3 # regen_perly.pl, DAPM 12-Feb-04
5 # Copyright (c) 2004, 2005 Larry Wall
7 # Given an input file perly.y, run bison on it and produce
8 # the following output files:
10 # perly.h standard bison header file with minor doctoring of
11 # #line directives plus adding a #ifdef PERL_CORE
13 # perly.tab the parser table C definitions extracted from the bison output
14 # plus an extra table generated by this script.
16 # perly.act the action case statements extracted from the bison output
18 # Note that perly.c is *not* regenerated - this is now a static file which
19 # is not dependent on perly.y any more.
21 # If a filename of the form foo.y is given on the command line, then
22 # this is used instead as the basename for all the files mentioned
25 # Note that temporary files of the form perlytmp.h and perlytmp.c are
26 # created and then deleted during this process
28 # Note also that this script is intended to be run on a UNIX system;
29 # it may work elsewhere but no specific attempt has been made to make it
33 sub usage { die "usage: $0 [ -b bison_executable ] [ file.y ]\n" }
38 BEGIN { require 'regen/regen_lib.pl'; }
42 if (@ARGV >= 2 and $ARGV[0] eq '-b') {
47 my $y_file = shift || 'perly.y';
49 usage unless @ARGV==0 && $y_file =~ /\.y$/;
51 (my $h_file = $y_file) =~ s/\.y$/.h/;
52 (my $act_file = $y_file) =~ s/\.y$/.act/;
53 (my $tab_file = $y_file) =~ s/\.y$/.tab/;
54 (my $tmpc_file = $y_file) =~ s/\.y$/tmp.c/;
55 (my $tmph_file = $y_file) =~ s/\.y$/tmp.h/;
57 # the yytranslate[] table generated by bison is ASCII/EBCDIC sensitive
59 die "$0: must be run on an ASCII system\n" unless ord 'A' == 65;
61 # check for correct version number. The constraints are:
62 # * must be >= 1.24 to avoid licensing issues.
63 # * it must generate the yystos[] table. Version 1.28 doesn't generate
65 # * Must produce output which is extractable by the regexes below
66 # * Must produce the right values.
67 # These last two constraints may well be met by earlier versions, but
68 # I simply haven't tested them yet. If it works for you, then modify
69 # the test below to allow that version too. DAPM Feb 04.
71 my $version = `$bison -V`;
72 unless ($version) { die <<EOF; }
73 Could not find a version of bison in your path. Please install bison.
76 unless ($version =~ /\b(1\.875[a-z]?|2\.[0134])\b/) { die <<EOF; }
78 You have the wrong version of bison in your path; currently 1.875
79 2.0, 2.1, 2.3 or 2.4 is required. Try installing
80 http://ftp.gnu.org/gnu/bison/bison-2.4.1.tar.gz
81 or similar. Your bison identifies itself as:
86 # creates $tmpc_file and $tmph_file
87 my_system("$bison -d -o $tmpc_file $y_file");
89 open my $ctmp_fh, '<', $tmpc_file or die "Can't open $tmpc_file: $!\n";
91 { local $/; $clines = <$ctmp_fh>; }
92 die "failed to read $tmpc_file: length mismatch\n"
93 unless length $clines == -s $tmpc_file;
96 my ($actlines, $tablines) = extract($clines);
98 $tablines .= make_type_tab($y_file, $tablines);
100 my ($act_fh, $tab_fh, $h_fh) = map {
101 open_new($_, '>', { by => $0, from => $y_file });
102 } $act_file, $tab_file, $h_file;
104 print $act_fh $actlines;
106 print $tab_fh $tablines;
110 # Wrap PERL_CORE round the symbol definitions. Also, the
111 # C<#line 30 "perly.y"> confuses the Win32 resource compiler and the
112 # C<#line 188 "perlytmp.h"> gets picked up by make depend, so remove them.
