standard models}

4862: Assume, from now on, that 

4863: \Sig\ is N-standard, and that \KK\ 

4864: consists of {\it N-standard} \Sig-algebras;

4865: for example,

4866: \ $\KK = \NStdAlgSigT$,

4867: \ for some \Sig-theory $T$.

4868: Then 

4869: \ \TSigTF,

4870: \ although it is pre-initial for \KK,

4871: might fail to be initial for \KK\

4872: for two reasons:

4873: it might not satisfy $T$,

4874: and it might not even be N-standard! 

4875: (We return to the second point below.)

4876: 

4877: An {\it initial N-standard model of\/} $T$

4878: is an initial algebra of \,\NStdAlgSigT.

4879: Any two initial N-standard models of $T$

4880: are \Sig-isomorphic.

4881: We denote any such model by 

4882: $$

4883: \INSSigT \ \eqdf\ \Init(\Sig,\,\NStdAlgSigT).

4884: $$

4885: \sn

4886: {\bf N-Standard Initiality Lemma\/}.\sl

4887: \ \ Suppose \FFF\ is valid for \NStdAlgSigT.

4888: \nl

4889: If \,$\TSigTF \in \NStdAlgSigT$

4890: \,then it is (\Sig-iso\-mor\-phic to) 

4891: \INSSigT.

4892: \endpr

4893: 

4894: \Defs

4895: Let $A$ be an N-standard \Sig-algebra.

4896: \itemm{(1)}

4897: $A$ has an {\it initial N-standard algebra specification} \SigT\

4898: \,if \ $A \,\cong\, \INSSigT$.

4899: \itemm{(2)}

4900: $A$ has an {\it initial N-standard algebra specification with

4901: hidden sorts and/or functions} \ $(\Sigp, T')$

4902: \ if \Sigp\ is an expansion of \Sig\ by sorts and/or functions, 

4903: $T'$ is a \Sigp-theory and 

4904: $$

4905: A \ \cong\ \Init\bigl(\Sig, \, \NStdAlg(\Sigp,T')\redSig \,\bigr).

4906: $$

4907: 

4908: Note that \INSSigT\ (if it exists)

4909: might not be an initial model of $T$,

4910: \ie, $T$ might have another, non-N-standard, initial model,

4911: as the following example demonstrates.

4912: 

4913: \Examplen{\,(Initial N-standard model of a theory

4914: which is not an initial model of that theory)}

4915: Let \Sig\ contain (in addition to the standard operations on \nats\ and \bools)

4916: a constant \,$\uuu:\bools$, \,and let $T$ contain the single axiom

4917: \ `$\uuu \ne \trues$'.

4918: Then the term algebra \ \TSig\

4919: trivially satisfies $T$,

4920: and is hence (by the Initiality Lemma of \S7.1)

4921: an {\it initial} model of $T$.

4922: It is {\it not N-standard}, since

4923: it has a 3-element carrier of sort \bools,

4924: with distinct denotations of \trues, \falses\ and \uuu.

4925: There is, however, also an {\it initial N-standard} model of $T$

4926: with an N-standard (2-element) carrier of sort \bools,

4927: formed by identifying \uuu\ and \falses.

4928: \endpr

4929: 

4930: 

4931: Now \TSigTF\ \ may fail to be N-standard for two reasons:

4932: that $T$ proves ``too little" or ``too much",

4933: roughly speaking.

4934: The first reason

4935: is connected with non-N-standard interpretations of the sorts \nats\

4936: and \bools.  

4937: Thus,

4938: there may be a function symbol $f$ in \Sig\

4939: with range sort \nats,

4940: without corresponding axioms in $T$ capable of ``reducing"

4941: \,$f(t)$, for some closed term $t$, to a numeral.

4942: Similarly (as in the above example), 

4943: not all closed boolean terms (\ie, terms of sort \bools)

4944: may be (provably in $T$) equal to \trues\ or \falses.

4945: (In the terminology of \cite{guttag-horning} the specification \SigT\ is not

4946: ``sufficiently complete".)

4947: The second reason is that $T$ may be {\it inconsistent},

4948: in the sense that it proves

4949: `$\trues = \falses$' (or, equivalently in a suitable 

4950: weak background theory, `$0 = 1$').

4951: This motivates the following definitions.

4952: Note that we must (to begin with) 

4953: speak of provability relative to some formal system

4954: \,\FFF, \,which will typically be 

4955: one of the system \,\CondEqSig\ \,or \,\CondBUEqSig\

4956: \,of Section 2.

4957: 

4958: \Defn3

4959: $T$ is {\it consistent} in \,\FFF\ \,if the equation `$\trues = \falses$'

4960: is not provable in \FFF\ from $T$.

4961: \endpr

4962: 

4963: \Defn4

4964: $T$ {\it determines \nats} in \FFF\

4965: if every closed term of sort \nats\ is, provably in \FFF\ from $T$,

4966: equal to a numeral;

4967: and $T$ {\it determines \bools} in \FFF\

4968: if every closed term of sort \bools\ is, provably in \FFF\ from $T$,

4969: equal to \trues\ or \falses.

4970: \endpr

4971: 

4972: \newpage

4973: 

4974: \Defn{5 \,(N-standardness axioms)}

4975: \sn

4976: ($a$) \,\NStdAxSig\ \, is the following set of conditional equations:

4977: $$

4978: \boxed{

4979: \gathered

4980: \ands(\trues,\trues) = \trues, \qquad

4981: \ands(\trues,\falses) = \ands(\falses,\trues)

4982:   = \ands(\falses,\falses) = \falses, \\

4983: \ors(\falses,\falses) = \falses, \qquad

4984: \ors(\trues,\trues) = \ors(\trues,\falses)

4985:   = \ors(\falses,\trues) = \trues, \\

4986: \nots(\trues) = \falses, \qquad \nots(\falses) = \trues, \\

4987: \ifs_s(\trues, \,\xtt_1^s, \xtt_2^s) \ = \ \xtt_1^s,

4988: \qquad \ifs_s(\falses, \,\xtt_1^s, \xtt_2^s) \ = \ \xtt_2^s, \\

4989: \eqnat(0,0) = \trues, \qquad \eqnat (\Ss \ztt,0) = 

4990: \eqnat(0,\Ss \ztt) = \falses, \\

4991: \eqnat(\Ss \ztt_1,\Ss \ztt_2) = \eqnat(\ztt_1,\ztt_2), \\

4992: \lsnat(0,\Ss \ztt) = \trues, \qquad \lsnat (\ztt,0) = \falses, \\

4993: \lsnat(\Ss \ztt_1,\Ss \ztt_2) = \lsnat(\ztt_1,\ztt_2), \\

4994: \eqs_s(\xtt^s,\xtt^s) = \trues, \\

4995: \eqs_s(\xtt^s_1, \xtt^s_2) = \trues \ \to \ t^s_1 = t^s_2.

4996: \endgathered

4997: }

4998: $$

4999: where, in the axioms for \,$\ifs_s$, 

5000: \,$s$ ranges over all \Sig-sorts other than \,\bools;

5001: \,and in the axioms for \,$\eqs_s$,

5002: \,$s$ ranges over all \Sig-{\it equality sorts} other than \,\nats,

5003: 

5004: \sn

5005: ($b$) \,$\NStdAxoSig$ 

5006: \,is the set of all {\it closed \Sig-substitution instances\/} of 

5007: \,\NStdAxSig.

5008: \endpr

5009: 

5010: %$$

5011: %\boxed{

5012: %\gathered

5013: %\ands(\trues,\trues) = \trues, \qquad

5014: %\ands(\trues,\falses) = \ands(\falses,\trues) 

5015: %  = \ands(\falses,\falses) = \falses, \\

5016: %\ors(\falses,\falses) = \falses, \qquad

5017: %\ors(\trues,\trues) = \ors(\trues,\falses) 

5018: %  = \ors(\falses,\trues) = \trues, \\

5019: %\nots(\trues) = \falses, \qquad \nots(\falses) = \trues, \\

5020: %\ifs_s(\trues, \,t_1^s, t_2^s) \ = \ t_1^s,

5021: %\qquad \ifs_s(\falses, \,t_1^s, t_2^s) \ = \ t_2^s, \\

5022: %\eqnat(\nbar,\,\nbar)\ = \ \trues \qquad\tx{for all \ninNN}, \\

5023: %\eqnat(\mbar,\,\nbar)\ = \ \falses \qquad\tx{for all $m,n\in\NN, \ m\ne n$}, \\

5024: %\lsnat(\mbar,\,\nbar)\ = \ \trues \qquad\tx{for all $m,n\in\NN, \ m< n$}, \\

5025: %\lsnat(\mbar,\,\nbar)\ = 

5026: %  \ \falses \qquad\tx{for all $m,n\in\NN, \ m\not< n$}, \\

5027: %\eqs_s(t^s,t^s) = \trues, \\

5028: %\eqs_s(t^s_1, t^s_2) = \trues \ \to \ t^s_1 = t^s_2

5029: %\endgathered

5030: %}

5031: %$$

5032: %\endpr

5033: 

5034: Note that

5035: \,$\NStdAxSig + \IndSig$ \,holds in any N-standard \Sig-algebra.

