* src/state.h, src/state.c (state_rule_lookaheads_print): New.
[bison.git] / TODO
CommitLineData
416bd7a9
MA
1-*- outline -*-
2
efea6231
AD
3* URGENT: Documenting C++ output
4Write a first documentation for C++ output.
5
3ae2b51f
AD
6* value_components_used
7Was defined but not used: where was it coming from? It can't be to
8check if %union is used, since the user is free to $<foo>n on her
9union, doesn't she?
10
efea6231
AD
11* yyerror, yyprint interface
12It should be improved, in particular when using Bison features such as
13locations, and YYPARSE_PARAMS. For the time being, it is recommended
14to #define yyerror and yyprint to steal internal variables...
15
ec3bc396 16* documentation
efea6231 17Explain $axiom (and maybe change its name: BTYacc names it `goal',
d7215705
AD
18byacc `$accept' probably based on AT&T Yacc, Meta `Start'...).
19Complete the glossary (item, axiom, ?).
ec3bc396 20
d7215705
AD
21* Error messages
22Some are really funky. For instance
23
24 type clash (`%s' `%s') on default action
25
26is really weird. Revisit them all.
27
28* Report documentation
ec3bc396
AD
29Extend with error. The hard part will probably be finding the right
30rule so that a single state does not exhibit to many yet undocumented
31``features''. Maybe an empty action ought to be presented too. Shall
32we try to make a single grammar with all these features, or should we
33have several very small grammars?
34
d7215705 35* Documentation
ec3bc396
AD
36Some history of Bison and some bibliography would be most welcome.
37Are there any Texinfo standards for bibliography?
38
6cbfbcc5
AD
39* Several %unions
40I think this is a pleasant (but useless currently) feature, but in the
41future, I want a means to %include other bits of grammars, and _then_
42it will be important for the various bits to define their needs in
43%union.
76551463 44
5c0a0514
AD
45When implementing multiple-%union support, bare the following in mind:
46
47- when --yacc, this must be flagged as an error. Don't make it fatal
48 though.
49
50- The #line must now appear *inside* the definition of yystype.
51 Something like
52
53 {
54 #line 12 "foo.y"
55 int ival;
56 #line 23 "foo.y"
57 char *sval;
58 }
59
ec3bc396
AD
60* --report=conflict-path
61Provide better assistance for understanding the conflicts by providing
d7215705
AD
62a sample text exhibiting the (LALR) ambiguity. See the paper from
63DeRemer and Penello: they already provide the algorithm.
308a2f76 64
eaff5ee3 65* Coding system independence
4358321a 66Paul notes:
eaff5ee3
AD
67
68 Currently Bison assumes 8-bit bytes (i.e. that UCHAR_MAX is
69 255). It also assumes that the 8-bit character encoding is
70 the same for the invocation of 'bison' as it is for the
71 invocation of 'cc', but this is not necessarily true when
72 people run bison on an ASCII host and then use cc on an EBCDIC
73 host. I don't think these topics are worth our time
74 addressing (unless we find a gung-ho volunteer for EBCDIC or
75 PDP-10 ports :-) but they should probably be documented
76 somewhere.
77
8b3ba7ff
AD
78* Output directory
79Akim:
80
81| I consider this to be a bug in bison:
82|
83| /tmp % mkdir src
84| /tmp % cp ~/src/bison/tests/calc.y src
85| /tmp % mkdir build && cd build
86| /tmp/build % bison ../src/calc.y
87| /tmp/build % cd ..
88| /tmp % ls -l build src
89| build:
90| total 0
91|
92| src:
93| total 32
94| -rw-r--r-- 1 akim lrde 27553 oct 2 16:31 calc.tab.c
95| -rw-r--r-- 1 akim lrde 3335 oct 2 16:31 calc.y
96|
97|
98| Would it be safe to change this behavior to something more reasonable?
99| Do you think some people depend upon this?
100
101Jim:
102
103Is it that behavior documented?
104If so, then it's probably not reasonable to change it.
105I've Cc'd the automake list, because some of automake's
106rules use bison through $(YACC) -- though I'll bet they
107all use it in yacc-compatible mode.
