[apple/icu.git] / icuSources / i18n / regexcst.pl

#!/usr/bin/perl
#  ********************************************************************
#  * COPYRIGHT:
#  * Copyright (c) 2002-2007, International Business Machines Corporation and
#  * others. All Rights Reserved.
#  ********************************************************************
#
#  regexcst.pl
#            Compile the regular expression paser state table data into initialized C data.
#            Usage:
#                   cd icu/source/i18n
#                   perl regexcst.pl < regexcst.txt > regexcst.h
#
#             The output file, regexcst.h, is included by some of the .cpp regex
#             implementation files.   This perl script is NOT run as part
#             of a normal ICU build.  It is run by hand when needed, and the
#             regexcst.h generated file is put back into cvs.
#
#             See regexcst.txt for a description of the input format for this script.
#
#             This script is derived from rbbicst.pl, which peforms the same function
#             for the Rule Based Break Iterator Rule Parser.  Perhaps they could be
#             merged?
#


$num_states = 1;         # Always the state number for the line being compiled.
$line_num  = 0;          # The line number in the input file.

$states{"pop"} = 255;    # Add the "pop"  to the list of defined state names.
                         # This prevents any state from being labelled with "pop",
                         #  and resolves references to "pop" in the next state field.

line_loop: while (<>) {
    chomp();
    $line = $_;
    @fields = split();
    $line_num++;

    # Remove # comments, which are any fields beginning with a #, plus all
    #  that follow on the line.
    for ($i=0; $i<@fields; $i++) {
        if ($fields[$i] =~ /^#/) {
            @fields = @fields[0 .. $i-1];
            last;
        }
    }
    # ignore blank lines, and those with no fields left after stripping comments..
    if (@fields == 0) {
        next;
    }

    #
    # State Label:  handling.
    #    Does the first token end with a ":"?  If so, it's the name  of a state.
    #    Put in a hash, together with the current state number,
    #        so that we can later look up the number from the name.
    #
    if (@fields[0] =~ /.*:$/) {
        $state_name = @fields[0];
        $state_name =~ s/://;        # strip off the colon from the state name.

        if ($states{$state_name} != 0) {
            print "  rbbicst: at line $line-num duplicate definition of state $state_name\n";
        }
        $states{$state_name} = $num_states;
        $stateNames[$num_states] = $state_name;

        # if the label was the only thing on this line, go on to the next line,
        # otherwise assume that a state definition is on the same line and fall through.
        if (@fields == 1) {
            next line_loop;
        }
        shift @fields;                       # shift off label field in preparation
                                             #  for handling the rest of the line.
    }

    #
    # State Transition line.
    #   syntax is this,
    #       character   [n]  target-state  [^push-state]  [function-name]
    #   where
    #      [something]   is an optional something
    #      character     is either a single quoted character e.g. '['
    #                       or a name of a character class, e.g. white_space
    #

    $state_line_num[$num_states] = $line_num;   # remember line number with each state
                                                #  so we can make better error messages later.
    #
    # First field, character class or literal character for this transition.
    #
    if ($fields[0] =~ /^'.'$/) {
        # We've got a quoted literal character.
        $state_literal_chars[$num_states] = $fields[0];
        $state_literal_chars[$num_states] =~ s/'//g;
    } else {
        # We've got the name of a character class.
        $state_char_class[$num_states] = $fields[0];
        if ($fields[0] =~ /[\W]/) {
            print "  rbbicsts:  at line $line_num, bad character literal or character class name.\n";
            print "     scanning $fields[0]\n";
            exit(-1);
        }
    }
    shift @fields;

    #
    # do the 'n' flag
    #
    $state_flag[$num_states] = "FALSE";
    if ($fields[0] eq "n") {
        $state_flag[$num_states] = "TRUE";
        shift @fields;
    }

    #
    # do the destination state.
    #
    $state_dest_state[$num_states] = $fields[0];
    if ($fields[0] eq "") {
        print "  rbbicsts:  at line $line_num, destination state missing.\n";
        exit(-1);
    }
    shift @fields;

    #
    # do the push state, if present.
    #
    if ($fields[0] =~ /^\^/) {
        $fields[0] =~ s/^\^//;
        $state_push_state[$num_states] = $fields[0];
        if ($fields[0] eq "" ) {
            print "  rbbicsts:  at line $line_num, expected state after ^ (no spaces).\n";
            exit(-1);
        }
        shift @fields;
    }

    #
    # Lastly, do the optional action name.
    #
    if ($fields[0] ne "") {
        $state_func_name[$num_states] = $fields[0];
        shift @fields;
    }