5036: 

5037: We use the terminology:

5038: $T$ {\it proves \NStdAxoSig\ in\/} \FFF\

5039: \,to mean: \,\NStdAxoSig\ is derivable from $T$ in \FFF.

5040: 

5041: We now state some lemmas which give sufficient conditions for 

5042: a term model \,\TSigTF\ \,to be N-standard.

5043: 

5044: 

5045: 

5046: \Lemman{1 \,(N-standardness lemma)}

5047: Suppose that in \,\FFF\

5048: \itemm{($i$)}

5049: $T$ is consistent,

5050: \itemm{($ii$)}

5051: $T$ determines \,\nats\ \,and \,\bools, \,and

5052: \itemm{($iii$)}

5053: $T$ proves \NStdAxoSig.

5054: \sn

5055: Then \,\TSigTF\ \,is N-standard.

5056: \endpr

5057: 

5058: \Lemman2

5059: If \Sig\ is strictly N-standard

5060: then \,\NStdAxoSig\ \,determines \,\nats\ \,and \,\bools\

5061: \,in \,\CondEqSig.

5062: \endpr

5063: 

5064: \Pf

5065: By structural induction on all closed \Sig-terms of sort

5066: \,\nats\ \,and \,\bools\ \,(simultaneous\-ly).

5067: \endpf

5068: 

5069: 

5070: The following is 

5071: an immediate consequence

5072: of Lemmas 1 and 2.

5073: 

5074: \newpage

5075: 

5076: \Lemman{3 \,(Strict N-standardness lemma)}

5077: Suppose \Sig\ is {\it strictly\/} N-standard, 

5078: \,\FFF\ \,is at least as strong as \,\CondEqSig,

5079: \,and in \,\FFF\

5080: \itemm{($i$)}

5081: $T$ is consistent, \,and

5082: \itemm{($ii$)}

5083: $T$ proves \NStdAxoSig\ \,(or \,\NStdAxSig).

5084: \sn

5085: Then \,\TSigTF\ \,is N-standard.

5086: \endpr

5087: 

5088: 

5089: \shead{7.3}{Conditional equational theories}

5090: We now give the analogue of Mal'cev's Theorem (\S7.1)

5091: for N-standard models of conditional equational theories.

5092: 

5093: \Thmn2

5094: Let $E$ be a conditional equational theory over \Sig.

5095: Suppose that in 

5096: \linebreak

5097: \CondEqSig, 

5098: \,$E$ is consistent, determines \nats\ and \bools,

5099: and proves \,\NStdAxoSig.

5100: Then 

5101: \linebreak

5102: $I \eqdf \Tbi(\Sig,\,E,\,\CondEqSig)$

5103: \,is an initial N-standard model of $E$.

5104: Furthermore, if \ $t_1, t_2$ \ are two closed \Sig-terms of the same sort,

5105: then the following are equivalent:

5106: \itemm{($i$)}

5107: $t_1$ and $t_2$ have the same value in $I$,

5108: \itemm{($ii$)}

5109: $t_1$ and $t_2$ have the same value in all N-standard models of $E$,

5110: \itemm{($iii$)}

5111: $t_1 = t_2$ \ is provable from $E$ \ in \,\CondEqSig,

5112: \itemm{($iv$)}

5113: $t_1 = t_2$ \ is provable from $E$ in \,$\FOLSig+\IndSig$.

5114: \endpr

5115: 

5116: \Pf

5117: By the N-standardness Lemma (\S7.2),

5118: $I$ is an N-standard algebra.

5119: As in Theorem 1, the main thing is to show

5120: that \,$I\ttstile E$.

5121: This is done exactly as in the proof of Theorem 1.

5122: It follows, by the N-standard Initiality Lemma (\S7.2),

5123: that $I$ is an {\it initial\/} N-standard model of $E$.

5124: The rest of the

5125: proof is similar to that for Theorem 1.

5126: Note for the implication \,$(iv)\tto(ii)$,

5127: \,we use the fact that the rule \,\IndSig\ \,is valid

5128: for N-standard \Sig-algebras.

5129: \endpf

5130: 

5131: 

5132: \Remarks

5133: (1) \,By Lemma 2 in \S7.2,

5134: the assumption in the theorem that $E$ determines \,\nats\ 

5135: \,and \,\bools\ \,can be replaced by the assumption that 

5136: \Sig\ is strictly N-standard.

5137: \sn

5138: (2) \,({\it Completeness and conservativity.\/})

5139: \,Here again, the implication \,$(ii)\tto(iii)$ 

5140: \,can be construed as a 

5141: completeness theorem,

5142: \,and \,$(iv)\tto(iii)$ \,

5143: as a conservativity theorem.

5144: (See the Remark in \S2.6 and the Remark following Theorem 1.)

5145: \sn

5146: (3) \,({\it The N-standardness axioms\/}.)

5147: We have ``incorporated" the N-standard\-ness axioms \,\NStdAxoSig\

5148: \,in the theory $E$, so to speak,

5149: by assuming that $E$ proves them.

5150: Another feasible approach would be to incorporate these axioms

5151: in the logics \,\CondEq, \,\CondBUEq\

5152: \,and \,\FOL,

5153: \,by adding them as axioms

5154: (as we did with the boundedness axioms \,\BddAx\ \,in \,\CondBUEq).

5155: This would entail some minor re-wording of the theorems.

5156: \endpr

5157: 

5158: We turn our attention to theories with

5159: syntactic structure more complicated than conditional equations.

5160: 

5161: \newpage

5162: 

5163: 

5164: \shead{7.4}{Conditional BU equational theories}

5165: We give the analogue of Mal'cev's Theorem for

5166: N-standard models of

5167: BU conditional equational theories.

5168: 

5169: \Thmn3

5170: Let $F$ be a conditional BU equational theory over \Sig.

5171: Suppose that in \,\CondBUEqSig, 

5172: \,$F$ 

5173: is consistent, determines \nats\ and \bools\ and

5174: proves \,\NStdAxoSig.

5175: Then \ $I \eqdf \Tbi(\Sig,\,F,\,\CondBUEqSig )$

5176: \ is an initial N-standard model of $F$.

5177: \ Furthermore, if $t_1, t_2$ are two closed \Sig-terms of the same sort,

5178: then the following are equivalent:

5179: \itemm{($i$)}

5180: $t_1$ and $t_2$ have the same value in $I$,

5181: \itemm{($ii$)}

5182: $t_1$ and $t_2$ have the same value in all N-standard models of $F$,

5183: \itemm{($iii$)}

5184: $t_1 = t_2$ \,is provable from \,$F$

5185: \,in \,\CondBUEqSig,

5186: \itemm{($iv$)}

5187: $t_1 = t_2$ \,is provable from \,$F$ 

5188: \,in \,$\FOLSig+\IndSig$.

5189: \endpr

5190: 

5191: \Pf

5192: By the N-standardness Lemma, $I$ is N-standard.

5193: As in Theorems 1 and 2, the main thing is to show that \,$I\ttstile F$.

5194: Again, since $I$ is a term model, 

5195: it is sufficient to show that $I$

5196: satisfies the set of closed substitution instances of $F$.

5197: First note that, by definition,

5198: $I$ satisfies precisely all closed equations provable

5199: from $F$ in CondBUEq, \ie, for any 

5200: {\it closed equation\/} $P$:

5201: $$

5202: I \ttstile P \ \ \llongtofrom \ \ F \tstile P

5203: \tagx

5204: $$

5205: where `$\tstile$' here means provability in \CondBUEq.

5206: Further, by use of the {\it boundedness axioms\/} \,\BddAx\ 

5207: \,of \CondBUEq\ (\S2.3),

5208: the same holds

5209: for any {\it closed BU equation} $Q$:

5210: $$

5211: I \ttstile Q \ \ \llongtofrom \ \ F \tstile Q.

5212: \tagxx

5213: $$

5214: For suppose

5215: \,$Q\,\ident \,\all \ztt < t P(\ztt)$,

5216: \ where $P(\ztt)$ is an equation. 

5217: Since $I$ is N-standard, 

5218: $$

5219: I\ttstile t \ = \ \nbar

5220: \tagxxx

5221: $$

5222: for some (unique) $n$.

5223: Then

5224: $$

5225: \align

5226: \ I \ttstile \all \ztt < t P(\ztt) \ \

5227:    &\llongtofrom \ \ \tx{for all $k<n$,}\ I \ttstile P(\kbar)\\

5228:    &\llongtofrom \ \ \tx{for all $k<n$,}\ F \tstile P(\bar k) 

5229:      \qquad \tx{by} \ \ (*)\\

5230:    &\llongtofrom \ \ F \tstile \all \ztt < t P(\ztt) 

5231:      \qqquad\quad\, \tx{by \,\BddAx\ \,and (\xxx)}.

5232: \endalign

5233: $$

5234: Now consider any closed instance

5235: \ $f \ \ident \ Q_1 \con \dots \con Q_m \to Q$ 

5236: \ of an axiom of $F$ 

5237: (where $Q_i$ and $Q$ are closed SU equations).

5238: Suppose \,$I \ttstile Q_i$ \,for $i = 1,\dots,m$.

5239: Then by ($**$) $Q_i$ is provable from $F$ in \CondBUEq.