108
109Pavel:
110
111Hello, Jim and others!
112
113> Is it that behavior documented?
114> If so, then it's probably not reasonable to change it.
115> I've Cc'd the automake list, because some of automake's
116> rules use bison through $(YACC) -- though I'll bet they
117> all use it in yacc-compatible mode.
118
119Yes, Automake currently used bison in Automake-compatible mode, but it
120would be fair for Automake to switch to the native mode as long as the
121processed files are distributed and "missing" emulates bison.
122
123In any case, the makefiles should specify the output file explicitly
124instead of relying on weird defaults.
125
126> | src:
127> | total 32
128> | -rw-r--r-- 1 akim lrde 27553 oct 2 16:31 calc.tab.c
129> | -rw-r--r-- 1 akim lrde 3335 oct 2 16:31 calc.y
130
131This is not _that_ ugly as it seems - with Automake you want to put
132sources where they belong - to the source directory.
133
134> | This is not _that_ ugly as it seems - with Automake you want to put
135> | sources where they belong - to the source directory.
136>
137> The difference source/build you are referring to is based on Automake
138> concepts. They have no sense at all for tools such as bison or gcc
139> etc. They have input and output. I do not want them to try to grasp
140> source/build. I want them to behave uniformly: output *here*.
141
142I realize that.
143
144It's unfortunate that the native mode of Bison behaves in a less uniform
145way than the yacc mode. I agree with your point. Bison maintainters may
146want to fix it along with the documentation.
147
148
fa770c86
AD
149* Unit rules
150Maybe we could expand unit rules, i.e., transform
151
152 exp: arith | bool;
153 arith: exp '+' exp;
154 bool: exp '&' exp;
155
156into
157
158 exp: exp '+' exp | exp '&' exp;
159
160when there are no actions. This can significantly speed up some
d7215705
AD
161grammars. I can't find the papers. In particular the book `LR
162parsing: Theory and Practice' is impossible to find, but according to
163`Parsing Techniques: a Practical Guide', it includes information about
164this issue. Does anybody have it?
fa770c86 165
51dec47b
AD
166* Stupid error messages
167An example shows it easily:
168
169src/bison/tests % ./testsuite -k calc,location,error-verbose -l
170GNU Bison 1.49a test suite test groups:
171
172 NUM: FILENAME:LINE TEST-GROUP-NAME
173 KEYWORDS
174
175 51: calc.at:440 Calculator --locations --yyerror-verbose
176 52: calc.at:442 Calculator --defines --locations --name-prefix=calc --verbose --yacc --yyerror-verbose
177 54: calc.at:445 Calculator --debug --defines --locations --name-prefix=calc --verbose --yacc --yyerror-verbose
178src/bison/tests % ./testsuite 51 -d
179## --------------------------- ##
180## GNU Bison 1.49a test suite. ##
181## --------------------------- ##
182 51: calc.at:440 ok
183## ---------------------------- ##
184## All 1 tests were successful. ##
185## ---------------------------- ##
186src/bison/tests % cd ./testsuite.dir/51
187tests/testsuite.dir/51 % echo "()" | ./calc
1881.2-1.3: parse error, unexpected ')', expecting error or "number" or '-' or '('
fa770c86
AD
189
190* read_pipe.c
191This is not portable to DOS for instance. Implement a more portable
192scheme. Sources of inspiration include GNU diff, and Free Recode.
193
aef1ffd5
AD
194* Memory leaks in the generator
195A round of memory leak clean ups would be most welcome. Dmalloc,
196Checker GCC, Electric Fence, or Valgrind: you chose your tool.
197
198* Memory leaks in the parser
199The same applies to the generated parsers. In particular, this is
200critical for user data: when aborting a parsing, when handling the
201error token etc., we often throw away yylval without giving a chance
202of cleaning it up to the user.
203
bcb05e75
MA
204* --graph
205Show reductions. []
206
704a47c4 207* Broken options ?
c3995d99 208** %no-lines [ok]
04a76783 209** %no-parser []
fbbf9b3b 210** %pure-parser []
04a76783
MA
211** %token-table []
212** Options which could use parse_dquoted_param ().