    #
    #  There should be no fields left on the line at this point.
    #
    if (@fields > 0) {
       print "  rbbicsts:  at line $line_num, unexpected extra stuff on input line.\n";
       print "     scanning $fields[0]\n";
   }
   $num_states++;
}

#
# We've read in the whole file, now go back and output the
#   C source code for the state transition table.
#
# We read all states first, before writing anything,  so that the state numbers
# for the destination states are all available to be written.
#

#
# Make hashes for the names of the character classes and
#      for the names of the actions that appeared.
#
for ($state=1; $state < $num_states; $state++) {
    if ($state_char_class[$state] ne "") {
        if ($charClasses{$state_char_class[$state]} == 0) {
            $charClasses{$state_char_class[$state]} = 1;
        }
    }
    if ($state_func_name[$state] eq "") {
        $state_func_name[$state] = "doNOP";
    }
    if ($actions{$state_action_name[$state]} == 0) {
        $actions{$state_func_name[$state]} = 1;
    }
}

#
# Check that all of the destination states have been defined
#
#
$states{"exit"} = 0;              # Predefined state name, terminates state machine.
for ($state=1; $state<$num_states; $state++) {
   if ($states{$state_dest_state[$state]} == 0 && $state_dest_state[$state] ne "exit") {
       print "Error at line $state_line_num[$state]: target state \"$state_dest_state[$state]\" is not defined.\n";
       $errors++;
   }
   if ($state_push_state[$state] ne "" && $states{$state_push_state[$state]} == 0) {
       print "Error at line $state_line_num[$state]: target state \"$state_push_state[$state]\" is not defined.\n";
       $errors++;
   }
}

die if ($errors>0);

print "//---------------------------------------------------------------------------------\n";
print "//\n";
print "// Generated Header File.  Do not edit by hand.\n";
print "//    This file contains the state table for the ICU Regular Expression Pattern Parser\n";
print "//    It is generated by the Perl script \"regexcst.pl\" from\n";
print "//    the rule parser state definitions file \"regexcst.txt\".\n";
print "//\n";
print "//   Copyright (C) 2002-2007 International Business Machines Corporation \n";
print "//   and others. All rights reserved.  \n";
print "//\n";
print "//---------------------------------------------------------------------------------\n";
print "#ifndef RBBIRPT_H\n";
print "#define RBBIRPT_H\n";
print "\n";
print "U_NAMESPACE_BEGIN\n";

#
# Emit the constants for indicies of Unicode Sets
#   Define one constant for each of the character classes encountered.
#   At the same time, store the index corresponding to the set name back into hash.
#
print "//\n";
print "// Character classes for regex pattern scanning.\n";
print "//\n";
$i = 128;                   # State Table values for Unicode char sets range from 128-250.
                            # Sets "default", "quoted", etc. get special handling.
                            #  They have no corresponding UnicodeSet object in the state machine,
                            #    but are handled by special case code.  So we emit no reference
                            #    to a UnicodeSet object to them here.
foreach $setName (keys %charClasses) {
    if ($setName eq "default") {
        $charClasses{$setName} = 255;}
    elsif ($setName eq "quoted") {
        $charClasses{$setName} = 254;}
    elsif ($setName eq "eof") {
        $charClasses{$setName} = 253;}
    else {
        # Normal character class.  Fill in array with a ptr to the corresponding UnicodeSet in the state machine.
       print "    static const uint8_t kRuleSet_$setName = $i;\n";
        $charClasses{$setName} = $i;
        $i++;
    }
}
print "\n\n";

#
# Emit the enum for the actions to be performed.
#
print "enum Regex_PatternParseAction {\n";
foreach $act (keys %actions) {
    print "    $act,\n";
}
print "    rbbiLastAction};\n\n";

#
# Emit the struct definition for transtion table elements.
#
print "//-------------------------------------------------------------------------------\n";
print "//\n";
print "//  RegexTableEl       represents the structure of a row in the transition table\n";
print "//                     for the pattern parser state machine.\n";
print "//-------------------------------------------------------------------------------\n";
print "struct RegexTableEl {\n";
print "    Regex_PatternParseAction      fAction;\n";
print "    uint8_t                       fCharClass;       // 0-127:    an individual ASCII character\n";
print "                                                    // 128-255:  character class index\n";
print "    uint8_t                       fNextState;       // 0-250:    normal next-state numbers\n";
print "                                                    // 255:      pop next-state from stack.\n";
print "    uint8_t                       fPushState;\n";
print "    UBool                         fNextChar;\n";
print "};\n\n";