5240: But then so is $Q$, 

5241: by repeated cuts

5242: of the sequent

5243: \ $Q_1, \dots, Q_m \seqt Q$ 

5244: \ corresponding to $f$

5245: with the sequents \ $\seqt Q_i$,

5246: and so $I \ttstile Q$.

5247: \endpf

5248: 

5249: \Remarks

5250: (1) \,As before,

5251: the assumption in the theorem that $F$ determines \,\nats\ 

5252: \,and \,\bools\ \,can be replaced by the assumption that 

5253: \Sig\ is strictly N-standard.

5254: \sn

5255: (2) \,({\it Completeness and conservativity.\/})

5256: \,Again, the implication \,$(ii)\tto(iii)$

5257: \,can be construed as a completeness theorem,

5258: and \,$(ii)\tto(iii)$ \,as a conservativity theorem.

5259: \endpr

5260: 

5261: 

5262: \shead{7.5}{Conditional SU equational theories}

5263: Now we turn to the 

5264: infinitary conditional SU equational logic (\S2.4).

5265: Although it will not be used further in the paper,

5266: it is interesting in its own right.

5267: 

5268: Remember that the infinitary \om-rule \,$\all_\om R$

5269: \,obviates the need for an induction rule.

5270: 

5271: \Thmn4

5272: Let $G$ be a conditional SU equational theory over \Sig.

5273: Suppose that in \,\CondSUEqomSig, 

5274: \,$G$ 

5275: is consistent, determines \nats\ and \bools\ and

5276: proves \,\NStdAxoSig.

5277: Then \ $I \eqdf  \Tbi(\Sig,\,G,\,\CondSUEqomSig)$

5278: \ is an initial N-standard model of $G$.

5279: \ Furthermore, if $t_1, t_2$ are two closed \Sig-terms of the same sort,

5280: then the following are equivalent:

5281: \itemm{($i$)}

5282: $t_1$ and $t_2$ have the same value in $I$,

5283: \itemm{($ii$)}

5284: $t_1$ and $t_2$ have the same value in all N-standard models of $G$,

5285: \itemm{($iii$)}

5286: $t_1 = t_2$ \,is provable from \,$G$

5287: \,in \,\CondSUEqomSig,

5288: \itemm{($iv$)}

5289: $t_1 = t_2$ \,is provable from \,$G$ 

5290: \,in \,\FOLomSig.

5291: \endpr

5292: 

5293: \Pf

5294: By the N-standardness Lemma, $I$ is N-standard.

5295: Again, the main thing is to show that 

5296: $I$ satisfies closed substitution instances of axioms of $G$.

5297: By definition, 

5298: for any {\it closed equation} $P$:

5299: $$

5300: I \ttstile P \ \ \llongtofrom \ \ G \tstile P

5301: \tagx

5302: $$

5303: where `$\tstile$' here means provability in \,\CondSUEqom.

5304: Further, by use of the $\all_\om R$ rule,

5305: the same holds

5306: for any {\it closed SU equation} $R$:

5307: $$

5308: I \ttstile R \ \ \llongtofrom \ \ G \tstile R.

5309: $$

5310: For suppose

5311: \,$R \ident \all \ztt P(\ztt)$,

5312: \ where $P(\ztt)$ is an equation. Then

5313: $$

5314: \align

5315: \ I \ttstile \all \ztt P(\ztt) \ \

5316:    &\llongtofrom \ \ \tx{for all $n$,}\ I \ttstile P(\bar n)\\

5317:    &\llongtofrom \ \ \tx{for all $n$,}\ G \tstile P(\bar n) 

5318:      \qquad \,\tx{by} \ \ (*)\\

5319:    &\llongtofrom \ \ G \ \tstile \all \ztt P(\ztt) 

5320:      \qqquad\quad \tx{by} \ \ \all_\om R

5321: \endalign

5322: $$

5323: The rest of the proof follows the pattern of Theorems 1, 2 and 3.

5324: \endpf

5325: 

5326: \Remarks

5327: (1) \,As before,

5328: the assumption in the theorem that $G$ determines \,\nats\ 

5329: \,and \,\bools\ \,can be replaced by the assumption that 

5330: \Sig\ is strictly N-standard.

5331: \sn

5332: (2) \,({\it Completeness and conservativity.\/})

5333: \,Once again, the implication \,$(ii)\tto(iii)$

5334: \,can be viewed as a completeness theorem,

5335: and \,$(iv)\tto(iii)$

5336: \,as a conservativity theorem.

5337: \endpr

5338: 

5339: 

5340: \shead{7.6}{Open term algebras}

5341: So far (Theorems 1, 2, 3 and 4)

5342: we have concentrated on closed term algebras.

5343: We could also formulate our results in a more general setting,

5344: namely, with term algebras constructed from

5345: open terms, \ie, terms containing free variables

5346: (from a given set $X$).

5347: 

5348: The problem here is that with open terms (an analogy of) the

5349: N-Standardness Lemma (\S7.2) will fail in general.

5350: However, under a certain syntactic condition 

5351: (the ``N-term condition" below),

5352: a version of this Lemma can still be formulated.

5353: 

5354: First we need some definitions and notation.

5355: Given a signature \Sig, and a set

5356: \,$X\sseq \VarSig$,

5357: \,let \,\TSigX\ \,be the set of \Sig-terms in $X$, 

5358: \ie, \Sig-terms containing variables from $X$ only.

5359: In particular, for $X=\nil$, \,we have 

5360: the set of closed \Sig-terms

5361: \,$\TSig = \Tbi(\Sig,\nil)$.

5362: 

5363: Given a first-order \Sig-theory $T$

5364: \,and formalism \,\FFF\

5365: \,which is valid for \,\AlgSigT,

5366: \,let \,\TSigXTF\ \,be the \Sig-{\it term algebra\/} formed 

5367: from \,\TSigX\

5368: \,by identifying terms provably equal from $T$ in \,\FFF.

5369: (The closed term algebra \,\TSigTF\ \,considered above

5370: corresponds to the special case \,$X=\nil$).

5371: 

5372: The algebra \,$I\eqdf\TSigXTF$ \,is 

5373: {\it free for \,$T$ over $X$\/}.

5374: This means that for every model $A$ of $T$,

5375: and every assignment \,$\rho:X\to A$

5376: \,of elements of $A$ to variables in $X$ (of the same sort),

5377: \,there is a {\it unique \Sig-homomorphism\/}

5378: \ $h:I\to A$ 

5379: \ such that \,$h\rest X \rho$.

5380: (This reduces to initiality in \,\AlgSigT\ \,when \,$X=\nil$.)

5381: 

5382: Note that $I$ need not itself be a model of $T$.

5383: However, this will be the case,

5384: provided $T$ satisfies certain syntactic conditions

5385: (\eg, if $T$ is a conditional equational theory;

5386: \cf\ Theorem 1 above).

5387: 

5388: Again, assuming that \Sig\ is N-standard, 

5389: we are interested in the question whether 

5390: $I$ is N-standard.  A useful criterion in this connection

5391: is the following syntactic condition on \Sig\ and $X$:

5392: 

5393: \pr

5394: {N-term Condition for $(\Sig, X)$}

5395: {\sl No \Sig-term of sort \,\nats\ \,or \,\bools\

5396: \,contains any variables from $X$.

5397: \endpr

5398: 

5399: 

5400: \Remarks

5401: (1) \,The N-term condition for $(\Sig,X)$ 

5402: is trivially satisfied when \,$X=\nil$. 

5403: \sn

5404: (2) \,When \Sig\ is strictly N-standard,

5405: it is equivalent to the condition:

5406: \sn

5407: 

5408: \ce{\sl there are no variables in $X$ of sort \,\nats\ \,or \,\bools.}

5409: 

5410: \sn

5411: This follows from Remark 3 in \S1.5.

5412: 

5413: 

5414: Now the theory given above, and specifically Theorems 1 to 4,

5415: can be generalised to the case of open term models

5416: \,\TSigXTF\ , where \,$(\Sig,X)$ \,satisfies the N-term condition.

5417: First, the N-standardness lemma becomes:

5418: 

5419: 

5420: \pr{N-Standardness Lemma$^{\bk X}$}\sl

5421: Suppose that \,$(\Sig,X)$ \,satisfies the N-term condition.

5422: \nl

5423: Suppose further that in \,\FFF

5424: \itemm{($i$)}

5425: $T$ is consistent,

5426: \itemm{($ii$)}

5427: $T$ determines \nats\ and \bools, \,and

5428: \itemm{($iii$)}

5429: $T$ proves \NStdAxoSig.

5430: \sn

5431: Then \,\TSigXTF\ \,is N-standard.

5432: \endpr

5433: 

5434: Next, the strict N-standardness lemma becomes

5435: (using Remark 2 above):

5436: 

5437: \pr{Strict N-Standardness Lemma$^{\bk X}$}\sl

5438: Suppose \Sig\ is {\it strictly\/} N-standard, 

5439: and there are no variables in $X$ of sort \,\nats\ \,or \,\bools.

5440: Suppose also \,\FFF\ \,is at least as strong as \,\CondEqSig,

5441: \,and in \,\FFF\

5442: \itemm{($i$)}

5443: $T$ is consistent, \,and

5444: \itemm{($ii$)}

5445: $T$ proves \NStdAxSig.