213Maybe transfered in lex.c.
214*** %skeleton [ok]
215*** %output []
216*** %file-prefix []
217*** %name-prefix []
ec93a213 218
fbbf9b3b 219** Skeleton strategy. []
c3a8cbaa
MA
220Must we keep %no-parser?
221 %token-table?
fbbf9b3b 222*** New skeletons. []
416bd7a9 223
c111e171 224* src/print_graph.c
31b53af2 225Find the best graph parameters. []
63c2d5de
MA
226
227* doc/bison.texinfo
1a4648ff 228** Update
c3a8cbaa 229informations about ERROR_VERBOSE. []
1a4648ff 230** Add explainations about
c3a8cbaa
MA
231skeleton muscles. []
232%skeleton. []
eeeb962b 233
704a47c4 234* testsuite
c3a8cbaa
MA
235** tests/pure-parser.at []
236New tests.
0f8d586a
AD
237
238* Debugging parsers
239
240From Greg McGary:
241
242akim demaille <akim.demaille@epita.fr> writes:
243
244> With great pleasure! Nonetheless, things which are debatable
245> (or not, but just `big') should be discuss in `public': something
246> like help- or bug-bison@gnu.org is just fine. Jesse and I are there,
247> but there is also Jim and some other people.
248
249I have no idea whether it qualifies as big or controversial, so I'll
250just summarize for you. I proposed this change years ago and was
251surprised that it was met with utter indifference!
252
253This debug feature is for the programs/grammars one develops with
254bison, not for debugging bison itself. I find that the YYDEBUG
255output comes in a very inconvenient format for my purposes.
256When debugging gcc, for instance, what I want is to see a trace of
257the sequence of reductions and the line#s for the semantic actions
258so I can follow what's happening. Single-step in gdb doesn't cut it
259because to move from one semantic action to the next takes you through
260lots of internal machinery of the parser, which is uninteresting.
261
262The change I made was to the format of the debug output, so that it
263comes out in the format of C error messages, digestible by emacs
264compile mode, like so:
265
266grammar.y:1234: foo: bar(0x123456) baz(0x345678)
267
268where "foo: bar baz" is the reduction rule, whose semantic action
269appears on line 1234 of the bison grammar file grammar.y. The hex
270numbers on the rhs tokens are the parse-stack values associated with
271those tokens. Of course, yytype might be something totally
272incompatible with that representation, but for the most part, yytype
273values are single words (scalars or pointers). In the case of gcc,
274they're most often pointers to tree nodes. Come to think of it, the
275right thing to do is to make the printing of stack values be
276user-definable. It would also be useful to include the filename &
277line# of the file being parsed, but the main filename & line# should
278continue to be that of grammar.y
279
280Anyway, this feature has saved my life on numerous occasions. The way
281I customarily use it is to first run bison with the traces on, isolate
282the sequence of reductions that interests me, put those traces in a
283buffer and force it into compile-mode, then visit each of those lines
284in the grammar and set breakpoints with C-x SPACE. Then, I can run
285again under the control of gdb and stop at each semantic action.
286With the hex addresses of tree nodes, I can inspect the values
287associated with any rhs token.
288
289You like?
cd6a695e
AD
290
291* input synclines
292Some users create their foo.y files, and equip them with #line. Bison
293should recognize these, and preserve them.
0e95c1dd
AD
294
295* BTYacc
296See if we can integrate backtracking in Bison. Contact the BTYacc
297maintainers.
298
0e95c1dd
AD
299* RR conflicts
300See if we can use precedence between rules to solve RR conflicts. See
301what POSIX says.
302
303* Precedence
304It is unfortunate that there is a total order for precedence. It
305makes it impossible to have modular precedence information. We should
306move to partial orders.
307
3c9160d9
AD
308This will be possible with a Bison parser for the grammar, as it will
309make it much easier to extend the grammar.