114 open my $tmph_fh, '<', $tmph_file or die "Can't open $tmph_file: $!\n";
116 my $endcore_done = 0;
117 # Token macros need to be generated manually on bison 2.4
118 my $gather_tokens = ($version =~ /\b2\.4\b/ ? undef : 0);
121 print $h_fh "#ifdef PERL_CORE\n" if $. == 1;
122 if (!$endcore_done and /YYSTYPE_IS_DECLARED/) {
123 print $h_fh "#endif /* PERL_CORE */\n";
126 next if /^#line \d+ ".*"/;
127 if (not defined $gather_tokens) {
128 $gather_tokens = 1 if /^\s* enum \s* yytokentype \s* \{/x;
130 elsif ($gather_tokens) {
131 if (/^\# \s* endif/x) { # The #endif just after the end of the token enum
133 $_ .= "\n/* Tokens. */\n$tokens";
136 my ($tok, $val) = /(\w+) \s* = \s* (\d+)/x;
137 $tokens .= "#define $tok $val\n" if $tok;
145 foreach ($act_fh, $tab_fh, $h_fh) {
146 read_only_bottom_close_and_rename($_, ['regen_perly.pl', $y_file]);
159 ^/* YYFINAL[^\n]+\n #optional comment
161 \# \s* define \s* YYFINAL # first #define
162 .*? # other defines + most tables
163 yystos\[\]\s*= # start of last table
165 }\s*; # end of last table
167 or die "Can't extract tables from $tmpc_file\n";
172 switch \s* \( \s* \w+ \s* \) \s* { \s*
176 (?: \s* /\* .*? \*/ \s* )* # optional C-comments
178 \#line [^\n]+"\Q$y_file\E"
183 (?: \s* /\* .*? \*/ \s* )* # optional C-comments
188 \#line[^\n]+\.simple"
193 or die "Can't extract actions from $tmpc_file\n";
196 # Remove extraneous comments from bison 2.4
197 $actlines =~ s!\s* /\* \s* Line \s* \d+ \s* of \s* yacc\.c \s* \*/!!gx;
199 # C<#line 188 "perlytmp.c"> gets picked up by make depend, so remove them.
200 $actlines =~ s/^#line \d+ "\Q$tmpc_file\E".*$//gm;
202 # convert yyvsp[nnn] into ps[nnn].val
204 $actlines =~ s/yyvsp\[(.*?)\]/ps[$1].val/g
205 or die "Can't convert value stack name\n";
207 return $actlines. "\n", $tablines. "\n";
210 # Generate a table, yy_type_tab[], that specifies for each token, what
211 # type of value it holds.
213 # Read the .y file and extract a list of all the token names and
214 # non-terminal names; then scan the string $tablines for the table yytname,
215 # which gives the token index of each token/non-terminal; then use this to
216 # create yy_type_tab.
218 # ie given (in perly.y),
225 # and (in $tablines),
227 # yytname[] = { "A" "B", "C", "D", "E" };
231 # typedef enum { toketype_ival, toketype_opval, toketype_pval } toketypes;
233 # static const toketypes yy_type_tab[]
234 # = { toketype_opval, toketype_ival, toketype_pval,
235 # toketype_opval, toketype_ival }
237 # where "E" has the default type. The default type is determined
238 # by the __DEFAULT__ comment next to the appropriate union member in
242 my ($y_file, $tablines) = @_;
246 open my $fh, '<', $y_file or die "Can't open $y_file: $!\n";
249 warn "$y_file:$.: dangerous assignment to $1: $_";
253 m{(\w+) \s* ; \s* /\* \s* __DEFAULT__}x
254 or die "$y_file: can't parse __DEFAULT__ line: $_";
255 die "$y_file: duplicate __DEFAULT__ line: $_"
256 if defined $default_token;
261 next unless /^%(token|type)/;
262 s/^%(token|type)\s+<(\w+)>\s+//
263 or die "$y_file: unparseable token/type line: $_";
264 $tokens{$_} = $2 for (split ' ', $_);
267 die "$y_file: no __DEFAULT__ token defined\n" unless $default_token;
268 $types{$default_token} = 1;
270 $tablines =~ /^\Qstatic const char *const yytname[] =\E\n
275 or die "Can't extract yytname[] from table string\n";
277 $fields =~ s{"([^"]+)"}
279 (defined $tokens{$1} ? $tokens{$1} : $default_token)
281 $fields =~ s/, \s* 0 \s* $//x
282 or die "make_type_tab: couldn't delete trailing ',0'\n";
285 "\ntypedef enum {\n\t"
286 . join(", ", map "toketype_$_", sort keys %types)
287 . "\n} toketypes;\n\n"
288 . "/* type of each token/terminal */\n"
289 . "static const toketypes yy_type_tab[] =\n{\n"
298 die "failed to execute command '@_': $!\n";
301 die sprintf "command '@_' died with signal %d\n",
305 die sprintf "command '@_' exited with value %d\n", $? >> 8;