5446: \sn

5447: Then \,\TSigXTF\ \,is N-standard.

5448: \endpr

5449: 

5450: Consider next, for example, Theorem 2.

5451: This can be reformulated as follows.

5452: 

5453: \Thmn{$\bk 2^{\bk X}$}

5454: Suppose \,$(\Sig,X)$ \,satisfies the N-term condition.

5455: Let $E$ be a conditional equational theory over \Sig.

5456: Suppose that in \CondEqSig,

5457: \,$E$ is consistent, determines \nats\ and \bools,

5458: and proves \,\NStdAxSig.

5459: Then \ $I \eqdf \Tbi(\Sig,X,\,E,\,\CondEqSig)$

5460: \ is an N-standard model of $E$,

5461: which is free for $E$ over $X$.

5462: \ Furthermore, if \ $t_1, t_2$ \ are two terms in \,\TSigX\ \,of the same sort,

5463: then the following are equivalent:

5464: \itemm{($i$)}

5465: $t_1$ and $t_2$ have the same value in $I$,

5466: \itemm{($ii$)}

5467: $t_1$ and $t_2$ have the same value in all N-standard models of $E$,

5468: \itemm{($iii$)}

5469: $t_1 = t_2$ \ is provable from $E$ \ in \,\CondEqSig,

5470: \itemm{($iv$)}

5471: $t_1 = t_2$ \ is provable from $E$ in \,$\FOLSig+\IndSig$.

5472: \endpr

5473: 

5474: The strict N-standardness Lemma$^X$,

5475: and Theorem 2$^X$, will be used in Section 9.

5476: 

5477: 

5478: \shead{7.7}{Reducing conditional BU to conditional equational specifications}

5479: We re-consider the work of \S3.3 from the viewpoint of initial algebra 

5480: specifications.

5481: 

5482: \Thmn{5 \ (BU elimination for initial algebra specifications)}

5483: Let $F$ be a conditional BU equational theory over \Sig.

5484: Then there is an expansion \Sigp\ of \Sig\ and a 

5485: conditional equational theory $E'$ over \Sigp\

5486: which is equivalent to $F$ (relative to N-standard models) in the sense that:

5487: \itemm{($i$)}

5488: if $A$ is an N-standard \Sig-model of $F$, 

5489: then it has a \Sigp-expansion which is

5490: a N-standard model of $E'$;

5491: \itemm{($ii$)}

5492: if \ $A \cong \INSSigF$ \ then

5493: it has a unique (up to \Sigp/\Sig-isomorphism) \Sigp-expansion $A'$ such that

5494: \ $A' \cong \INS(\Sig', E')$;

5495: \itemm{($iii$)}

5496: if $A'$ is an N-standard \Sigp-model of $E'$, then its \Sig-reduct $A$

5497: is an N-standard model of $F$;

5498: and if 

5499: \ $A' \cong \INS(\Sig', E')$

5500: \ then \ $A \cong \INSSigF$.

5501: \sn

5502: If $F$ contains $q$ occurrences of BU quantifiers, 

5503: then \Sigp\ \,expands \,\Sig\ by one new sort and $q$ new function symbols.

5504: Moreover, if

5505: $F$ is finite, with $e$ axioms (say),

5506: then so is $E'$, with $e+4q$ axioms.

5507: \endpr

5508: 

5509: \Pf

5510: The idea, again, is to incorporate in the signature,

5511: for each BU quantifier occurring in $F$,

5512: a {\it characteristic function}

5513: for that quantifier.

5514: The problem with adjoining a boolean-valued function symbol

5515: \ $\fs:\nats\times u \to \bools$

5516: \ satisfying ($**$) in 

5517: the BU elimination theorem in \S3.3,

5518: is in the case that $A$ is an initial N-standard model of $F$.

5519: In order that its \Sigp-expansion $A'$ be {\it N-standard},

5520: the value of $\fs(n,x)$ must be either \trues\ or \falses\

5521: for every value of the arguments $n,x$.

5522: Furthermore, in order that $A'$ also be {\it initial},

5523: the \Sig-homomorphism $h$ from $A$ to every N-standard model $B$ of $F$

5524: must be extendible to a

5525: \Sigp-homomorphism $h'$ from $A'$ to the \Sigp-expansion $B'$ of $B$.

5526: However, the rhs of ($**$) in \S3.3 will hold ``more often" in $B$ than in $A$

5527: (since $B$ is a homomorphic image of $A$),

5528: with a corresponding {\it change} in the value of $\fs(n,x)$

5529: from \falses\ to \trues!

5530: Hence $h$ cannot, in general, be extended as desired.

5531: (Making \fs\ a 0,1-valued function will cause exactly the same problem.)

5532:  

5533: We therefore adjoin a special sort \Ds\ for the range of

5534: such functions \fs,

5535: with a constant \ds\ which takes the place of `\trues' in ($**$) in \S3.3.

5536: (The point is that when the condition on the rhs of ($**$) fails,

5537: $\fs(n,x)$ is not ``forced" to equal anything else at all.)

5538: Now for each BU quantifier as in ($*$) of \S3.3,

5539: {\it adjoin} to the signature the function symbol 

5540: $$

5541: \fs: \nats\times u \ \to \ \Ds,

5542: $$

5543: and adjoin the axioms formed from (\xxx) and (\xxxx) in \S3.3 

5544: by replacing `\trues' by `\ds' throughout.

5545: In this way we {\it replace} $F$ by a conditional

5546: equational theory $E'$ in \Sigp, with the stated properties.

5547: \endpf

5548: 

5549: \Remark

5550: If $A$ is an N-standard model of $F$, 

5551: then its N-standard \Sigp-expansion $A'$

5552: modelling $E'$,

5553: given by part ($i$) of the theorem,

5554: is not (in general) uniquely determined.

5555: However, the added condition of initiality (on $A$ and $A'$)

5556: determines $A'$ uniquely.

5557: \endpr

5558: 

5559: 

5560: 

5561: \Shead8{Initiality-preserving operators on N-standard algebras}

5562: In this section we combine the theory of Section 5

5563: (``computability $\implies$ algebraic specifiability")

5564: with the initial algebra theory of Section 7.

5565: 

5566: \shead{8.1}{Initiality preserving operators and the HEP}

5567: Assume now (as in \S3.1) that

5568: \Sigp\ and \Sigpp\ are N-standard signatures with 

5569: \,$\Sig \subset \Sig' \subset \Sigpp$,

5570: \,and \,$\Phi : \NStdAlgSig \to \NStdAlg(\Sig')$

5571: \, is an expanding operator over \Sig.

5572: Recall Definitions 5 and 7 in \S3.1.

5573: 

5574: \Defn1

5575: \Ph\ is {\it initiality preserving\/}

5576: ({\it w.r.t\. \Sig\ and\/} \Sigp)

5577: iff for all 

5578: \,$\KK \subseteq \NStdAlgSig$

5579: \,and \,$A \in \NStdAlgSig$,

5580: \ $A$ is initial in \KK\ iff \APhi\ is initial in \KKPhi.

5581: \endpr

5582: 

5583: \Lemman1

5584: Suppose \Ph\ is initiality preserving,

5585: and \,\SigpTp\ \,specifies \Ph\ uniformly over \Sig.

5586: Then for any \Sig-theory $T$ and N-standard \Sig-algebra $A$,

5587: $$

5588: A \cong \INSSigT \ \ifff

5589: \ \APhi \cong \INS(\Sigp, \, T + T').

5590: $$

5591: \endpr

5592: 

5593: \newpage

5594: 

5595: \Lemman2

5596: Suppose 

5597: \,$\Phi(A) = \Psi(A)\redSigp$

5598: \,for all \,$A \in \NStdAlgSig$,

5599: \,where 

5600: $$

5601: \Psi : \NStdAlgSig \to \NStdAlg(\Sig'')

5602: $$

5603: is an expanding operator 

5604: which is initiality preserving w.r.t\. \Sig\ and \Sigpp.

5605: Then 

5606: \Ph\ is initiality preserving, and

5607: for any \Sigpp-theory \,$T''$ \,and N-standard \Sig-algebra $A$,

5608: if \,\SigppTpp\ \,specifies \Ps\ uniformly over 

5609: \Sig, \,then

5610: \,\SigppTpp\ \,specifies \Ph\ uniformly over \Sig\

5611: with hidden sorts and/or functions; and

5612: for any \Sig-theory T\ \,and N-standard \Sig-algebra $A$,

5613: $$

5614: \align

5615: A \cong \INSSigT \

5616: &\ifff

5617: \ \APsi \cong \INS(\Sigpp, T+T'') \\

5618: &\ifff

5619: \ \APhi \cong \INS(\Sigpp, T+T'')\redSigp \\

5620: &\ifff

5621: \ \APhi \cong \Init\bigl(\Sigp, \,\NStdAlg(\Sigpp, T + T'') \redSigp \,\bigr).

5622: \endalign

5623: $$

5624: \endpr

5625: 

5626: \Pf

5627: From Lemma 1.

