Debian Codegraph @

doc/todo

raw
1 2006-08-19 JoeB:
2 * There is something inconsistent about SAVEDEGREE
3 versus DEGREEPRINTOUTPUT. I suspect that the former
4 is mistakenly investigated instead of the latter,
5 in e. g. GROEBNERFINISH.
6
7 * The GLOBAL variable !#!$NoHigherGbRedNeeded!*
8 either should be put into use (by adding a
9 new minor strategy mode), or removed.
10
11 2006-08-18 JoeB:
12 * In 'real and true' commutative TDEGREVLEX order,
13 with non-homogeneous input, the refined Buchberger
14 criterion does not work. There may be subsequent
15 monomials, which are lower in the most significant
16 variable than is the LeadMon, but which still are
17 lower, because their totaldegrees are. Thus, it
18 is not sure that the power of this variable in the
19 LeadMon also is a factor of the entire polynomial.
20
21 2006-08-09 JoeB, VU:
22 * Make MAXDEG inhibition of calculation of higher
23 critical pairs efficient. (Now, seemingly, the
24 calculated but discarded pairs clutter the heap.)
25
26 2006-08-08 JoeB:
27 * ReducePol (nowadays, at least) returns a genaugredor,
28 NOT a genaugredand! Before changing the protocol,
29 go through the situation a little further. However,
30 I call PreenRedand on the result of ReducePol, in
31 FixNGbe; and it is NOT ALWAYS a noop. (Formerly,
32 the Redand2Redor wasn't performed until FixGbe was
33 applied on all current degree new Gbe's
34 simultaneously, at the end of the calculations
35 for this degree.)
36
37 * Refine RABBIT mode settings: SETRABBIT, GETRABBIT.
38
39 2006-08-07 JoeB:
40 * The SAWS mode incopatibilities now are extended,
41 by NOT making AUTOREDUCEINPUT and e.g.
42 AutoReduceFromTo working in itemwise mode. Probably
43 rather extensive testing is necessary to get SAWS
44 working in ANY mode!
45
46 2006-08-06 JoeB:
47 * MONFACTORP is ONLY used as a boolean, but in the
48 non-commutative case does MUCH more.
49
50 2006-08-05 JoeB:
51 * DestructRedor2Redand is sometimes defined as
52 NILNOOPFCN1, sometimes as IBIDNOOPFCN. The former
53 seems most in accordance with the documentation.
54
55 * The manual documents nowadays non-existing procedures;
56 at least POLALGOUT. Define them, with obsoleteness
57 warnings!
58
59 2005-06-25 JoeB, VU:
60 * It should be possible to stop ANICK at a certain
61 homological degree (altough we go on in totaldegree).
62 Something like SETMAXHOMDEGREE?
63
64 * We should add a dynamic way to add new monoid
65 orders. In this context, some calculation of a
66 new commutative monomial quotient could be useful.
67
68 2005-06-21 VU:
69 * Autosave mode; able to restart calculation after
70 a chrash. Must be coupled with "restore" and
71 "continue". We have to save the following
72 monomials: Leading ones in Groebner basis elements;
73 those in S-polynomials.
74
75 2005-07-16 JoeB, VU:
76 * If e.g. we (in this order) call ANICK,
77 SETNORESOLUTION, and ANICK again, the second call
78 to ANICK does not work. The reason is that only
79 when the anick module is loaded the setup changes
80 are made.
81 One remedy of course could be to move the settings
82 to the beginning of call of ANICK; or to make them
83 explicitly, if ANICKISLOADED. However, this might
84 lessen the flexibility a little.
85
86 * Fix the output mode default in such a way that
87 algebraic output is standard, when outvars are set.
88 Perhaps a new, flexible mode; or just changing the
89 default setting?
90
91 2005-07-15 JoeB:
92 * There are situations where partial results are not
93 exhibited, since no call to FLUSHCHANNEL was
94 included. This seems to be the case in the anick
95 module; and probably in some other cases, where
96 the SC&VU test suite did not produce (complete)
97 output. Insert them!
98
99 * Include tests of positive characteristics. Correct
100 such clisp compilation as is set up without proper
101 macros inserted.
102
103 2005-02-19 JoeB:
104 * If such do not exist, then define procedures for
105 adding to, but not replacing, InPols. Say,
106 ADDLISPFORMINPUT and ADDALGFORMINPUT; the latter NOT
107 accepting a vars statement. DONE in v. 0.988/JoeB
108
109 2005-01-19 JoeB:
110 * The customising procedures should have alternatives
111 working as hooks rather than supplanting the
112 calling procedures. We may also want interruptions
113 (pausing with the possibility of performing a number
114 of dialogue or lisp commands, but possibility to
115 resume calculations), breaking (to top level,
116 possibly by creating an error), and quitting (which
117 is simple to do by means of QUIT). Hooks could be
118 created by separate procedures, or by letting a
119 `reserved' identifier as return value cause `hooking',
120 i.e., continuation with the rest of the calling
121 procedure.