#
# emit the state transition table
#
print "static const struct RegexTableEl gRuleParseStateTable[] = {\n";
print "    {doNOP, 0, 0, 0, TRUE}\n";    # State 0 is a dummy.  Real states start with index = 1.
for ($state=1; $state < $num_states; $state++) {
    print "    , {$state_func_name[$state],";
    if ($state_literal_chars[$state] ne "") {
        $c = $state_literal_chars[$state];
        printf(" %d /* $c */,", ord($c));   #  use numeric value, so EBCDIC machines are ok.
    }else {
        print " $charClasses{$state_char_class[$state]},";
    }
    print " $states{$state_dest_state[$state]},";

    # The push-state field is optional.  If omitted, fill field with a zero, which flags
    #   the state machine that there is no push state.
    if ($state_push_state[$state] eq "") {
        print "0, ";
    } else {
        print " $states{$state_push_state[$state]},";
    }
    print " $state_flag[$state]} ";

    # Put out a C++ comment showing the number (index) of this state row,
    #   and, if this is the first row of the table for this state, the state name.
    print "    //  $state ";
    if ($stateNames[$state] ne "") {
        print "     $stateNames[$state]";
    }
    print "\n";
};
print " };\n";


#
# emit a mapping array from state numbers to state names.
#
#    This array is used for producing debugging output from the pattern parser.
#
print "static const char * const RegexStateNames[] = {";
for ($state=0; $state<$num_states; $state++) {
    if ($stateNames[$state] ne "") {
        print "     \"$stateNames[$state]\",\n";
    } else {
        print "    0,\n";
    }
}
print "    0};\n\n";

print "U_NAMESPACE_END\n";
print "#endif\n";
Commit	Line	Data
b75a7d8f A	1	#!/usr/bin/perl
	2	# ********************************************************************
	3	# * COPYRIGHT:
46f4442e	4	# * Copyright (c) 2002-2007, International Business Machines Corporation and
b75a7d8f A	5	# * others. All Rights Reserved.
	6	# ********************************************************************
	7	#
	8	# regexcst.pl
	9	# Compile the regular expression paser state table data into initialized C data.
	10	# Usage:
	11	# cd icu/source/i18n
	12	# perl regexcst.pl < regexcst.txt > regexcst.h
	13	#
	14	# The output file, regexcst.h, is included by some of the .cpp regex
	15	# implementation files. This perl script is NOT run as part
	16	# of a normal ICU build. It is run by hand when needed, and the
	17	# regexcst.h generated file is put back into cvs.
	18	#
	19	# See regexcst.txt for a description of the input format for this script.
	20	#
	21	# This script is derived from rbbicst.pl, which peforms the same function
	22	# for the Rule Based Break Iterator Rule Parser. Perhaps they could be
	23	# merged?
	24	#
b75a7d8f A	25
	26
	27	$num_states = 1; # Always the state number for the line being compiled.
	28	$line_num = 0; # The line number in the input file.
	29
	30	$states{"pop"} = 255; # Add the "pop" to the list of defined state names.
	31	# This prevents any state from being labelled with "pop",
	32	# and resolves references to "pop" in the next state field.
	33
	34	line_loop: while (<>) {
	35	chomp();
	36	$line = $_;
	37	@fields = split();
	38	$line_num++;
	39
	40	# Remove # comments, which are any fields beginning with a #, plus all
	41	# that follow on the line.
	42	for ($i=0; $i<@fields; $i++) {
	43	if ($fields[$i] =~ /^#/) {
	44	@fields = @fields[0 .. $i-1];
	45	last;
	46	}
	47	}
	48	# ignore blank lines, and those with no fields left after stripping comments..
	49	if (@fields == 0) {
	50	next;
	51	}
	52
	53	#
	54	# State Label: handling.
	55	# Does the first token end with a ":"? If so, it's the name of a state.
	56	# Put in a hash, together with the current state number,
	57	# so that we can later look up the number from the name.
	58	#
	59	if (@fields[0] =~ /.*:$/) {
	60	$state_name = @fields[0];
	61	$state_name =~ s/://; # strip off the colon from the state name.
	62
	63	if ($states{$state_name} != 0) {
	64	print " rbbicst: at line $line-num duplicate definition of state $state_name\n";
	65	}
	66	$states{$state_name} = $num_states;
	67	$stateNames[$num_states] = $state_name;
	68
	69	# if the label was the only thing on this line, go on to the next line,
	70	# otherwise assume that a state definition is on the same line and fall through.
	71	if (@fields == 1) {
	72	next line_loop;
	73	}
	74	shift @fields; # shift off label field in preparation
	75	# for handling the rest of the line.
	76	}
	77
	78	#
	79	# State Transition line.
	80	# syntax is this,
	81	# character [n] target-state [^push-state] [function-name]
	82	# where
	83	# [something] is an optional something
	84	# character is either a single quoted character e.g. '['
	85	# or a name of a character class, e.g. white_space
	86	#
	87
	88	$state_line_num[$num_states] = $line_num; # remember line number with each state
89	# so we can make better error messages later.
90	#
91	# First field, character class or literal character for this transition.
92	#
93	if ($fields[0] =~ /^'.'$/) {
94	# We've got a quoted literal character.
95	$state_literal_chars[$num_states] = $fields[0];
96	$state_literal_chars[$num_states] =~ s/'//g;
97	} else {
98	# We've got the name of a character class.
99	$state_char_class[$num_states] = $fields[0];
100	if ($fields[0] =~ /[\W]/) {
101	print " rbbicsts: at line $line_num, bad character literal or character class name.\n";
102	print " scanning $fields[0]\n";
103	exit(-1);
104	}
105	}
106	shift @fields;
107
108	#
109	# do the 'n' flag
110	#
111	$state_flag[$num_states] = "FALSE";
112	if ($fields[0] eq "n") {
113	$state_flag[$num_states] = "TRUE";
114	shift @fields;
115	}
116
117	#
118	# do the destination state.
119	#
120	$state_dest_state[$num_states] = $fields[0];
121	if ($fields[0] eq "") {
122	print " rbbicsts: at line $line_num, destination state missing.\n";
123	exit(-1);
124	}
125	shift @fields;
126
127	#
128	# do the push state, if present.
129	#
130	if ($fields[0] =~ /^\^/) {
131	$fields[0] =~ s/^\^//;
132	$state_push_state[$num_states] = $fields[0];
133	if ($fields[0] eq "" ) {
134	print " rbbicsts: at line $line_num, expected state after ^ (no spaces).\n";
135	exit(-1);
136	}
137	shift @fields;
138	}
139
140	#
141	# Lastly, do the optional action name.
142	#
143	if ($fields[0] ne "") {
144	$state_func_name[$num_states] = $fields[0];
145	shift @fields;
146	}
147
148	#
149	# There should be no fields left on the line at this point.
150	#
151	if (@fields > 0) {
152	print " rbbicsts: at line $line_num, unexpected extra stuff on input line.\n";
153	print " scanning $fields[0]\n";
154	}
155	$num_states++;
156	}
157
158	#
159	# We've read in the whole file, now go back and output the
160	# C source code for the state transition table.
161	#
162	# We read all states first, before writing anything, so that the state numbers
163	# for the destination states are all available to be written.
164	#
165
166	#
167	# Make hashes for the names of the character classes and
168	# for the names of the actions that appeared.
169	#
170	for ($state=1; $state < $num_states; $state++) {
171	if ($state_char_class[$state] ne "") {
172	if ($charClasses{$state_char_class[$state]} == 0) {
173	$charClasses{$state_char_class[$state]} = 1;
174	}
175	}
176	if ($state_func_name[$state] eq "") {
177	$state_func_name[$state] = "doNOP";