5628: \endpf

5629: 

5630: \Defn2

5631: \Ph\ has the {\it homomorphism extension property (HEP) 

5632: (w.r.t\. \Sig\ and \Sigp)}

5633: iff every homomorphism \ $h: A \to B$ \ between N-standard \Sig-algebras

5634: can be {\it extended uniquely} to a homomorphism

5635: \ $h^\Phi : \APhi \to B^\Phi$

5636: \ between their images under \Ph.

5637: \endpr

5638: 

5639: \Lemman3

5640: If \Ph\ has the HEP,

5641: then \Ph\ is initiality preserving.

5642: \endpr

5643: 

5644: We will apply the above theory to three cases:

5645: array specifications in \S8.2, and specifications

5646: for PR and \muPRx\ computable functions in \S8.3 and \S8.4 respectively.

5647: 

5648: 

5649: 

5650: \shead{8.2}{Initial algebra specification of array algebras}

5651: Recall the array specification 

5652: \,$(\Sigx,\,\ArrAxSig)$ \,defined in \S3.2.

5653: 

5654: \Lemman1

5655: The array construction \,$A \mapsto \Ax$ 

5656: \,(\S1.6) has the HEP, and (hence) is initiality preserving.

5657: \endpr

5658: 

5659: \Lemman2

5660: For any N-standard \Sig-algebra $A$ and \Sig-theory $T$:

5661: $$

5662: A \cong \INSSigT \ \ifff \  \Ax \cong \INS(\Sigx, \, T + \ArrAxSig).

5663: $$

5664: \endpr

5665: 

5666: \Pf

5667: By \S8.1, Lemma 1, and \S3.2, Theorem 1.

5668: \endpf

5669: 

5670: Of particular interest is the case that $T$ is a {\it conditional BU

5671: equational\/} theory:

5672: 

5673: 

5674: \Thmn1

5675: If a \Sig-algebra $A$ has an initial N-standard algebra specification 

5676: by a set of conditional BU equations,

5677: then so does \Ax.

5678: Moreover, if the specification for $A$ is finite, 

5679: with $e$ axioms (say),

5680: then so is that for \Ax, with at most

5681: \,$e + 8s$\,\ axioms, 

5682: where $s$ is the number of sorts in \Sig.

5683: \endpr

5684: 

5685: Next, from the BU elimination theorem

5686: for initial algebras (Theorem 5 in Section 7)

5687: we can reduce such a specification for \Ax\ further to one with

5688: {\it conditional equations} only.

5689: 

5690: 

5691: \Thmn2

5692: If a \Sig-algebra $A$ has an initial N-standard algebra specification 

5693: by a set of conditional equations,

5694: then so does \Ax\

5695: (with hidden sorts and functions).

5696: Moreover, if the specification for $A$ is finite, 

5697: with $e$ axioms (say),

5698: then so is that for \Ax, with at most

5699: \,$e + 12s$\,\ axioms, where $s$ is the number of sorts in \Sig.

5700: \endpr

5701: 

5702: \Pf

5703: First apply Theorem 1

5704: (or Lemma 2) above.

5705: Then replace the equality axiom for \sx\ in \ArrAxSig,

5706: which is a conditional BU \Sigx-equation (\S3.2), by a conditional 

5707: \Sigx-equation, for each \Sig-equality sort $s$ other than \nats,

5708: by BU elimination (Theorem 5 in \S7.7, applied to \Sigx).

5709: \endpf

5710: 

5711: \shead{8.3}{Initial algebra specifications for PR computable functions}

5712: Now we apply the above theory to the results in \S5.1.

5713: 

5714: \Lemman1

5715: For each \,\PRSig\ \,derivation \al,

5716: the operator $(**)$ (\S5.1) has the HEP, and is (therefore)

5717: initiality preserving.

5718: Hence the operator $(*)$ is initiality preserving.

5719: \endpr

5720: 

5721: \Pf

5722: By structural induction on \al.

5723: \endpf

5724: 

5725: Hence, by Theorem 1 in Section 5 and Lemma 2 in \S8.1:

5726: 

5727: \Lemman2

5728: For each \PRSig\ derivation \al,

5729: and for each N-standard \Sig-algebra $A$ and \Sig-theory $T$:

5730: $$

5731: \align

5732: A \cong \INSSigT \

5733: &\ifff

5734: \ (A, \galA, \falA) \,\cong \,\INS(\Sigal, \,T+\Eal) \\

5735: &\ifff

5736: \ \AfalA \,\cong \,\INS(\Sigal, \,T+\Eal)\redSigf \\

5737: &\ifff

5738: \ \AfalA \,\cong 

5739: \,\Init\bigl(\Sigf, \,\NStdAlg(\Sigal, \, T+ \Eal)\redSigf\,\bigr).

5740: \endalign

5741: $$

5742: \rm

5743: Here \,$\Sigf = \Sig \cup \{\fal \}$.

5744: (Remember, \,$\Sigal = \Sig \cup \{ \gal, \fal \}$,

5745: \,where \,\gal\ is the list of auxiliary functions of \al.)

5746: Of particular interest is the case that $T$ is a 

5747: {\it conditional equational} theory:

5748: 

5749: \Thmn3

5750: Let $f$ be a PR function on a \Sig-algebra $A$.

5751: If $A$ has an initial N-standard algebra specification 

5752: by a set of conditional equations,

5753: then so does \Af\

5754: (with hidden functions).

5755: \endpr

5756: 

5757: 

5758: \shead{8.4}{Initial algebra specifications for \muPRx\ computable functions}

5759: We turn to \,\muPRx\ \,computability (\S5.2).

5760: The problem here (as noted in \S5.2)

5761: is that even if the computed function is total,

5762: the auxiliary functions need not be.

5763: However, by applying the totality lemma (\S5.2),

5764: we are able restrict our attention to total derivations.

5765: 

5766: \Lemman1

5767: For each \,\muPRxSig\ \,derivation \al\ 

5768: and each N-standard \Sig-algebra $A$ on which \falA\ is total,

5769: the operator (\xxx) (\S5.2) has the HEP, and is (therefore)

5770: initiality preserving.

5771: Hence the operator $(*)$ (\S5.1) is initiality preserving.

5772: \endpr

5773: 

5774: \Pf

5775: By structural induction on \al.

5776: \endpf

5777: 

5778: Hence, by Theorem 2 in Section 5 and 

5779: Lemma 2 in \S8.1, we have:

5780: 

5781: \Lemman2

5782: For each \muPRxSig\ derivation \al,

5783: each N-standard \Sig-algebra $A$ on which

5784: \falA\ is total, and each \Sig-theory $T$:

5785: $$

5786: \multline

5787: A \isom \INSSigT \ \ifff \\

5788: \AfalA\ \cong \ \Init\bigl

5789: (\Sigf, \,\NStdAlg(\Sigalx, \, T+ \ArrAxSig +\Falhat)\redSigf\,\bigr).

5790: \endmultline

5791: $$

5792: where \alhat\ is the total derivation for 

5793: \,\fal\ \,given by the totality lemma,

5794: and \,\Falhat\ \,is the conditional BU specification for \alhat.

5795: \endpr

5796: Here, as before,  \,$\Sigf = \Sig \cup \{\fal \}$.

5797: Of particular interest are the two cases that $T$ is a 

5798: {\it conditional BU equational} theory,

5799: and a {\it conditional equational} theory.

5800: First, assuming the former:

5801: 

5802: 

5803: \Thmn4

5804: Let $f$ be a total \muPRx\ function on a \Sig-algebra $A$.

5805: If $A$ has an initial N-standard algebra specification \SigF,

5806: where $F$ is a set of conditional BU equations,

5807: then likewise \Af\

5808: has such a specification $(\Sigf,\Ff)$

5809: with hidden sorts and functions,

5810: where \Ff\ is also a set of conditional BU equations.

5811: Moreover, \Ff\

5812: can be obtained by adjoining to $F$

5813: an instantiation \,$\FU(\kbar)$

5814: \,of some universal conditional BU equational specification 

5815: \,$\FU(\ztt)$,

5816: \,which depends only on \Sig\ and the type of $f$.

5817: \endpr

5818: 

5819: The universal specification \,$\FU(\ztt)$

5820: \,in this theorem is obtained as in Theorem 3 in Section 5.

5821: 

5822: Finally, by assuming $T$ in Lemma 2 

5823: is a conditional equational theory,

5824: and applying Theorem 4 above and then 

5825: BU elimination for initial algebras (Theorem 5 in Section 7):

5826: 

5827: 

5828: \Thmn5

5829: Let $f$ be a total \muPRx\ function on a \Sig-algebra $A$.

5830: If $A$ has an initial N-standard algebra specification \SigE,

5831: where $E$ is a set of conditional equations,

5832: then likewise \Af\

5833: has such a specification $(\Sigf,\Ef)$

5834: with hidden sorts and functions,

5835: where \Ef\ is also a set of conditional equations.

5836: Moreover, \Ef\

5837: can be obtained by adjoining to $E$

5838: an instantiation \,$\EU(\kbar)$

5839: \,of some universal conditional equational specification 

5840: \,$\EU(\ztt)$,

5841: \,which depends only on \Sig\ and the type of $f$.

