TODO

   1 -*- outline -*-
   2
   3 * Header guards
   4
   5 From Franc,ois: should we keep the directory part in the CPP guard?
   6
   7
   8 * Yacc.c: CPP Macros
   9
  10 Do some people use YYPURE, YYLSP_NEEDED like we do in the test suite?
  11 They should not: it is not documented.  But if they need to, let's
  12 find something clean (not like YYLSP_NEEDED...).
  13
  14
  15 * Installation
  16
  17 * Documentation
  18 Before releasing, make sure the documentation ("Understanding your
  19 parser") refers to the current `output' format.
  20
  21 * lalr1.cc
  22 ** vector
  23 Move to using vector, drop stack.hh.
  24
  25 ** I18n
  26 Catch up with yacc.c.
  27
  28 * Report
  29
  30 **  GLR
  31 How would Paul like to display the conflicted actions?  In particular,
  32 what when two reductions are possible on a given lookahead token, but one is
  33 part of $default.  Should we make the two reductions explicit, or just
  34 keep $default?  See the following point.
  35
  36 ** Disabled Reductions
  37 See `tests/conflicts.at (Defaulted Conflicted Reduction)', and decide
  38 what we want to do.
  39
  40 ** Documentation
  41 Extend with error productions.  The hard part will probably be finding
  42 the right rule so that a single state does not exhibit too many yet
  43 undocumented ``features''.  Maybe an empty action ought to be
  44 presented too.  Shall we try to make a single grammar with all these
  45 features, or should we have several very small grammars?
  46
  47 ** --report=conflict-path
  48 Provide better assistance for understanding the conflicts by providing
  49 a sample text exhibiting the (LALR) ambiguity.  See the paper from
  50 DeRemer and Penello: they already provide the algorithm.
  51
  52 ** Statically check for potential ambiguities in GLR grammars.  See
  53 <http://www.i3s.unice.fr/~schmitz/papers.html#expamb> for an approach.
  54
  55
  56 * Extensions
  57
  58 ** Labeling the symbols
  59 Have a look at the Lemon parser generator: instead of $1, $2 etc. they
  60 can name the values.  This is much more pleasant.  For instance:
  61
  62        exp (res): exp (a) '+' exp (b) { $res = $a + $b; };
  63
  64 I love this.  I have been bitten too often by the removal of the
  65 symbol, and forgetting to shift all the $n to $n-1.  If you are
  66 unlucky, it compiles...
  67
  68 But instead of using $a etc., we can use regular variables.  And
  69 instead of using (), I propose to use `:' (again).  Paul suggests
  70 supporting `->' in addition to `:' to separate LHS and RHS. In other
  71 words:
  72
  73        r:exp -> a:exp '+' b:exp { r = a + b; };
  74
  75 That requires an significant improvement of the grammar parser.  Using
  76 GLR would be nice.  It also requires that Bison know the type of the
  77 symbols (which will be useful for %include anyway).  So we have some
  78 time before...
  79
  80 Note that there remains the problem of locations: `@r'?
  81
  82
  83 ** $-1
  84 We should find a means to provide an access to values deep in the
  85 stack.  For instance, instead of
  86
  87         baz: qux { $$ = $<foo>-1 + $<bar>0 + $1; }
  88
  89 we should be able to have:
  90
  91   foo($foo) bar($bar) baz($bar): qux($qux) { $baz = $foo + $bar + $qux; }
  92
  93 Or something like this.
  94
  95 ** %if and the like
  96 It should be possible to have %if/%else/%endif.  The implementation is
  97 not clear: should it be lexical or syntactic.  Vadim Maslow thinks it
  98 must be in the scanner: we must not parse what is in a switched off
  99 part of %if.  Akim Demaille thinks it should be in the parser, so as
 100 to avoid falling into another CPP mistake.