310
0e95c1dd 311* Parsing grammars
3c9160d9 312Rewrite the reader in Flex/Bison. There will be delicate parts, in
9306c70c
AD
313particular, expect the scanner to be hard to write. Many interesting
314features cannot be implemented without such a new reader.
f294a2c2 315
d7215705
AD
316I'm on it! I already have a proto that parses (but the actions are
317not fully written yet). -- Akim
318
69991a58
AD
319* $undefined
320From Hans:
321- If the Bison generated parser experiences an undefined number in the
322character range, that character is written out in diagnostic messages, an
323addition to the $undefined value.
324
325Suggest: Change the name $undefined to undefined; looks better in outputs.
326
327* Default Action
328From Hans:
329- For use with my C++ parser, I transported the "switch (yyn)" statement
330that Bison writes to the bison.simple skeleton file. This way, I can remove
331the current default rule $$ = $1 implementation, which causes a double
332assignment to $$ which may not be OK under C++, replacing it with a
333"default:" part within the switch statement.
334
335Note that the default rule $$ = $1, when typed, is perfectly OK under C,
336but in the C++ implementation I made, this rule is different from
337$<type_name>$ = $<type_name>1. I therefore think that one should implement
338a Bison option where every typed default rule is explicitly written out
339(same typed ruled can of course be grouped together).
340
3c9160d9
AD
341Note: Robert Anisko handles this. He knows how to do it.
342
0164db68
AD
343* Warnings
344It would be nice to have warning support. See how Autoconf handles
345them, it is fairly well described there. It would be very nice to
346implement this in such a way that other programs could use
347lib/warnings.[ch].
348
9306c70c
AD
349Don't work on this without first announcing you do, as I already have
350thought about it, and know many of the components that can be used to
351implement it.
352
69991a58
AD
353* Pre and post actions.
354From: Florian Krohm <florian@edamail.fishkill.ibm.com>
355Subject: YYACT_EPILOGUE
356To: bug-bison@gnu.org
357X-Sent: 1 week, 4 days, 14 hours, 38 minutes, 11 seconds ago
358
359The other day I had the need for explicitly building the parse tree. I
360used %locations for that and defined YYLLOC_DEFAULT to call a function
361that returns the tree node for the production. Easy. But I also needed
362to assign the S-attribute to the tree node. That cannot be done in
363YYLLOC_DEFAULT, because it is invoked before the action is executed.
364The way I solved this was to define a macro YYACT_EPILOGUE that would
365be invoked after the action. For reasons of symmetry I also added
366YYACT_PROLOGUE. Although I had no use for that I can envision how it
367might come in handy for debugging purposes.
76551463 368All is needed is to add
69991a58
AD
369
370#if YYLSP_NEEDED
371 YYACT_EPILOGUE (yyval, (yyvsp - yylen), yylen, yyloc, (yylsp - yylen));
372#else
373 YYACT_EPILOGUE (yyval, (yyvsp - yylen), yylen);
374#endif
375
376at the proper place to bison.simple. Ditto for YYACT_PROLOGUE.
377
378I was wondering what you think about adding YYACT_PROLOGUE/EPILOGUE
379to bison. If you're interested, I'll work on a patch.
380
d7215705
AD
381* Move to Graphviz
382Well, VCG seems really dead. Move to Graphviz instead. Also, equip
383the parser with a means to create the (visual) parse tree.
384
f294a2c2
AD
385-----
386
387Copyright (C) 2001, 2002 Free Software Foundation, Inc.
388
389This file is part of GNU Autoconf.
390
391GNU Autoconf is free software; you can redistribute it and/or modify
392it under the terms of the GNU General Public License as published by
393the Free Software Foundation; either version 2, or (at your option)
394any later version.
395
396GNU Autoconf is distributed in the hope that it will be useful,
397but WITHOUT ANY WARRANTY; without even the implied warranty of
398MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
399GNU General Public License for more details.
400
401You should have received a copy of the GNU General Public License
402along with autoconf; see the file COPYING. If not, write to
403the Free Software Foundation, Inc., 59 Temple Place - Suite 330,
404Boston, MA 02111-1307, USA.