5842: \endpr

5843: 

5844: 

5845: \Shead9{Computability of algebraically specifiable functions}

5846: In this section we prove (partial) converses to the results of 

5847: Section 5.

5848: First we need a definition.

5849: 

5850: \Defn{(Strong specifiability)}

5851: Let \KK\ be a \Sig-class, 

5852: let \,$\Sig' \supseteq \Sig \cup \{ \fs \}$

5853: \,and let $T$ be a \Sigp-theory.

5854: We say that $T$ {\it strongly specifies\/} a family

5855: \,\curly{\fA \mid\AinKK}

5856: \,(possibly with hidden sorts and/or functions)

5857: \,iff

5858: \nl

5859: ($i$) $T$ specifies \,\curly{\fA \mid\AinKK},

5860: \,and further

5861: \nl

5862: ($ii$) for every \,$A,B\in \KK$ \,with \,$B\subalg A$,

5863: \,$f^B = \fA\rest B$.

5864: \endpr

5865: \n

5866: (Here \,$\fA\rest B$ \,denotes the restriction of \fA\ to $B$.)

5867: 

5868: The significance of this concept is seen by rephrasing it 

5869: in either of the following two ways.

5870: 

5871: \Lemman1

5872: Let \KK\ be a \Sig-class, 

5873: let \,$\Sig' \supseteq \Sig \cup \{ \fs \}$

5874: \,and let $T$ be a \Sigp-theory.

5875: $T$ {\it strongly specifies\/} a family

5876: \,\curly{\fA \mid\AinKK}

5877: \,(possibly with hidden sorts and/or functions)

5878: \,iff

5879: \nl

5880: ($i$) $T$ specifies \,\curly{\fA \mid\AinKK},

5881: \,and further \,

5882: \nl

5883: ($ii'$) for every \,$A,B\in \KK$ \,with \,$B\subalg A$,

5884: \,$B$ is closed under \fA.

5885: \endpr

5886: 

5887: 

5888: \Lemman2

5889: Let \KK\ be a \Sig-class which is closed under finitely generated subalgebras,

5890: let \,$\Sig' \supseteq \Sig \cup \{ \fs \}$

5891: \,and let $T$ be a \Sigp-theory.

5892: $T$ {\it strongly specifies\/} a family

5893: \,\curly{\fA \mid\AinKK}

5894: \,(possibly with hidden sorts and/or functions)

5895: \,iff

5896: \nl

5897: ($i$) $T$ specifies \,\curly{\fA \mid\AinKK},

5898: \,and further \,

5899: \nl

5900: ($ii''$) for every \,\AinKK\ 

5901: \,and every finitely generated \,$B\subalg A$,

5902: \,$B$ is closed under \fA.

5903: \endpr

5904: 

5905: We consider algebras and functions specified by

5906: conditional equational theories.

5907: We have to assume now that these theories have

5908: {\it effective axiomatisations\/}:

5909: that the axioms are finite, for example, or 

5910: at least recursively enumerable.

5911: 

5912: We will also make use of Theorem $2^X$ in \S7.6.

5913: Recall the remarks preceding the theorem there,

5914: that the N-term condition for $(\Sig,X)$ follows from either

5915: \,($i$) $X = \nil$;

5916: \,or

5917: \,($ii$) strict N-standardness of \,\Sig,

5918: together with $X$ containing no variables of sort \,\nats\ \,or \,\bools.

5919: 

5920: We will prove two theorems, making each of these assumptions 

5921: in turn.

5922: 

5923: 

5924: \shead{9.1}{Computability of specifiable function on minimal algebras}

5925: We first consider a partial converse,

5926: using Remark 1 on the N-term condition (\S7.6),

5927: that is, restricting our attention

5928: to minimal models (\ie, models in which every element 

5929: is named by a closed term).

5930: We use the notation \,\MinNStdAlgSigT\

5931: \,for the set of minimal N-standard \Sig-models of a theory $T$.

5932: 

5933: \Thmn1

5934: Suppose \Sig\ is N-standard.

5935: Let $E$ be an r.e\. conditional equational \Sig-theory

5936: which in \,\CondEqSig\ is consistent, determines

5937: \,\nats\ \,and \,\bools\ \,and proves 

5938: \NStdAxoSig.

5939: Suppose $\Sig' \supseteq \Sig \cup \{ \fs \}$,

5940: \,and let $E'$ be an r.e\. conditional equational \Sigp-theory 

5941: which strongly specifies \,\curly{f^A\mid \AinMinNStdAlgSigE}

5942: \,(possibly with hidden sorts and/or functions).

5943: Assume also that \,$E+E'$ 

5944: \,determines \,\nats\ \,and \,\bools,  

5945: \,and is conservative over $E$,

5946: \,in \,\CondEqSig,

5947: and also that 

5948: all sorts of \,\dom{f} \,other than \,\bools\ \,are equality sorts.

5949: Then \fA\ is uniformly \,\muPRx\ computable over 

5950: \,\AinMinNStdAlgSigE.

5951: \endpr

5952: 

5953: \Pf

5954: We will describe a pseudo-\WhilexSig\ algorithm

5955: for computing \fA\ uniformly over minimal N-standard \Sig-models $A$ of $E$.

5956: Suppose \,$\fs:\utos$,

5957: \,where \,$u = \tuptimes{s}1n$.

5958: In general, some of the $s_i$ are \nats\ or \bools,

5959: and the others not.

5960: Suppose (w.l.o.g.) that for some $m<n$,

5961: sorts \,$s_{m+1},\dots,s_n$ \,are all either \,\nats\ \,or \,\bools,

5962: \,and sorts \,\tup{s}1m \,are not.

5963: Write \,$u=v\times w$

5964: \,where \,$v=\tuptimes{s}1m$

5965: \,and

5966: \,$w=\tuptimes{s}{m+1}n$.

5967: By assumption, sorts \,\tup{s}1m\ \,are equality sorts.

5968: 

5969: For any \,\AinMinNStdAlgSigE,

5970: we will show how to compute 

5971: $$

5972: f^A: \ A^u \ \to \ A_s.

5973: $$

5974: Choose a tuple 

5975: \,$k= (\tup{k}1{n-m})\in\Aw$

5976: \,(of naturals and truth values), and

5977: consider the function

5978: $$

5979: f_k^A \eqdf f(\ \cdot\ ,\,k): \ \Av \ \to \ \As.

5980: $$

5981: We will show how to compute $f_k^A$

5982: uniformly in the (numerical and boolean) parameters $k$.

5983: 

5984: Let \,$I = \Tbi(\Sig,\,E, \,\CondEqSig)$

5985: %\,the free closed \Sig-algebra for $E$ in \,\CondEqSig,

5986: \,and \,$J = \Tbi(\Sig',\,E+E', \,\CondEq(\Sigp))$

5987: %\,the free closed \Sigp-algebra for \,$E+E'$ \,in \,\CondEq(\Sigp).

5988: \,(recall the definitions in \S7.1).

5989: By the N-Standardness Lemma (and the conservativity assumption

5990: for \,$E+E'$ \,over $E$),

5991: both $I$ and $J$ are N-standard.

5992: (Below we denote elements of these

5993: algebras by `$[t]$', \ie, suitable equivalence classes of terms $t$,

5994: or tuples thereof.

5995: We also write \,\kbar\ \,for the tuple of numerals and/or truth constants

5996: corresponding to $k$.)

5997: 

5998: Note that the identity mapping on \,\TSig\ \,induces a \Sig-homomorphism

5999: $$

6000: \iI:\ I \ \to \ \JSig.

6001: $$

6002: By conservativity of \,$E+E'$ \,over $E$, \,\iI\ \,is injective.

6003: Hence \,$I\subalg \JSig$.

6004: 

6005: Further, the function $f^J$ specified by $E'$ on \,\JSig\ \,is clearly

6006: the same as that defined ``naturally" on $J$ by

6007: \ $f^J([t]) = [\fs(t)]$.

6008: By the strong specification assumption,

6009: $$

6010: f^I = f^J\rest I.

6011: $$

6012: Hence for any closed \Sig-term $t_0$,

6013: $$

6014: f_k^J ([t_0]) \ = \ f_k^I([t_0]) \ = \ [t]

6015: $$

6016: for some closed \Sig-term $t$.

6017: By definition of $J$, this means that the equation

6018: $$

6019: \fs(t_0, \,\kbar) \ = \ t

6020: \tagx

6021: $$

6022: is {\it provable\/} from \,$E+E'$ \,in \,\CondEq(\Sigp).

6023: 

6024: Now take any \,\AinMinNStdAlgSigE,

6025: \,and any \,\ainAv.

6026: Since $A$ is minimal, there is a tuple of closed \Sig-terms \,$t_0:v$ 

6027: \,such that 

6028: \,$t_0^A = a$.

6029: By Theorem 2 of Section 6 applied to \Sigp, there is a \Sigp-homomorphism 

6030: $$

6031: h \,: \ J \ \to \ (A,\,f^A, \,\dots\,)

6032: $$

6033: with \,$h([t_0]) = a$.

6034: Hence, since ($*$) holds in $J$, it also holds in 

6035: (the \Sigp-expansion of) $A$, with \,`\fs' \,interpreted as \fA.