 101
 102 ** -D, --define-muscle NAME=VALUE
 103 To define muscles via cli.  Or maybe support directly NAME=VALUE?
 104
 105 ** XML Output
 106 There are couple of available extensions of Bison targeting some XML
 107 output.  Some day we should consider including them.  One issue is
 108 that they seem to be quite orthogonal to the parsing technique, and
 109 seem to depend mostly on the possibility to have some code triggered
 110 for each reduction.  As a matter of fact, such hooks could also be
 111 used to generate the yydebug traces.  Some generic scheme probably
 112 exists in there.
 113
 114 XML output for GNU Bison and gcc
 115    http://www.cs.may.ie/~jpower/Research/bisonXML/
 116
 117 XML output for GNU Bison
 118    http://yaxx.sourceforge.net/
 119
 120 * Unit rules
 121 Maybe we could expand unit rules, i.e., transform
 122
 123         exp: arith | bool;
 124         arith: exp '+' exp;
 125         bool: exp '&' exp;
 126
 127 into
 128
 129         exp: exp '+' exp | exp '&' exp;
 130
 131 when there are no actions.  This can significantly speed up some
 132 grammars.  I can't find the papers.  In particular the book `LR
 133 parsing: Theory and Practice' is impossible to find, but according to
 134 `Parsing Techniques: a Practical Guide', it includes information about
 135 this issue.  Does anybody have it?
 136
 137
 138
 139 * Documentation
 140
 141 ** History/Bibliography
 142 Some history of Bison and some bibliography would be most welcome.
 143 Are there any Texinfo standards for bibliography?
 144
 145
 146
 147 * Java, Fortran, etc.
 148
 149
 150 * Coding system independence
 151 Paul notes:
 152
 153         Currently Bison assumes 8-bit bytes (i.e. that UCHAR_MAX is
 154         255).  It also assumes that the 8-bit character encoding is
 155         the same for the invocation of 'bison' as it is for the
 156         invocation of 'cc', but this is not necessarily true when
 157         people run bison on an ASCII host and then use cc on an EBCDIC
 158         host.  I don't think these topics are worth our time
 159         addressing (unless we find a gung-ho volunteer for EBCDIC or
 160         PDP-10 ports :-) but they should probably be documented
 161         somewhere.
 162
 163         More importantly, Bison does not currently allow NUL bytes in
 164         tokens, either via escapes (e.g., "x\0y") or via a NUL byte in
 165         the source code.  This should get fixed.
 166
 167 * --graph
 168 Show reductions.
 169
 170 * Broken options ?
 171 ** %no-parser
 172 ** %token-table
 173 ** Skeleton strategy
 174 Must we keep %no-parser?  %token-table?
 175
 176 * src/print_graph.c
 177 Find the best graph parameters.
 178
 179 * BTYacc
 180 See if we can integrate backtracking in Bison.  Charles-Henri de
 181 Boysson <de-boy_c@epita.fr> is working on this, and already has some
 182 results.  Vadim Maslow, the maintainer of BTYacc was contacted, and we
 183 stay in touch with him.  Adjusting the Bison grammar parser will be
 184 needed to support some extra BTYacc features.  This is less urgent.
 185
 186 ** Keeping the conflicted actions
 187 First, analyze the differences between byacc and btyacc (I'm referring
 188 to the executables).  Find where the conflicts are preserved.
 189
 190 ** Compare with the GLR tables
 191 See how isomorphic the way BTYacc and the way the GLR adjustments in
 192 Bison are compatible.  *As much as possible* one should try to use the
 193 same implementation in the Bison executables.  I insist: it should be
 194 very feasible to use the very same conflict tables.
 195
 196 ** Adjust the skeletons
 197 Import the skeletons for C and C++.
 198
 199 ** Improve the skeletons
 200 Have them support yysymprint, yydestruct and so forth.
 201
 202
 203 * Precedence
 204
 205 ** Partial order
 206 It is unfortunate that there is a total order for precedence.  It
 207 makes it impossible to have modular precedence information.  We should
 208 move to partial orders (sounds like series/parallel orders to me).
 209
 210 ** Correlation b/w precedence and associativity
 211 Also, I fail to understand why we have to assign the same
 212 associativity to operators with the same precedence.  For instance,
 213 why can't I decide that the precedence of * and / is the same, but the
 214 latter is nonassoc?