178	}
179	if ($actions{$state_action_name[$state]} == 0) {
180	$actions{$state_func_name[$state]} = 1;
181	}
182	}
183
184	#
185	# Check that all of the destination states have been defined
186	#
187	#
188	$states{"exit"} = 0; # Predefined state name, terminates state machine.
189	for ($state=1; $state<$num_states; $state++) {
190	if ($states{$state_dest_state[$state]} == 0 && $state_dest_state[$state] ne "exit") {
191	print "Error at line $state_line_num[$state]: target state \"$state_dest_state[$state]\" is not defined.\n";
192	$errors++;
193	}
194	if ($state_push_state[$state] ne "" && $states{$state_push_state[$state]} == 0) {
195	print "Error at line $state_line_num[$state]: target state \"$state_push_state[$state]\" is not defined.\n";
196	$errors++;
197	}
198	}
199
200	die if ($errors>0);
201
202	print "//---------------------------------------------------------------------------------\n";
203	print "//\n";
204	print "// Generated Header File. Do not edit by hand.\n";
205	print "// This file contains the state table for the ICU Regular Expression Pattern Parser\n";
206	print "// It is generated by the Perl script \"regexcst.pl\" from\n";
207	print "// the rule parser state definitions file \"regexcst.txt\".\n";
208	print "//\n";
46f4442e	209	print "// Copyright (C) 2002-2007 International Business Machines Corporation \n";
b75a7d8f A	210	print "// and others. All rights reserved. \n";
	211	print "//\n";
	212	print "//---------------------------------------------------------------------------------\n";
	213	print "#ifndef RBBIRPT_H\n";
	214	print "#define RBBIRPT_H\n";
	215	print "\n";
	216	print "U_NAMESPACE_BEGIN\n";
	217
	218	#
	219	# Emit the constants for indicies of Unicode Sets
	220	# Define one constant for each of the character classes encountered.
	221	# At the same time, store the index corresponding to the set name back into hash.
	222	#
	223	print "//\n";
	224	print "// Character classes for regex pattern scanning.\n";
	225	print "//\n";
	226	$i = 128; # State Table values for Unicode char sets range from 128-250.
	227	# Sets "default", "quoted", etc. get special handling.
	228	# They have no corresponding UnicodeSet object in the state machine,
	229	# but are handled by special case code. So we emit no reference
	230	# to a UnicodeSet object to them here.
	231	foreach $setName (keys %charClasses) {
	232	if ($setName eq "default") {
	233	$charClasses{$setName} = 255;}
	234	elsif ($setName eq "quoted") {
	235	$charClasses{$setName} = 254;}
	236	elsif ($setName eq "eof") {
	237	$charClasses{$setName} = 253;}
	238	else {
	239	# Normal character class. Fill in array with a ptr to the corresponding UnicodeSet in the state machine.
	240	print " static const uint8_t kRuleSet_$setName = $i;\n";
	241	$charClasses{$setName} = $i;
	242	$i++;
	243	}
	244	}
	245	print "\n\n";
	246
	247	#
	248	# Emit the enum for the actions to be performed.
	249	#
	250	print "enum Regex_PatternParseAction {\n";
	251	foreach $act (keys %actions) {
	252	print " $act,\n";
	253	}
	254	print " rbbiLastAction};\n\n";
	255
	256	#
	257	# Emit the struct definition for transtion table elements.
	258	#
	259	print "//-------------------------------------------------------------------------------\n";
	260	print "//\n";
	261	print "// RegexTableEl represents the structure of a row in the transition table\n";
	262	print "// for the pattern parser state machine.\n";
	263	print "//-------------------------------------------------------------------------------\n";
	264	print "struct RegexTableEl {\n";
	265	print " Regex_PatternParseAction fAction;\n";
	266	print " uint8_t fCharClass; // 0-127: an individual ASCII character\n";
	267	print " // 128-255: character class index\n";
	268	print " uint8_t fNextState; // 0-250: normal next-state numbers\n";
	269	print " // 255: pop next-state from stack.\n";
	270	print " uint8_t fPushState;\n";
	271	print " UBool fNextChar;\n";
	272	print "};\n\n";
	273
274	#
275	# emit the state transition table
276	#
277	print "static const struct RegexTableEl gRuleParseStateTable[] = {\n";
278	print " {doNOP, 0, 0, 0, TRUE}\n"; # State 0 is a dummy. Real states start with index = 1.
279	for ($state=1; $state < $num_states; $state++) {
280	print " , {$state_func_name[$state],";
281	if ($state_literal_chars[$state] ne "") {
282	$c = $state_literal_chars[$state];
283	printf(" %d /* $c */,", ord($c)); # use numeric value, so EBCDIC machines are ok.
284	}else {
285	print " $charClasses{$state_char_class[$state]},";
286	}
287	print " $states{$state_dest_state[$state]},";
288
289	# The push-state field is optional. If omitted, fill field with a zero, which flags
290	# the state machine that there is no push state.
291	if ($state_push_state[$state] eq "") {
292	print "0, ";
293	} else {
294	print " $states{$state_push_state[$state]},";
295	}
296	print " $state_flag[$state]} ";
297
298	# Put out a C++ comment showing the number (index) of this state row,
299	# and, if this is the first row of the table for this state, the state name.
300	print " // $state ";
301	if ($stateNames[$state] ne "") {
302	print " $stateNames[$state]";
303	}
304	print "\n";
305	};
306	print " };\n";
307
308
309	#
310	# emit a mapping array from state numbers to state names.
311	#
312	# This array is used for producing debugging output from the pattern parser.
313	#
314	print "static const char * const RegexStateNames[] = {";
315	for ($state=0; $state<$num_states; $state++) {
316	if ($stateNames[$state] ne "") {
317	print " \"$stateNames[$state]\",\n";
318	} else {
319	print " 0,\n";
320	}
321	}
322	print " 0};\n\n";
323
324	print "U_NAMESPACE_END\n";
325	print "#endif\n";
326
327
328