6036: 

6037: This suggests the following algorithm for \,$f_k^A$

6038: with $A$ minimal.

6039: \ With {\it inputs} \ainAv:

6040: first generate all (G\"odel numbers of)

6041: tuples of closed \Sig-terms of type $v$, until you find a tuple $t_0$

6042: with \,$t_0^A=a$. 

6043: (This is where we use computability of equality on type $v$.)

6044: Then

6045: {\it generate} all G\"odel numbers of theorems of $E+E'$

6046: until you find one of the form

6047: \,$\cnr{\fs(t_0) = t}$,

6048: \,for some closed \Sig-term $t$.

6049: Then the {\it output} is \,$t^A$.

6050: 

6051: The search is effective in the 

6052: {\it term evaluation\/} function for closed \Sig-terms in $A$,

6053: by recursive enumerability of $E$ and $E'$.

6054: Further, since term evaluation is \PRx\ computable (\cite[\S4]{tz:hb}),

6055: this algorithm can be formalised as a \,\muPRxSig\ \,derivation 

6056: for \fA,

6057: as desired.

6058: \endpf

6059: 

6060: 

6061: \Remark

6062: The assumption that the sorts of \,\dom{f} \,are equality sorts

6063: can clearly be weakened to the assumption that

6064: equality is

6065: (uniformly over \,\MinNStdAlgSigE)

6066: computable on these sorts.

6067: \endpr

6068: 

6069: 

6070: 

6071: \shead{9.2}{Computability of specifiable function in strictly N-standard

6072: algebras}

6073: We consider a second partial converse,

6074: using Remark 2 

6075: on the N-term condition,

6076: \ie, no free variables of sort \,\nats\ \,or \,\bools,

6077: plus strict N-standardness.

6078: \Thmn2

6079: Suppose \Sig\ is strictly N-standard.

6080: Let $E$ be an r.e\. conditional equational \Sig-theory

6081: which in \,\CondEqSig\ is consistent

6082: and proves \,\NStdAxSig.

6083: Suppose $\Sig' \supseteq \Sig \cup \{ \fs \}$

6084: \,is also strictly N-standard

6085: and proves \,\NStdAx(\Sigp).

6086: Let $E'$ be an r.e\. conditional equational \Sigp-theory 

6087: which strongly specifies \,\curly{f^A\mid \AinNStdAlgSigE}

6088: \,(possibly with hidden sorts and/or functions).

6089: Assume also that \,$E+E'$ \,is conservative over $E$  in \,\CondEq(\Sigp).

6090: Then \,\fA\ \,is uniformly \,\muPRx\ computable over \,\AinNStdAlgSigE.

6091: \endpr

6092: 

6093: \Pf

6094: We will describe a pseudo-\WhilexSig\ algorithm

6095: for computing \fA\ uniformly over \,\AinNStdAlgSigE.

6096: Suppose \,$\fs:\utos$,

6097: \,where \,$u = \tuptimes{s}1n$.

6098: In general, some of the $s_i$ are \nats\ or \bools,

6099: and the others not.

6100: Suppose (w.l.o.g.) that for some $m<n$,

6101: sorts \,$s_{m+1},\dots,s_n$ \,are all either \,\nats\ \,or \,\bools,

6102: \,and sorts \,\tup{s}1m \,are not.

6103: Write \,$u=v\times w$

6104: \,where \,$v=\tuptimes{s}1m$

6105: \,and

6106: \,$w=\tuptimes{s}{m+1}n$.

6107: 

6108: For any \,\AinNStdAlgSigE,

6109: we will show how to compute 

6110: $$

6111: f^A: \ A^u \ \to \ A_s.

6112: $$

6113: Choose a tuple 

6114: \,$k= (\tup{k}1{n-m})\in\Aw$

6115: \,(of naturals and truth values), and

6116: consider the function

6117: $$

6118: f_k^A \eqdf f(\ \cdot\ ,\,k): \ \Av \ \to \ \As.

6119: $$

6120: We will show how to compute $f_k^A$

6121: uniformly in the (numerical and boolean) parameters $k$.

6122: 

6123: Choose a tuple of variables \,$\xtt:v$

6124: (\ie, of the same product type as $a$).

6125: Let \,$I = \Tbi(\Sig,\,\xtt,\,E, \,\CondEqSig)$

6126: %\,the free \Sig-algebra for $E$ in \,\CondEqSig\ \,over \xtt,

6127: \,and \,$J = \Tbi(\Sig',\,\xtt,\,E+E', \,\CondEq(\Sigp))$

6128: \,(recall the definitions in \S7.6).

6129: %\,the free \Sigp-algebra for \,$E+E'$ \,in \,\CondEq(\Sigp) \,over \xtt.

6130: By the strict N-standardness Lemma$^X$ (\S7.6),

6131: both $I$ and $J$ are N-standard.

6132: 

6133: Note that the identity mapping on \,\TSigx\ \,induces a \Sig-homomorphism

6134: $$

6135: \iI:\ I \ \to \ \JSig.

6136: $$

6137: By conservativity of \,$E+E'$ \,over $E$, \,\iI\ \,is injective.

6138: Hence \,$I\subalg \JSig$.

6139: 

6140: Further, the function $f^J$ specified by $E'$ on \,\JSig\ \,is clearly

6141: the same as that defined naturally on $J$ by

6142: \ $f^J([t]) = [\fs(t)]$.

6143: By the strong specification assumption,

6144: $$

6145: f^I \ = \ f^J\rest I.

6146: $$

6147: Hence

6148: $$

6149: f_k^J (\xtt) \ = \ f_k^I(\xtt) \ = \ [t]

6150: $$

6151: for some \,\tinTSigx.

6152: By definition of $J$, this means that the equation

6153: $$

6154: \fs(\xtt, \,\kbar) \ = \ t

6155: \tagx

6156: $$

6157: is {\it provable\/} from \,$E+E'$ \,in \,\CondEq(\Sigp).

6158: %Note that $t$ is a \Sig-term (in \xtt).

6159: 

6160: Now take any \,\AinNStdAlgSigE,

6161: \,and any \,\ainAv.

6162: By Theorem $2^X$ applied to \Sigp, there is a \Sigp-homomorphism 

6163: $$

6164: h \,: \ J \ \to \ (A,\,f^A, \,\dots\,)

6165: $$

6166: where \,$h(\xtt) = a$.

6167: Hence, since ($*$) holds in $J$, it also holds in 

6168: (the \Sigp-expansion of) $A$, with \,`\fs' \,interpreted as \fA\

6169: and $a$ assigned to \,\xtt.

6170: 

6171: This suggests the following algorithm for \,$f_k^A$.

6172: \ With {\it inputs} \ainAv:

6173: {\it generate} all G\"odel numbers of theorems of $E+E'$

6174: until you find one of the form

6175: \,$\cnr{\fs(\xtt) = t}$,

6176: \,for some \Sig-term $t$ (in \xtt).

6177: This search is effective, by recursive enumerability of $E$ and $E'$.

6178: Then the {\it output} is 

6179: the {\it evaluation of the term\/} $t$ in $A$ with $a$ assigned to \,\xtt.

6180: 

6181: Since term evaluation is \PRx\ computable \cite[\S4]{tz:hb},

6182: this algorithm can be formalised as a \,\muPRxSig\ \,derivation for \fA,

6183: as desired.

6184: \endpf

6185: 

6186: \Remarks

6187: \itemm{(1)}

6188: The above algorithm gives,

6189: for each tuple of numerical and boolean arguments $k$,

6190: a {\it fixed\/} term \,\tinTSigx\ 

6191: \,as the value of \,$f_k^A(a)$

6192: \,for all \,\AinNStdAlgSigE\

6193: \,and all \,\ainAv.

6194: \itemm{(2)}

6195: Theorems similar to Theorems 1 and 2 can be formulated for conditional 

6196: BU equational theories and specifications,

6197: using a variation of Theorem 3 (instead of Theorem 2) in Section 7.

6198: \endpr

6199: 

6200: \mn

6201: 

6202: \itemm{\bf 9.3}

6203: {\bf Significance of strong specifiability;

6204: \ Equivalence of specifiability and 

6205: \linebreak

6206: computability}

6207: \sn

6208: We want to combine some of the above results into an

6209: equivalence result between computability and specifiability.

6210: 

6211: Note that by the Locality Theorem for \While\ \,computations

6212: \cite[\S2.8]{tz:hb}, if $f$ is \,\muPRx\

6213: computable on an algebra $A$,

6214: then any subalgebra of $A$ is closed under $f$.

6215: This suggests the following formulations for equivalence theorems,

6216: which are simple consequences of the above theorems

6217: and the lemmas on strong specifiability at the beginning of this section.

6218: 

6219: We give one formulation (Theorem 3) for minimal algebras (\cf\ Theorem 1),

6220: and another (Theorem 4) for strictly N-standard algebras (\cf\ Theorem 2).

6221: 

6222: \Thmn3

6223: Suppose \Sig\ is N-standard.

6224: Let $E$ be an r.e\.   conditional equational \Sig-theory,

6225: which in \,\CondEqSig\  

6226: \,is consistent, determines \,\nats\ \,and \,\bools\

6227: \,and proves \,\NStdAxoSig.