 215
 216 If there is really no profound motivation, we should find a new syntax
 217 to allow specifying this.
 218
 219 ** RR conflicts
 220 See if we can use precedence between rules to solve RR conflicts.  See
 221 what POSIX says.
 222
 223
 224 * $undefined
 225 From Hans:
 226 - If the Bison generated parser experiences an undefined number in the
 227 character range, that character is written out in diagnostic messages, an
 228 addition to the $undefined value.
 229
 230 Suggest: Change the name $undefined to undefined; looks better in outputs.
 231
 232
 233 * Default Action
 234 From Hans:
 235 - For use with my C++ parser, I transported the "switch (yyn)" statement
 236 that Bison writes to the bison.simple skeleton file. This way, I can remove
 237 the current default rule $$ = $1 implementation, which causes a double
 238 assignment to $$ which may not be OK under C++, replacing it with a
 239 "default:" part within the switch statement.
 240
 241 Note that the default rule $$ = $1, when typed, is perfectly OK under C,
 242 but in the C++ implementation I made, this rule is different from
 243 $<type_name>$ = $<type_name>1. I therefore think that one should implement
 244 a Bison option where every typed default rule is explicitly written out
 245 (same typed ruled can of course be grouped together).
 246
 247 Note: Robert Anisko handles this.  He knows how to do it.
 248
 249
 250 * Warnings
 251 It would be nice to have warning support.  See how Autoconf handles
 252 them, it is fairly well described there.  It would be very nice to
 253 implement this in such a way that other programs could use
 254 lib/warnings.[ch].
 255
 256 Don't work on this without first announcing you do, as I already have
 257 thought about it, and know many of the components that can be used to
 258 implement it.
 259
 260
 261 * Pre and post actions.
 262 From: Florian Krohm <florian@edamail.fishkill.ibm.com>
 263 Subject: YYACT_EPILOGUE
 264 To: bug-bison@gnu.org
 265 X-Sent: 1 week, 4 days, 14 hours, 38 minutes, 11 seconds ago
 266
 267 The other day I had the need for explicitly building the parse tree. I
 268 used %locations for that and defined YYLLOC_DEFAULT to call a function
 269 that returns the tree node for the production. Easy. But I also needed
 270 to assign the S-attribute to the tree node. That cannot be done in
 271 YYLLOC_DEFAULT, because it is invoked before the action is executed.
 272 The way I solved this was to define a macro YYACT_EPILOGUE that would
 273 be invoked after the action. For reasons of symmetry I also added
 274 YYACT_PROLOGUE. Although I had no use for that I can envision how it
 275 might come in handy for debugging purposes.
 276 All is needed is to add
 277
 278 #if YYLSP_NEEDED
 279     YYACT_EPILOGUE (yyval, (yyvsp - yylen), yylen, yyloc, (yylsp - yylen));
 280 #else
 281     YYACT_EPILOGUE (yyval, (yyvsp - yylen), yylen);
 282 #endif
 283
 284 at the proper place to bison.simple. Ditto for YYACT_PROLOGUE.
 285
 286 I was wondering what you think about adding YYACT_PROLOGUE/EPILOGUE
 287 to bison. If you're interested, I'll work on a patch.
 288
 289 * Better graphics
 290 Equip the parser with a means to create the (visual) parse tree.
 291
 292 -----
 293
 294 Copyright (C) 2001, 2002, 2003, 2004, 2006 Free Software Foundation,
 295 Inc.
 296
 297 This file is part of Bison, the GNU Compiler Compiler.
 298
 299 Bison is free software; you can redistribute it and/or modify
 300 it under the terms of the GNU General Public License as published by
 301 the Free Software Foundation; either version 2, or (at your option)
 302 any later version.
 303
 304 Bison is distributed in the hope that it will be useful,
 305 but WITHOUT ANY WARRANTY; without even the implied warranty of
 306 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 307 GNU General Public License for more details.
 308
 309 You should have received a copy of the GNU General Public License
 310 along with Bison; see the file COPYING.  If not, write to
 311 the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
 312 Boston, MA 02110-1301, USA.