122
123 2004-10-12 JoeB , VU:
124 * (noncommify)(simple)vars x,y;x*y-y^2;
125 yields heap overflow already at degree 99. Why?
126 If in addition the Hilbert series limitation is
127 active, with the limitations set to the true
128 Hilbert series (by means of setting the DEFAULT to
129 (PLUS2 HSDEG 1)), and (of course) PBSERIES is ON,
130 then the run seems to be markedly slower; and it
131 collapses by heap overflow already at degree 80
132 (approximately). Again, why?
133
134 * Check the action of POSLEADCOEFFS. UnSign-ing
135 seems to be automatic in the SparseContents submode,
136 but not at all 'activatible' by modes or switches
137 in the dense case. I think I removed these troubles
138 a long time ago. Perhaps, it is done in e.g.
139 DenseContentChar0SubtractRedand1, in some oldfashioned
140 variant.
141 To-day, UNSIGNLC!'S is called explicitly by aninterf.sl
142 procedures (ultimately called by ResolutionFixcDegEnd);
143 but no-where else.
144
145
146 2004-09-30 JoeB , VU:
147 * Fix merging of doc's from rather different sources.
148 Some rather old doc's contain things never added to
149 the newer ones.
150
151 * Investigate precisely what troubles may arise from
152 running inhomogeneous input. Write one or several
153 strategies for doing this properly. At least one
154 of these should be compatible with ELIM orderings.
155
156 * Re-organise the "safe" stuff into the mode handling.
157 Include handling of constant terms.
158
159 * Incorporate corrections from older versions (???)
160 (alternatively: find and eliminate brand new bugs)
161 concerning repeated series calculation and
162 inner embdim (which seemingly is not quite cleared
163 even by CLEARRING).
164
165 * Check and update SAWS.
166
167 * Make a manual version in a `booklet printable' form.
168 DONE circa 2005-06-15 by SV & VU.
169
170 * Investigate, possibly write new, procedures for solving
171 linear equations.
172
173 * Employ the latter in calculating rational Hilbert series
174 in the non-commutative case.
175
176 2004-05-04 JoeB:
177 * Enter more `free floating' modes into the compact
178 mode handler. This includes: SAWS modifications;
179 MODULE or TWOMODULES module variable numbers;
180 homogeneity.
181
182 * Fix the common lisp independence of compile.sl
183 et cetera.
184
185 2003-11-01 JoeB: Common Lisp now have troubles with
186 lambda expressions. I am not sure of how to overcome
187 this, but I have a memory that we found some (partial?)
188 remedies earlier.
189
190 This touches the new FINDMODECLASHES family of
191 procedures (by means of ModeSettingP), and also
192 employing the procedure alternative for argument
193 to SETHSERIESMINIMUMDEFAULT.
194
195 2003-10-29 JoeB:
196 (* Checksetup: E.g., modules do not work within the
197 commutative case now. Add some "Incompatibilities
198 checker". DONE 2004-05-04.)
199
200 * The DIALOGUE is rather underdeveloped. Test its clisp
201 abilities, inter alia!
202
203 (* CLEARIDEAL removes the variable number information.
204 This should not be so. On the other hand, a new
205 CLEARRING might do this - and much more.
206 [FIXED 2003-11-01] )
207
208 2003-08-28 JoeB:
209 * Move (at least the source files of) the new run
210 test files from bin/Linux_elf to tests.
211
212 * Merge the loose pieces of documentation - such as
213 ReadMe and ReadMe0.91.var (with info on log files);
214 KnownBugs and known.bugs; News, ... .
215
216 * Make a 'make' based version of mkbergman.
217
218 2003-08-18 JoeB: Fix the linear and constant input
219 problems by adding a new mode difference. Some efficient
220 procedures be replaced by safer ones, when linear or
221 constant inputs are acceptable. (May or may not be
222 ord/altweights dependent, simultaneously.)
223
224 Affected procedures include:
225 LC2, ALGFORMINPUT (or one of its auxiliaries),
226
227 2002-10-09 JoeB: Check why (nc) FixNGbe collapses with
228 linear input.
229
230 2002-06-28 JoeB: Move setting of EMBDIM from PBINIT to
231 noncommMonomialInit ... and replace it perhaps by
232 !&!*InnerEmbDim!&!* setting, like in commMonomialInit?
233 Problem, of course: We may do calculations with embdim
234 unset, if PB is off, but not if it is on. Thus perhaps:
235 Make a tentative setting in noncommMonomialInit, and an
236 obligatoric one (if not yet set) in PBSERIES.
237
238 Also, unset !&!*InnerEmbDim!&!* in case of clearing the
239 variables. Actually, this should perhaps be reset when variables
240 are?
241
242 1999-11-26 JoeB: Make the last argument to mkbergman in unix
243 environments facultative; using cd $0 and an adequate
244 cd ../.. ... /.. in order to get to the bmroot and $cwd extract
245 it (at a suggestion from SV). However, warn for the use of this
246 alternative!
247 Check if caseconv now is OK, so that one may apply it even
248 when the xsrc and xauxil directories did exist.
249
250 1999-06-28 JoeB: Make the HSERIESMIN procedures self-loading.
251 Introduce CLEARing of hseriesminima. Make setting hseriesmin
252 strategy to load hseries (if not done).
253
254 1999-06-27 JoeB: According to SV, the single instance of a
255 case non-distinction potential name clash would be
256 noncommMonTimes2 versus noncommMONTIMES2 (in ncmonom.sl).
257 Perhaps should rename one of them, in order to facilitate
258 for others to adapt the bm sources in other settings?
259
260 1997-10-12 JoeB: Add ncCALCPOLRING (or better ncCALCFREERING),
261 which sets INVHILBERTSERIES appropriately, and which in the
262 same cases as in the next note calculates HILBERTSERIES up to
263 MAXDEG.
264
265 1997-10-12 ufn: Suggestion: if Maxdeg was greater than obtained
266 maximum degree in GB, continue to calculate Hilbert(Poincare)series
267 untill Maxdeg.
268
269 1997-09-18 JoeB: Fix weight setting and variable setting in such
270 ways that they are consistent from the user point of view. If
271 necessary, distinguish internal weight lists and dplists from
272 interface forms.
273
274 1997-09-08 JoeB: Fix the non-comm exponent output. Introduce new
275 modifications: a PRINTEXPONENTS flag (default: T); flag deciding
276 whether the exponents be collected or spread. DONE 09-18
277 And a flag or mode setter deciding whether the first or the last
278 variable be the highest in standard orders.
279
280 1997-08-27 JoeB: DOCUMENT setweights. Add getweights, clearweights,
281 printweights. Make it compatible with series calculation, Anick,
282 stagger, (de)homogenisation.
283 Make (de)homogenisation variants in the non-commutative case.
284 Write the file output checker.
285
286 1997-08-15: JoeB: Either include YESP in slext.sl, or exclude it from
287 the module series procedures.
288
289 1997-05-22: JoeB: Investigate if the AddTo!&Char0Lists calls in stg.sl
290 for AutoReduceFix, StagFormSPol, ReduceSubstantialPol, and
291 STAGGERFINISH are adequate.
292
293 1997-04-09: JoeB: Fix a nice and neat way of calculating pbseries etc.
294 to a given degree; and DOCUMENT this!!
295
296 1997-04-07: JoeB: Remove the Char2RedorMonMult definition (or rather,
297 replace the Char0 one by this).
298 I also seem to have started replacing DestructRedandTermPlus by
299 DestructQPolTermPlus, but I have not completed this?
300
301 1997-03-03: JoeB: Check through files outside Char0 for definitions of
302 or occurrences of Char0FOO variants. I believe that RETURNING to Char0
303 might not work now; it should be fixed as above, and by making ALL
304 FOO-out-of-Char0 initiations at the end of char0.sl, by a call to
305 RESTORECHAR0, if possible.
306
307 1996-12-17: Jan S & Ralf: Add possibility to INPUT H series as numerator
308 + denominator degree.
309
310 1996-12-09: Ralf: Add (at least) the orders you wish to use at elimination
311 problems: block orders with specific variables (of various degrees); 2
312 degrevlex blocks being enough.
313 In a Hilbert series driven algorithm, add Moreno's 'criterium' (optionally):
314 If a reduction leads to 0 in the beginning, then they mostly continue
315 in the same manner. Thus, skip such calculations. If not all new Gbe's
316 which should be there are found, in a new pass calculate them to the
317 bitter end.
318
319 1996-11-28 Ralf: Imitate Gotzmann? (Dynamically calculate several
320 series, for to see if by Gotzmanns theorem no more should be calculated.)
321 Likewise, if 0 reached, interrupt!
322
323 1996-11-27:
324 - Fix stagger as `proper' and revertible mode. (Fix stagsubs.sl accordingly.)
325 - Fix file and dynamic prime logarithm table making.
326 - Fix `generic' system saver (in slext.sl).
327 - Integrate the newdomain content into other souce files.
328 - Integrate comm and nc series handling. (Should sermul be integrated, too?)
329 (Hilbert series driven algorithms included.)
330 - Set up proper modes for anick.
331 - Collect documentation from various sources (incl. Lund and ~sanya).
332 Integrate it, on the doc area.
333 - Make easy-to-access handling of several rings, G-bases, series, et
334 cetera simultaneously.
335
336 Most of this should be in the version 1.0!
337 The following probably should wait:
338 - Adopt bergman to the latest Psl and Reduce versions.
339 - Set up proper modes for (V.U. and S.C.) module Gbases et cetera.
340 - Make easy-to-access handling of several rings, G-bases, series, et
341 cetera simultaneously.
342 - Make a general command interpreter tool; to be used in alginput,
343 and as a base of a `macaulay-like' interpreter.
344 - Make bergman available with Common Lisp.
345 - Possibly, make bergman available with C.
346 - Trace strategic submode mode
347 /JoeB