6228: Let \,$\fbi = \ang{\fA\mid\AinMinNStdAlgSigE}$

6229: \,be a family of functions on 

6230: \MinNStdAlgSigE.

6231: Assume that all sorts of \,\dom{\fbi} 

6232: \,other than \,\bools\ \,are equality sorts.

6233: Then the following are equivalent:

6234: \itemm{($i$)}

6235: \fbi\ \,is \,\muPRx\ computable uniformly on \,\MinNStdAlgSigE;

6236: \itemm{($ii$)}

6237: \fbi\ \,is strongly specifiable uniformly on \,\MinNStdAlgSigE,

6238: \,with hidden sorts and functions,

6239: by a finite set of conditional equations

6240: which (together with $E$) is conservative over $E$ in \CondEqSig.

6241: %\itemm{($iii$)}

6242: %\f\ \,is strongly specifiable uniformly on \,\NStdAlgSigE,

6243: %\,with hidden sorts and functions,

6244: %by a finite set of conditional BU equations

6245: %which (together with $T$) is conservative for equations 

6246: %over $T$ in \CondBUEqSig.

6247: \endpr

6248: 

6249: 

6250: \Thmn4

6251: Suppose \Sig\ is strictly N-standard.

6252: Let $E$ be an r.e\.   conditional equational \Sig-theory,

6253: which in \,\CondEqSig\  

6254: \,is consistent

6255: and proves \,\NStdAxoSig.

6256: Let \,$\fbi = \ang{\fA\mid\AinNStdAlgSigE}$

6257: \,be a family of functions on 

6258: \,\NStdAlgSigE.

6259: Then the following are equivalent:

6260: \itemm{($i$)}

6261: \fbi\ \,is \,\muPRx\ computable uniformly on \,\NStdAlgSigE;

6262: \itemm{($ii$)}

6263: \fbi\ \,is strongly specifiable uniformly on \,\NStdAlgSigE,

6264: \,with hidden sorts and functions,

6265: by a finite set of conditional equations

6266: which (together with $E$) is conservative over $E$ in \CondEqSig,

6267: and such that the signature of these equations is also strictly N-standard.

6268: \endpr

6269: 

6270: \Remarkn{\,(Herbrand-G\"odel computability on \NN)}

6271: %Definability of functions on many-sorted structures

6272: %by conditional (BU) equations

6273: %is a generalisation of the notion of {\it Herbrand-G\"odel computability}

6274: %on \NN.

6275: The above theorem 

6276: generalises the classical equivalence result on \,\NNN\ \cite{kleene:im}.

6277: %strengthening Remark 3 in \S5.2.

6278: \endpr

6279: 

6280: 

6281: \Shead{10}{Concluding remarks and future directions}

6282: \sshead{10.1}{Computation on Topological Data Types}

6283: We have extended the theory of algebraic specifications

6284: from the world of countable computable algebras to that of

6285: all algebras, and especially metric algebras, by means of

6286: abstract computability theory. Topological data types and

6287: algebraic specifications play a fundamental role in many areas of

6288: computing, including semantics and scientific computation.

6289: 

6290: Our main theorems concern the transformation of abstract

6291: algorithms to algebraic specifications and provide some basic

6292: techniques for the theory of specifying and verifying

6293: abstract computations. An obvious question is:

6294: 

6295: {\displaytext

6296: Under what circumstances can the conditional equations be

6297: replaced by equations in our theory?

6298: 

6299: }

6300: However, the converse results on the derivation of

6301: algorithms from algebraic specifications need strengthening

6302: to provide completeness or equivalence theorems. Improving

6303: results in the reverse direction

6304: is an important problem,

6305: as stated in the Introduction.

6306: There is much more to this topic than the results 

6307: in Section 9.

6308: A key technical problem in this area is:

6309: 

6310: {\displaytext

6311: To develop general techniques for solving equations, conditional

6312: equations and other algebraic formulae in topological algebras.

6313: 

6314: }

6315: 

6316: In semantics, for example, special cases of the problem are

6317: common.  Semantic modelling makes heavy use of fixed-point

6318: equations. One thinks of the introduction of metric methods

6319: into semantics by M. Nivat 

6320: (see \cite{nivat,arnold-nivat:tcs,arnold-nivat:fi}), or their use in

6321: concurrency theory by De Bakker and others

6322: \cite{db-z,db-rutten,db-devink}.

6323: Studies of the methods of equation

6324: solving in ultrametric algebras, including equivalence

6325: between metric, algebraic and domain-theoretic techniques,

6326: are in Stoltenberg-Hansen and Tucker 

6327: \cite{stolt-jvt88,stolt-jvt91,stolt-jvt93};

6328: see also \cite{stolt94}.

6329: 

6330: In scientific computation, numerical methods are concerned

6331: with obtaining computable solutions from differential and

6332: integral equations. Mathematical models of systems in the

6333: world are specified by sets of equations, from which

6334: algorithms are sought to compute their solutions and hence

6335: to simulate the system. Our main theorems and examples in

6336: Section 5 show the opposite: if a system can be

6337: approximately simulated on a computer then there exist

6338: algebraic specifications that capture the system's behaviour. Such

6339: results seem to be new and, in our view, draw loci that

6340: help delimit the computability theory of physical

6341: systems. We conjecture that it is possible

6342: 

6343: {\sl

6344: To show that certain parts of the theory of numerical

6345: approximation of differential and integral equations are special

6346: instances of a general theory of algebraic specifications.

6347: 

6348: }

6349: This is an exciting and difficult problem with many obvious applications.

6350: 

6351: Given the wealth of algorithms and theory in numerical

6352: methods, it seems to us that relatively little is known

6353: about the computational and logical scope and limits of

6354: equations, the classical mathematical methods of science.

6355: Progress in the area has awaited the creation of stable

6356: computation theories for topological data types. Over the

6357: past decade, computability theory for topological spaces

6358: and algebras has developed dramatically. Several general

6359: approaches have produced deep results and have been shown

6360: to be equivalent. Some approaches are

6361: 

6362: \bull

6363: metric spaces \cite{moscho64},

6364: \bull

6365: axiomatic computation structures \cite{pourel-richards},

6366: \bull

6367: type two effectivity \cite{weih:book},

6368: \bull

6369: algebraic domain representability \cite{stolt-jvt88,stolt-jvt95,blanck97},

6370: \bull

6371: continuous domain representability \cite{edalat97},

6372: \bull

6373: abstract computability \cite{tz:top,tz:hb,bss}

6374: 

6375: The equivalence of the first five approaches is proved in

6376: \cite{stolt-jvt99:tcs}.

6377: The equivalence of all these with the last one is proved in

6378: \cite{brattka97,brattka:thesis,tz:top,tz:abs-conc}.

6379: 

6380: However, this computability theory needs to be complemented

6381: by a logical theory which includes equation solving in topological

6382: algebras.

6383: 

6384: 

6385: \shead{10.2}{Theory of computable data types}

6386: The theory of algebraic specifications of computable

6387: (semicomputable, and cosemicomputable) data types contains

6388: many techniques for proving special properties of algebraic

6389: specifications, and showing the equivalence or

6390: non-equivalence of specification methods.  Can some of

6391: these results be generalised? We believe the answer is yes,

6392: but not without much further work. Many results depend on

6393: special techniques of classical computability theory on the

6394: natural numbers. The theory for computable algebras uses

6395: representations by recursive algebras of numbers. It is

6396: possible to make a representation theory for topological

6397: algebras based on Baire space \,$\NN^\NN$ \,using the {\it type two

6398: effectivity\/} methods of Klaus Weihrauch \cite{weih:book}.

6399: The use of the Diophantine Theorem  for r.e\. sets is more

6400: difficult: the theory of r.e\. sets in abstract

6401: computability differs from the classical case, and no

6402: Diophantine Theorem is known (even for minimal algebras).

6403: 

6404: Since abstract computability theory is uniform over classes

6405: of algebras, our results on specifications are uniform,

6406: yielding parameterised specifications. As we have seen,

6407: this process is surprisingly delicate because it leads to

6408: questions about standardness. In abstract computations it

6409: is natural to augment an algebra by basic data types such

6410: as booleans, naturals and finite arrays. These have an

6411: effect on the axiomatisations.  There are other important

6412: additional types, of both theoretical and practical

6413: interest, that may be used to augment a given data type

6414: and are in need of a standard algebraic

6415: specification theory, including:

6416: \itemm{$(i)$}

6417: infinite streams (necessary for developing the theory of interactive systems);

6418: \itemm{$(ii)$}

6419: real numbers (necessary for developing the theory of metric algebras

6420: and normed linear spaces).

6421: 

6422: \n

6423: An attempt to extend the specification methods

6424: of this paper to both these data types,

6425: using infinitary equational specifications,

6426: is made in \cite{tz:fef} 

6427: 

6428: 

6429: Finally, we note there are several other basic properties

6430: of specifications in need of investigation, especially term

6431: rewriting properties.

6432: 

6433: 

6434: \bigskip

6435: \bigskip

6436: \goodbreak

6437: 

6438: \cbb{References}\bigskip

6439: 

6440: \bibliographystyle{alpha}

6441: \bibliography{abbrev,bib}

6442: 

6443: \enddocument

6444: