3 # !!!!!!!!!!!!!! IF YOU MODIFY THIS FILE !!!!!!!!!!!!!!!!!!!!!!!!!
4 # Any files created or read by this program should be listed in 'mktables.lst'
5 # Use -makelist to regenerate it.
7 # Needs 'no overloading' to run faster on miniperl. Code commented out at the
8 # subroutine objaddr can be used instead to work as far back (untested) as
9 # 5.8: needs pack "U". But almost all occurrences of objaddr have been
10 # removed in favor of using 'no overloading'. You also would have to go
11 # through and replace occurrences like:
12 # my $addr = do { no overloading; pack 'J', $self; }
14 # my $addr = main::objaddr $self;
15 # (or reverse commit 9b01bafde4b022706c3d6f947a0963f821b2e50b
16 # that instituted the change to main::objaddr, and subsequent commits that
17 # changed 0+$self to pack 'J', $self.)
20 BEGIN { # Get the time the script started running; do it at compiliation to
21 # get it as close as possible
36 sub DEBUG () { 0 } # Set to 0 for production; 1 for development
37 my $debugging_build = $Config{"ccflags"} =~ /-DDEBUGGING/;
39 ##########################################################################
41 # mktables -- create the runtime Perl Unicode files (lib/unicore/.../*.pl),
42 # from the Unicode database files (lib/unicore/.../*.txt), It also generates
43 # a pod file and a .t file
45 # The structure of this file is:
46 # First these introductory comments; then
47 # code needed for everywhere, such as debugging stuff; then
48 # code to handle input parameters; then
49 # data structures likely to be of external interest (some of which depend on
50 # the input parameters, so follows them; then
51 # more data structures and subroutine and package (class) definitions; then
52 # the small actual loop to process the input files and finish up; then
53 # a __DATA__ section, for the .t tests
55 # This program works on all releases of Unicode through at least 6.0. The
56 # outputs have been scrutinized most intently for release 5.1. The others
57 # have been checked for somewhat more than just sanity. It can handle all
58 # existing Unicode character properties in those releases.
60 # This program is mostly about Unicode character (or code point) properties.
61 # A property describes some attribute or quality of a code point, like if it
62 # is lowercase or not, its name, what version of Unicode it was first defined
63 # in, or what its uppercase equivalent is. Unicode deals with these disparate
64 # possibilities by making all properties into mappings from each code point
65 # into some corresponding value. In the case of it being lowercase or not,
66 # the mapping is either to 'Y' or 'N' (or various synonyms thereof). Each
67 # property maps each Unicode code point to a single value, called a "property
68 # value". (Hence each Unicode property is a true mathematical function with
69 # exactly one value per code point.)
71 # When using a property in a regular expression, what is desired isn't the
72 # mapping of the code point to its property's value, but the reverse (or the
73 # mathematical "inverse relation"): starting with the property value, "Does a
74 # code point map to it?" These are written in a "compound" form:
75 # \p{property=value}, e.g., \p{category=punctuation}. This program generates
76 # files containing the lists of code points that map to each such regular
77 # expression property value, one file per list
79 # There is also a single form shortcut that Perl adds for many of the commonly
80 # used properties. This happens for all binary properties, plus script,
81 # general_category, and block properties.
83 # Thus the outputs of this program are files. There are map files, mostly in
84 # the 'To' directory; and there are list files for use in regular expression
85 # matching, all in subdirectories of the 'lib' directory, with each
86 # subdirectory being named for the property that the lists in it are for.
87 # Bookkeeping, test, and documentation files are also generated.
89 my $matches_directory = 'lib'; # Where match (\p{}) files go.
90 my $map_directory = 'To'; # Where map files go.
94 # The major data structures of this program are Property, of course, but also
95 # Table. There are two kinds of tables, very similar to each other.
96 # "Match_Table" is the data structure giving the list of code points that have
97 # a particular property value, mentioned above. There is also a "Map_Table"
98 # data structure which gives the property's mapping from code point to value.
99 # There are two structures because the match tables need to be combined in
100 # various ways, such as constructing unions, intersections, complements, etc.,
101 # and the map ones don't. And there would be problems, perhaps subtle, if
102 # a map table were inadvertently operated on in some of those ways.
103 # The use of separate classes with operations defined on one but not the other
104 # prevents accidentally confusing the two.
106 # At the heart of each table's data structure is a "Range_List", which is just
107 # an ordered list of "Ranges", plus ancillary information, and methods to
108 # operate on them. A Range is a compact way to store property information.
109 # Each range has a starting code point, an ending code point, and a value that
110 # is meant to apply to all the code points between the two end points,
111 # inclusive. For a map table, this value is the property value for those
112 # code points. Two such ranges could be written like this:
113 # 0x41 .. 0x5A, 'Upper',
114 # 0x61 .. 0x7A, 'Lower'
116 # Each range also has a type used as a convenience to classify the values.
117 # Most ranges in this program will be Type 0, or normal, but there are some
118 # ranges that have a non-zero type. These are used only in map tables, and
119 # are for mappings that don't fit into the normal scheme of things. Mappings
120 # that require a hash entry to communicate with utf8.c are one example;
121 # another example is mappings for charnames.pm to use which indicate a name
122 # that is algorithmically determinable from its code point (and vice-versa).
123 # These are used to significantly compact these tables, instead of listing
124 # each one of the tens of thousands individually.
126 # In a match table, the value of a range is irrelevant (and hence the type as
127 # well, which will always be 0), and arbitrarily set to the null string.
128 # Using the example above, there would be two match tables for those two
129 # entries, one named Upper would contain the 0x41..0x5A range, and the other
130 # named Lower would contain 0x61..0x7A.
132 # Actually, there are two types of range lists, "Range_Map" is the one
133 # associated with map tables, and "Range_List" with match tables.
134 # Again, this is so that methods can be defined on one and not the other so as
135 # to prevent operating on them in incorrect ways.
137 # Eventually, most tables are written out to files to be read by utf8_heavy.pl
138 # in the perl core. All tables could in theory be written, but some are
139 # suppressed because there is no current practical use for them. It is easy
140 # to change which get written by changing various lists that are near the top
141 # of the actual code in this file. The table data structures contain enough
142 # ancillary information to allow them to be treated as separate entities for
143 # writing, such as the path to each one's file. There is a heading in each
144 # map table that gives the format of its entries, and what the map is for all
145 # the code points missing from it. (This allows tables to be more compact.)
147 # The Property data structure contains one or more tables. All properties
148 # contain a map table (except the $perl property which is a
149 # pseudo-property containing only match tables), and any properties that
150 # are usable in regular expression matches also contain various matching
151 # tables, one for each value the property can have. A binary property can
152 # have two values, True and False (or Y and N, which are preferred by Unicode
153 # terminology). Thus each of these properties will have a map table that
154 # takes every code point and maps it to Y or N (but having ranges cuts the
155 # number of entries in that table way down), and two match tables, one
156 # which has a list of all the code points that map to Y, and one for all the
157 # code points that map to N. (For each of these, a third table is also
158 # generated for the pseudo Perl property. It contains the identical code
159 # points as the Y table, but can be written, not in the compound form, but in
160 # a "single" form like \p{IsUppercase}.) Many properties are binary, but some
161 # properties have several possible values, some have many, and properties like
162 # Name have a different value for every named code point. Those will not,
163 # unless the controlling lists are changed, have their match tables written
164 # out. But all the ones which can be used in regular expression \p{} and \P{}
165 # constructs will. Generally a property will have either its map table or its
166 # match tables written but not both. Again, what gets written is controlled
167 # by lists which can easily be changed. Properties have a 'Type', like
168 # binary, or string, or enum depending on how many match tables there are and
169 # the content of the maps. This 'Type' is different than a range 'Type', so
170 # don't get confused by the two concepts having the same name.
172 # For information about the Unicode properties, see Unicode's UAX44 document:
174 my $unicode_reference_url = 'http://www.unicode.org/reports/tr44/';
176 # As stated earlier, this program will work on any release of Unicode so far.
177 # Most obvious problems in earlier data have NOT been corrected except when
178 # necessary to make Perl or this program work reasonably. For example, no
179 # folding information was given in early releases, so this program uses the
180 # substitute of lower case, just so that a regular expression with the /i
181 # option will do something that actually gives the right results in many
182 # cases. There are also a couple other corrections for version 1.1.5,
183 # commented at the point they are made. As an example of corrections that
184 # weren't made (but could be) is this statement from DerivedAge.txt: "The
185 # supplementary private use code points and the non-character code points were
186 # assigned in version 2.0, but not specifically listed in the UCD until
187 # versions 3.0 and 3.1 respectively." (To be precise it was 3.0.1 not 3.0.0)
188 # More information on Unicode version glitches is further down in these
189 # introductory comments.
191 # This program works on all non-provisional properties as of 6.0, though the
192 # files for some are suppressed from apparent lack of demand for them. You
193 # can change which are output by changing lists in this program.
195 # The old version of mktables emphasized the term "Fuzzy" to mean Unicode's
196 # loose matchings rules (from Unicode TR18):
198 # The recommended names for UCD properties and property values are in
199 # PropertyAliases.txt [Prop] and PropertyValueAliases.txt
200 # [PropValue]. There are both abbreviated names and longer, more
201 # descriptive names. It is strongly recommended that both names be
202 # recognized, and that loose matching of property names be used,
203 # whereby the case distinctions, whitespace, hyphens, and underbar
205 # The program still allows Fuzzy to override its determination of if loose
206 # matching should be used, but it isn't currently used, as it is no longer
207 # needed; the calculations it makes are good enough.
209 # SUMMARY OF HOW IT WORKS:
213 # A list is constructed containing each input file that is to be processed
215 # Each file on the list is processed in a loop, using the associated handler
217 # The PropertyAliases.txt and PropValueAliases.txt files are processed
218 # first. These files name the properties and property values.
219 # Objects are created of all the property and property value names
220 # that the rest of the input should expect, including all synonyms.
221 # The other input files give mappings from properties to property
222 # values. That is, they list code points and say what the mapping
223 # is under the given property. Some files give the mappings for
224 # just one property; and some for many. This program goes through
225 # each file and populates the properties from them. Some properties
226 # are listed in more than one file, and Unicode has set up a
227 # precedence as to which has priority if there is a conflict. Thus
228 # the order of processing matters, and this program handles the
229 # conflict possibility by processing the overriding input files
230 # last, so that if necessary they replace earlier values.
231 # After this is all done, the program creates the property mappings not
232 # furnished by Unicode, but derivable from what it does give.
233 # The tables of code points that match each property value in each
234 # property that is accessible by regular expressions are created.
235 # The Perl-defined properties are created and populated. Many of these
236 # require data determined from the earlier steps
237 # Any Perl-defined synonyms are created, and name clashes between Perl
238 # and Unicode are reconciled and warned about.
239 # All the properties are written to files
240 # Any other files are written, and final warnings issued.
242 # For clarity, a number of operators have been overloaded to work on tables:
243 # ~ means invert (take all characters not in the set). The more
244 # conventional '!' is not used because of the possibility of confusing
245 # it with the actual boolean operation.
247 # - means subtraction
248 # & means intersection
249 # The precedence of these is the order listed. Parentheses should be
250 # copiously used. These are not a general scheme. The operations aren't
251 # defined for a number of things, deliberately, to avoid getting into trouble.
252 # Operations are done on references and affect the underlying structures, so
253 # that the copy constructors for them have been overloaded to not return a new
254 # clone, but the input object itself.
256 # The bool operator is deliberately not overloaded to avoid confusion with
257 # "should it mean if the object merely exists, or also is non-empty?".
259 # WHY CERTAIN DESIGN DECISIONS WERE MADE
261 # This program needs to be able to run under miniperl. Therefore, it uses a
262 # minimum of other modules, and hence implements some things itself that could
263 # be gotten from CPAN
265 # This program uses inputs published by the Unicode Consortium. These can
266 # change incompatibly between releases without the Perl maintainers realizing
267 # it. Therefore this program is now designed to try to flag these. It looks
268 # at the directories where the inputs are, and flags any unrecognized files.
269 # It keeps track of all the properties in the files it handles, and flags any
270 # that it doesn't know how to handle. It also flags any input lines that
271 # don't match the expected syntax, among other checks.
273 # It is also designed so if a new input file matches one of the known
274 # templates, one hopefully just needs to add it to a list to have it
277 # As mentioned earlier, some properties are given in more than one file. In
278 # particular, the files in the extracted directory are supposedly just
279 # reformattings of the others. But they contain information not easily
280 # derivable from the other files, including results for Unihan, which this
281 # program doesn't ordinarily look at, and for unassigned code points. They
282 # also have historically had errors or been incomplete. In an attempt to
283 # create the best possible data, this program thus processes them first to
284 # glean information missing from the other files; then processes those other
285 # files to override any errors in the extracted ones. Much of the design was
286 # driven by this need to store things and then possibly override them.
288 # It tries to keep fatal errors to a minimum, to generate something usable for
289 # testing purposes. It always looks for files that could be inputs, and will
290 # warn about any that it doesn't know how to handle (the -q option suppresses
293 # Why have files written out for binary 'N' matches?
294 # For binary properties, if you know the mapping for either Y or N; the
295 # other is trivial to construct, so could be done at Perl run-time by just
296 # complementing the result, instead of having a file for it. That is, if
297 # someone types in \p{foo: N}, Perl could translate that to \P{foo: Y} and
298 # not need a file. The problem is communicating to Perl that a given
299 # property is binary. Perl can't figure it out from looking at the N (or
300 # No), as some non-binary properties have these as property values. So
301 # rather than inventing a way to communicate this info back to the core,
302 # which would have required changes there as well, it was simpler just to
303 # add the extra tables.
305 # Why is there more than one type of range?
306 # This simplified things. There are some very specialized code points that
307 # have to be handled specially for output, such as Hangul syllable names.
308 # By creating a range type (done late in the development process), it
309 # allowed this to be stored with the range, and overridden by other input.
310 # Originally these were stored in another data structure, and it became a
311 # mess trying to decide if a second file that was for the same property was
312 # overriding the earlier one or not.
314 # Why are there two kinds of tables, match and map?
315 # (And there is a base class shared by the two as well.) As stated above,
316 # they actually are for different things. Development proceeded much more
317 # smoothly when I (khw) realized the distinction. Map tables are used to
318 # give the property value for every code point (actually every code point
319 # that doesn't map to a default value). Match tables are used for regular
320 # expression matches, and are essentially the inverse mapping. Separating
321 # the two allows more specialized methods, and error checks so that one
322 # can't just take the intersection of two map tables, for example, as that
325 # There are no match tables generated for matches of the null string. These
326 # would look like qr/\p{JSN=}/ currently without modifying the regex code.
327 # Perhaps something like them could be added if necessary. The JSN does have
328 # a real code point U+110B that maps to the null string, but it is a
329 # contributory property, and therefore not output by default. And it's easily
330 # handled so far by making the null string the default where it is a
335 # This program is written so it will run under miniperl. Occasionally changes
336 # will cause an error where the backtrace doesn't work well under miniperl.
337 # To diagnose the problem, you can instead run it under regular perl, if you
340 # There is a good trace facility. To enable it, first sub DEBUG must be set
341 # to return true. Then a line like
343 # local $to_trace = 1 if main::DEBUG;
345 # can be added to enable tracing in its lexical scope or until you insert
348 # local $to_trace = 0 if main::DEBUG;
350 # then use a line like "trace $a, @b, %c, ...;
352 # Some of the more complex subroutines already have trace statements in them.
353 # Permanent trace statements should be like:
355 # trace ... if main::DEBUG && $to_trace;
357 # If there is just one or a few files that you're debugging, you can easily
358 # cause most everything else to be skipped. Change the line
360 # my $debug_skip = 0;
362 # to 1, and every file whose object is in @input_file_objects and doesn't have
363 # a, 'non_skip => 1,' in its constructor will be skipped.
365 # To compare the output tables, it may be useful to specify the -annotate
366 # flag. This causes the tables to expand so there is one entry for each
367 # non-algorithmically named code point giving, currently its name, and its
368 # graphic representation if printable (and you have a font that knows about
369 # it). This makes it easier to see what the particular code points are in
370 # each output table. The tables are usable, but because they don't have
371 # ranges (for the most part), a Perl using them will run slower. Non-named
372 # code points are annotated with a description of their status, and contiguous
373 # ones with the same description will be output as a range rather than
374 # individually. Algorithmically named characters are also output as ranges,
375 # except when there are just a few contiguous ones.
379 # The program would break if Unicode were to change its names so that
380 # interior white space, underscores, or dashes differences were significant
381 # within property and property value names.
383 # It might be easier to use the xml versions of the UCD if this program ever
384 # would need heavy revision, and the ability to handle old versions was not
387 # There is the potential for name collisions, in that Perl has chosen names
388 # that Unicode could decide it also likes. There have been such collisions in
389 # the past, with mostly Perl deciding to adopt the Unicode definition of the
390 # name. However in the 5.2 Unicode beta testing, there were a number of such
391 # collisions, which were withdrawn before the final release, because of Perl's
392 # and other's protests. These all involved new properties which began with
393 # 'Is'. Based on the protests, Unicode is unlikely to try that again. Also,
394 # many of the Perl-defined synonyms, like Any, Word, etc, are listed in a
395 # Unicode document, so they are unlikely to be used by Unicode for another
396 # purpose. However, they might try something beginning with 'In', or use any
397 # of the other Perl-defined properties. This program will warn you of name
398 # collisions, and refuse to generate tables with them, but manual intervention
399 # will be required in this event. One scheme that could be implemented, if
400 # necessary, would be to have this program generate another file, or add a
401 # field to mktables.lst that gives the date of first definition of a property.
402 # Each new release of Unicode would use that file as a basis for the next
403 # iteration. And the Perl synonym addition code could sort based on the age
404 # of the property, so older properties get priority, and newer ones that clash
405 # would be refused; hence existing code would not be impacted, and some other
406 # synonym would have to be used for the new property. This is ugly, and
407 # manual intervention would certainly be easier to do in the short run; lets
408 # hope it never comes to this.
412 # This program can generate tables from the Unihan database. But it doesn't
413 # by default, letting the CPAN module Unicode::Unihan handle them. Prior to
414 # version 5.2, this database was in a single file, Unihan.txt. In 5.2 the
415 # database was split into 8 different files, all beginning with the letters
416 # 'Unihan'. This program will read those file(s) if present, but it needs to
417 # know which of the many properties in the file(s) should have tables created
418 # for them. It will create tables for any properties listed in
419 # PropertyAliases.txt and PropValueAliases.txt, plus any listed in the
420 # @cjk_properties array and the @cjk_property_values array. Thus, if a
421 # property you want is not in those files of the release you are building
422 # against, you must add it to those two arrays. Starting in 4.0, the
423 # Unicode_Radical_Stroke was listed in those files, so if the Unihan database
424 # is present in the directory, a table will be generated for that property.
425 # In 5.2, several more properties were added. For your convenience, the two
426 # arrays are initialized with all the 6.0 listed properties that are also in
427 # earlier releases. But these are commented out. You can just uncomment the
428 # ones you want, or use them as a template for adding entries for other
431 # You may need to adjust the entries to suit your purposes. setup_unihan(),
432 # and filter_unihan_line() are the functions where this is done. This program
433 # already does some adjusting to make the lines look more like the rest of the
434 # Unicode DB; You can see what that is in filter_unihan_line()
436 # There is a bug in the 3.2 data file in which some values for the
437 # kPrimaryNumeric property have commas and an unexpected comment. A filter
438 # could be added for these; or for a particular installation, the Unihan.txt
439 # file could be edited to fix them.
441 # HOW TO ADD A FILE TO BE PROCESSED
443 # A new file from Unicode needs to have an object constructed for it in
444 # @input_file_objects, probably at the end or at the end of the extracted
445 # ones. The program should warn you if its name will clash with others on
446 # restrictive file systems, like DOS. If so, figure out a better name, and
447 # add lines to the README.perl file giving that. If the file is a character
448 # property, it should be in the format that Unicode has by default
449 # standardized for such files for the more recently introduced ones.
450 # If so, the Input_file constructor for @input_file_objects can just be the
451 # file name and release it first appeared in. If not, then it should be
452 # possible to construct an each_line_handler() to massage the line into the
455 # For non-character properties, more code will be needed. You can look at
456 # the existing entries for clues.
458 # UNICODE VERSIONS NOTES
460 # The Unicode UCD has had a number of errors in it over the versions. And
461 # these remain, by policy, in the standard for that version. Therefore it is
462 # risky to correct them, because code may be expecting the error. So this
463 # program doesn't generally make changes, unless the error breaks the Perl
464 # core. As an example, some versions of 2.1.x Jamo.txt have the wrong value
465 # for U+1105, which causes real problems for the algorithms for Jamo
466 # calculations, so it is changed here.
468 # But it isn't so clear cut as to what to do about concepts that are
469 # introduced in a later release; should they extend back to earlier releases
470 # where the concept just didn't exist? It was easier to do this than to not,
471 # so that's what was done. For example, the default value for code points not
472 # in the files for various properties was probably undefined until changed by
473 # some version. No_Block for blocks is such an example. This program will
474 # assign No_Block even in Unicode versions that didn't have it. This has the
475 # benefit that code being written doesn't have to special case earlier
476 # versions; and the detriment that it doesn't match the Standard precisely for
477 # the affected versions.
479 # Here are some observations about some of the issues in early versions:
481 # The number of code points in \p{alpha} halved in 2.1.9. It turns out that
482 # the reason is that the CJK block starting at 4E00 was removed from PropList,
483 # and was not put back in until 3.1.0
485 # Unicode introduced the synonym Space for White_Space in 4.1. Perl has
486 # always had a \p{Space}. In release 3.2 only, they are not synonymous. The
487 # reason is that 3.2 introduced U+205F=medium math space, which was not
488 # classed as white space, but Perl figured out that it should have been. 4.0
489 # reclassified it correctly.
491 # Another change between 3.2 and 4.0 is the CCC property value ATBL. In 3.2
492 # this was erroneously a synonym for 202. In 4.0, ATB became 202, and ATBL
493 # was left with no code points, as all the ones that mapped to 202 stayed
494 # mapped to 202. Thus if your program used the numeric name for the class,
495 # it would not have been affected, but if it used the mnemonic, it would have
498 # \p{Script=Hrkt} (Katakana_Or_Hiragana) came in 4.0.1. Before that code
499 # points which eventually came to have this script property value, instead
500 # mapped to "Unknown". But in the next release all these code points were
501 # moved to \p{sc=common} instead.
503 # The default for missing code points for BidiClass is complicated. Starting
504 # in 3.1.1, the derived file DBidiClass.txt handles this, but this program
505 # tries to do the best it can for earlier releases. It is done in
506 # process_PropertyAliases()
508 ##############################################################################
510 my $UNDEF = ':UNDEF:'; # String to print out for undefined values in tracing
512 my $MAX_LINE_WIDTH = 78;
514 # Debugging aid to skip most files so as to not be distracted by them when
515 # concentrating on the ones being debugged. Add
517 # to the constructor for those files you want processed when you set this.
518 # Files with a first version number of 0 are special: they are always
519 # processed regardless of the state of this flag.
522 # Set to 1 to enable tracing.
525 { # Closure for trace: debugging aid
526 my $print_caller = 1; # ? Include calling subroutine name
527 my $main_with_colon = 'main::';
528 my $main_colon_length = length($main_with_colon);
531 return unless $to_trace; # Do nothing if global flag not set
535 local $DB::trace = 0;
536 $DB::trace = 0; # Quiet 'used only once' message
540 # Loop looking up the stack to get the first non-trace caller
545 $line_number = $caller_line;
546 (my $pkg, my $file, $caller_line, my $caller) = caller $i++;
547 $caller = $main_with_colon unless defined $caller;
549 $caller_name = $caller;
552 $caller_name =~ s/.*:://;
553 if (substr($caller_name, 0, $main_colon_length)
556 $caller_name = substr($caller_name, $main_colon_length);
559 } until ($caller_name ne 'trace');
561 # If the stack was empty, we were called from the top level
562 $caller_name = 'main' if ($caller_name eq ""
563 || $caller_name eq 'trace');
566 foreach my $string (@input) {
567 #print STDERR __LINE__, ": ", join ", ", @input, "\n";
568 if (ref $string eq 'ARRAY' || ref $string eq 'HASH') {
569 $output .= simple_dumper($string);
572 $string = "$string" if ref $string;
573 $string = $UNDEF unless defined $string;
575 $string = '""' if $string eq "";
576 $output .= " " if $output ne ""
578 && substr($output, -1, 1) ne " "
579 && substr($string, 0, 1) ne " ";
584 print STDERR sprintf "%4d: ", $line_number if defined $line_number;
585 print STDERR "$caller_name: " if $print_caller;
586 print STDERR $output, "\n";
591 # This is for a rarely used development feature that allows you to compare two
592 # versions of the Unicode standard without having to deal with changes caused
593 # by the code points introduced in the later verson. Change the 0 to a SINGLE
594 # dotted Unicode release number (e.g. 2.1). Only code points introduced in
595 # that release and earlier will be used; later ones are thrown away. You use
596 # the version number of the earliest one you want to compare; then run this
597 # program on directory structures containing each release, and compare the
598 # outputs. These outputs will therefore include only the code points common
599 # to both releases, and you can see the changes caused just by the underlying
600 # release semantic changes. For versions earlier than 3.2, you must copy a
601 # version of DAge.txt into the directory.
602 my $string_compare_versions = DEBUG && 0; # e.g., v2.1;
603 my $compare_versions = DEBUG
604 && $string_compare_versions
605 && pack "C*", split /\./, $string_compare_versions;
608 # Returns non-duplicated input values. From "Perl Best Practices:
609 # Encapsulated Cleverness". p. 455 in first edition.
612 # Arguably this breaks encapsulation, if the goal is to permit multiple
613 # distinct objects to stringify to the same value, and be interchangeable.
614 # However, for this program, no two objects stringify identically, and all
615 # lists passed to this function are either objects or strings. So this
616 # doesn't affect correctness, but it does give a couple of percent speedup.
618 return grep { ! $seen{$_}++ } @_;
621 $0 = File::Spec->canonpath($0);
623 my $make_test_script = 0; # ? Should we output a test script
624 my $write_unchanged_files = 0; # ? Should we update the output files even if
625 # we don't think they have changed
626 my $use_directory = ""; # ? Should we chdir somewhere.
627 my $pod_directory; # input directory to store the pod file.
628 my $pod_file = 'perluniprops';
629 my $t_path; # Path to the .t test file
630 my $file_list = 'mktables.lst'; # File to store input and output file names.
631 # This is used to speed up the build, by not
632 # executing the main body of the program if
633 # nothing on the list has changed since the
635 my $make_list = 1; # ? Should we write $file_list. Set to always
636 # make a list so that when the pumpking is
637 # preparing a release, s/he won't have to do
639 my $glob_list = 0; # ? Should we try to include unknown .txt files
641 my $output_range_counts = $debugging_build; # ? Should we include the number
642 # of code points in ranges in
644 my $annotate = 0; # ? Should character names be in the output
646 # Verbosity levels; 0 is quiet
647 my $NORMAL_VERBOSITY = 1;
651 my $verbosity = $NORMAL_VERBOSITY;
655 my $arg = shift @ARGV;
657 $verbosity = $VERBOSE;
659 elsif ($arg eq '-p') {
660 $verbosity = $PROGRESS;
661 $| = 1; # Flush buffers as we go.
663 elsif ($arg eq '-q') {
666 elsif ($arg eq '-w') {
667 $write_unchanged_files = 1; # update the files even if havent changed
669 elsif ($arg eq '-check') {
670 my $this = shift @ARGV;
671 my $ok = shift @ARGV;
673 print "Skipping as check params are not the same.\n";
677 elsif ($arg eq '-P' && defined ($pod_directory = shift)) {
678 -d $pod_directory or croak "Directory '$pod_directory' doesn't exist";
680 elsif ($arg eq '-maketest' || ($arg eq '-T' && defined ($t_path = shift)))
682 $make_test_script = 1;
684 elsif ($arg eq '-makelist') {
687 elsif ($arg eq '-C' && defined ($use_directory = shift)) {
688 -d $use_directory or croak "Unknown directory '$use_directory'";
690 elsif ($arg eq '-L') {
692 # Existence not tested until have chdir'd
695 elsif ($arg eq '-globlist') {
698 elsif ($arg eq '-c') {
699 $output_range_counts = ! $output_range_counts
701 elsif ($arg eq '-annotate') {
703 $debugging_build = 1;
704 $output_range_counts = 1;
708 $with_c .= 'out' if $output_range_counts; # Complements the state
710 usage: $0 [-c|-p|-q|-v|-w] [-C dir] [-L filelist] [ -P pod_dir ]
711 [ -T test_file_path ] [-globlist] [-makelist] [-maketest]
713 -c : Output comments $with_c number of code points in ranges
714 -q : Quiet Mode: Only output serious warnings.
715 -p : Set verbosity level to normal plus show progress.
716 -v : Set Verbosity level high: Show progress and non-serious
718 -w : Write files regardless
719 -C dir : Change to this directory before proceeding. All relative paths
720 except those specified by the -P and -T options will be done
721 with respect to this directory.
722 -P dir : Output $pod_file file to directory 'dir'.
723 -T path : Create a test script as 'path'; overrides -maketest
724 -L filelist : Use alternate 'filelist' instead of standard one
725 -globlist : Take as input all non-Test *.txt files in current and sub
727 -maketest : Make test script 'TestProp.pl' in current (or -C directory),
729 -makelist : Rewrite the file list $file_list based on current setup
730 -annotate : Output an annotation for each character in the table files;
731 useful for debugging mktables, looking at diffs; but is slow,
732 memory intensive; resulting tables are usable but slow and
734 -check A B : Executes $0 only if A and B are the same
739 # Stores the most-recently changed file. If none have changed, can skip the
741 my $most_recent = (stat $0)[9]; # Do this before the chdir!
743 # Change directories now, because need to read 'version' early.
744 if ($use_directory) {
745 if ($pod_directory && ! File::Spec->file_name_is_absolute($pod_directory)) {
746 $pod_directory = File::Spec->rel2abs($pod_directory);
748 if ($t_path && ! File::Spec->file_name_is_absolute($t_path)) {
749 $t_path = File::Spec->rel2abs($t_path);
751 chdir $use_directory or croak "Failed to chdir to '$use_directory':$!";
752 if ($pod_directory && File::Spec->file_name_is_absolute($pod_directory)) {
753 $pod_directory = File::Spec->abs2rel($pod_directory);
755 if ($t_path && File::Spec->file_name_is_absolute($t_path)) {
756 $t_path = File::Spec->abs2rel($t_path);
760 # Get Unicode version into regular and v-string. This is done now because
761 # various tables below get populated based on it. These tables are populated
762 # here to be near the top of the file, and so easily seeable by those needing
764 open my $VERSION, "<", "version"
765 or croak "$0: can't open required file 'version': $!\n";
766 my $string_version = <$VERSION>;
768 chomp $string_version;
769 my $v_version = pack "C*", split /\./, $string_version; # v string
771 # The following are the complete names of properties with property values that
772 # are known to not match any code points in some versions of Unicode, but that
773 # may change in the future so they should be matchable, hence an empty file is
774 # generated for them.
775 my @tables_that_may_be_empty = (
776 'Joining_Type=Left_Joining',
778 push @tables_that_may_be_empty, 'Script=Common' if $v_version le v4.0.1;
779 push @tables_that_may_be_empty, 'Title' if $v_version lt v2.0.0;
780 push @tables_that_may_be_empty, 'Script=Katakana_Or_Hiragana'
781 if $v_version ge v4.1.0;
783 # The lists below are hashes, so the key is the item in the list, and the
784 # value is the reason why it is in the list. This makes generation of
785 # documentation easier.
787 my %why_suppressed; # No file generated for these.
789 # Files aren't generated for empty extraneous properties. This is arguable.
790 # Extraneous properties generally come about because a property is no longer
791 # used in a newer version of Unicode. If we generated a file without code
792 # points, programs that used to work on that property will still execute
793 # without errors. It just won't ever match (or will always match, with \P{}).
794 # This means that the logic is now likely wrong. I (khw) think its better to
795 # find this out by getting an error message. Just move them to the table
796 # above to change this behavior
797 my %why_suppress_if_empty_warn_if_not = (
799 # It is the only property that has ever officially been removed from the
800 # Standard. The database never contained any code points for it.
801 'Special_Case_Condition' => 'Obsolete',
803 # Apparently never official, but there were code points in some versions of
804 # old-style PropList.txt
805 'Non_Break' => 'Obsolete',
808 # These would normally go in the warn table just above, but they were changed
809 # a long time before this program was written, so warnings about them are
811 if ($v_version gt v3.2.0) {
812 push @tables_that_may_be_empty,
813 'Canonical_Combining_Class=Attached_Below_Left'
816 # These are listed in the Property aliases file in 6.0, but Unihan is ignored
817 # unless explicitly added.
818 if ($v_version ge v5.2.0) {
819 my $unihan = 'Unihan; remove from list if using Unihan';
820 foreach my $table (qw (
824 kCompatibilityVariant
838 $why_suppress_if_empty_warn_if_not{$table} = $unihan;
842 # Properties that this program ignores.
843 my @unimplemented_properties = (
844 'Unicode_Radical_Stroke' # Remove if changing to handle this one.
847 # There are several types of obsolete properties defined by Unicode. These
848 # must be hand-edited for every new Unicode release.
849 my %why_deprecated; # Generates a deprecated warning message if used.
850 my %why_stabilized; # Documentation only
851 my %why_obsolete; # Documentation only
854 my $simple = 'Perl uses the more complete version of this property';
855 my $unihan = 'Unihan properties are by default not enabled in the Perl core. Instead use CPAN: Unicode::Unihan';
857 my $other_properties = 'other properties';
858 my $contributory = "Used by Unicode internally for generating $other_properties and not intended to be used stand-alone";
859 my $why_no_expand = "Deprecated by Unicode: less useful than UTF-specific calculations",
862 'Grapheme_Link' => 'Deprecated by Unicode: Duplicates ccc=vr (Canonical_Combining_Class=Virama)',
863 'Jamo_Short_Name' => $contributory,
864 'Line_Break=Surrogate' => 'Deprecated by Unicode because surrogates should never appear in well-formed text, and therefore shouldn\'t be the basis for line breaking',
865 'Other_Alphabetic' => $contributory,
866 'Other_Default_Ignorable_Code_Point' => $contributory,
867 'Other_Grapheme_Extend' => $contributory,
868 'Other_ID_Continue' => $contributory,
869 'Other_ID_Start' => $contributory,
870 'Other_Lowercase' => $contributory,
871 'Other_Math' => $contributory,
872 'Other_Uppercase' => $contributory,
876 # There is a lib/unicore/Decomposition.pl (used by Normalize.pm) which
877 # contains the same information, but without the algorithmically
878 # determinable Hangul syllables'. This file is not published, so it's
879 # existence is not noted in the comment.
880 'Decomposition_Mapping' => 'Accessible via Unicode::Normalize',
882 'ISO_Comment' => 'Apparently no demand for it, but can access it through Unicode::UCD::charinfo. Obsoleted, and code points for it removed in Unicode 5.2',
883 'Unicode_1_Name' => "$simple, and no apparent demand for it, but can access it through Unicode::UCD::charinfo. If there is no later name for a code point, then this one is used instead in charnames",
885 'Simple_Case_Folding' => "$simple. Can access this through Unicode::UCD::casefold",
886 'Simple_Lowercase_Mapping' => "$simple. Can access this through Unicode::UCD::charinfo",
887 'Simple_Titlecase_Mapping' => "$simple. Can access this through Unicode::UCD::charinfo",
888 'Simple_Uppercase_Mapping' => "$simple. Can access this through Unicode::UCD::charinfo",
890 'Name' => "Accessible via 'use charnames;'",
891 'Name_Alias' => "Accessible via 'use charnames;'",
893 FC_NFKC_Closure => 'Supplanted in usage by NFKC_Casefold; otherwise not useful',
894 Expands_On_NFC => $why_no_expand,
895 Expands_On_NFD => $why_no_expand,
896 Expands_On_NFKC => $why_no_expand,
897 Expands_On_NFKD => $why_no_expand,
900 # The following are suppressed because they were made contributory or
901 # deprecated by Unicode before Perl ever thought about supporting them.
902 foreach my $property ('Jamo_Short_Name', 'Grapheme_Link') {
903 $why_suppressed{$property} = $why_deprecated{$property};
906 # Customize the message for all the 'Other_' properties
907 foreach my $property (keys %why_deprecated) {
908 next if (my $main_property = $property) !~ s/^Other_//;
909 $why_deprecated{$property} =~ s/$other_properties/the $main_property property (which should be used instead)/;
913 if ($v_version ge 4.0.0) {
914 $why_stabilized{'Hyphen'} = 'Use the Line_Break property instead; see www.unicode.org/reports/tr14';
915 if ($v_version ge 6.0.0) {
916 $why_deprecated{'Hyphen'} = 'Supplanted by Line_Break property values; see www.unicode.org/reports/tr14';
919 if ($v_version ge 5.2.0 && $v_version lt 6.0.0) {
920 $why_obsolete{'ISO_Comment'} = 'Code points for it have been removed';
921 if ($v_version ge 6.0.0) {
922 $why_deprecated{'ISO_Comment'} = 'No longer needed for chart generation; otherwise not useful, and code points for it have been removed';
926 # Probably obsolete forever
927 if ($v_version ge v4.1.0) {
928 $why_suppressed{'Script=Katakana_Or_Hiragana'} = 'Obsolete. All code points previously matched by this have been moved to "Script=Common"';
931 # This program can create files for enumerated-like properties, such as
932 # 'Numeric_Type'. This file would be the same format as for a string
933 # property, with a mapping from code point to its value, so you could look up,
934 # for example, the script a code point is in. But no one so far wants this
935 # mapping, or they have found another way to get it since this is a new
936 # feature. So no file is generated except if it is in this list.
937 my @output_mapped_properties = split "\n", <<END;
940 # If you are using the Unihan database, you need to add the properties that
941 # you want to extract from it to this table. For your convenience, the
942 # properties in the 6.0 PropertyAliases.txt file are listed, commented out
943 my @cjk_properties = split "\n", <<'END';
944 #cjkAccountingNumeric; kAccountingNumeric
945 #cjkOtherNumeric; kOtherNumeric
946 #cjkPrimaryNumeric; kPrimaryNumeric
947 #cjkCompatibilityVariant; kCompatibilityVariant
949 #cjkIRG_GSource; kIRG_GSource
950 #cjkIRG_HSource; kIRG_HSource
951 #cjkIRG_JSource; kIRG_JSource
952 #cjkIRG_KPSource; kIRG_KPSource
953 #cjkIRG_KSource; kIRG_KSource
954 #cjkIRG_TSource; kIRG_TSource
955 #cjkIRG_USource; kIRG_USource
956 #cjkIRG_VSource; kIRG_VSource
957 #cjkRSUnicode; kRSUnicode ; Unicode_Radical_Stroke; URS
960 # Similarly for the property values. For your convenience, the lines in the
961 # 6.0 PropertyAliases.txt file are listed. Just remove the first BUT NOT both
963 my @cjk_property_values = split "\n", <<'END';
964 ## @missing: 0000..10FFFF; cjkAccountingNumeric; NaN
965 ## @missing: 0000..10FFFF; cjkCompatibilityVariant; <code point>
966 ## @missing: 0000..10FFFF; cjkIICore; <none>
967 ## @missing: 0000..10FFFF; cjkIRG_GSource; <none>
968 ## @missing: 0000..10FFFF; cjkIRG_HSource; <none>
969 ## @missing: 0000..10FFFF; cjkIRG_JSource; <none>
970 ## @missing: 0000..10FFFF; cjkIRG_KPSource; <none>
971 ## @missing: 0000..10FFFF; cjkIRG_KSource; <none>
972 ## @missing: 0000..10FFFF; cjkIRG_TSource; <none>
973 ## @missing: 0000..10FFFF; cjkIRG_USource; <none>
974 ## @missing: 0000..10FFFF; cjkIRG_VSource; <none>
975 ## @missing: 0000..10FFFF; cjkOtherNumeric; NaN
976 ## @missing: 0000..10FFFF; cjkPrimaryNumeric; NaN
977 ## @missing: 0000..10FFFF; cjkRSUnicode; <none>
980 # The input files don't list every code point. Those not listed are to be
981 # defaulted to some value. Below are hard-coded what those values are for
982 # non-binary properties as of 5.1. Starting in 5.0, there are
983 # machine-parsable comment lines in the files the give the defaults; so this
984 # list shouldn't have to be extended. The claim is that all missing entries
985 # for binary properties will default to 'N'. Unicode tried to change that in
986 # 5.2, but the beta period produced enough protest that they backed off.
988 # The defaults for the fields that appear in UnicodeData.txt in this hash must
989 # be in the form that it expects. The others may be synonyms.
990 my $CODE_POINT = '<code point>';
991 my %default_mapping = (
993 # Bidi_Class => Complicated; set in code
994 Bidi_Mirroring_Glyph => "",
996 Canonical_Combining_Class => 0,
997 Case_Folding => $CODE_POINT,
998 Decomposition_Mapping => $CODE_POINT,
999 Decomposition_Type => 'None',
1000 East_Asian_Width => "Neutral",
1001 FC_NFKC_Closure => $CODE_POINT,
1002 General_Category => 'Cn',
1003 Grapheme_Cluster_Break => 'Other',
1004 Hangul_Syllable_Type => 'NA',
1006 Jamo_Short_Name => "",
1007 Joining_Group => "No_Joining_Group",
1008 # Joining_Type => Complicated; set in code
1009 kIICore => 'N', # Is converted to binary
1010 #Line_Break => Complicated; set in code
1011 Lowercase_Mapping => $CODE_POINT,
1018 Numeric_Type => 'None',
1019 Numeric_Value => 'NaN',
1020 Script => ($v_version le 4.1.0) ? 'Common' : 'Unknown',
1021 Sentence_Break => 'Other',
1022 Simple_Case_Folding => $CODE_POINT,
1023 Simple_Lowercase_Mapping => $CODE_POINT,
1024 Simple_Titlecase_Mapping => $CODE_POINT,
1025 Simple_Uppercase_Mapping => $CODE_POINT,
1026 Titlecase_Mapping => $CODE_POINT,
1027 Unicode_1_Name => "",
1028 Unicode_Radical_Stroke => "",
1029 Uppercase_Mapping => $CODE_POINT,
1030 Word_Break => 'Other',
1033 # Below are files that Unicode furnishes, but this program ignores, and why
1034 my %ignored_files = (
1035 'CJKRadicals.txt' => 'Unihan data',
1036 'Index.txt' => 'An index, not actual data',
1037 'NamedSqProv.txt' => 'Not officially part of the Unicode standard; Append it to NamedSequences.txt if you want to process the contents.',
1038 'NamesList.txt' => 'Just adds commentary',
1039 'NormalizationCorrections.txt' => 'Data is already in other files.',
1040 'Props.txt' => 'Adds nothing to PropList.txt; only in very early releases',
1041 'ReadMe.txt' => 'Just comments',
1042 'README.TXT' => 'Just comments',
1043 'StandardizedVariants.txt' => 'Only for glyph changes, not a Unicode character property. Does not fit into current scheme where one code point is mapped',
1044 'EmojiSources.txt' => 'Not of general utility: for Japanese legacy cell-phone applications',
1045 'IndicMatraCategory.txt' => 'Provisional',
1046 'IndicSyllabicCategory.txt' => 'Provisional',
1047 'ScriptExtensions.txt' => 'Provisional',
1050 ### End of externally interesting definitions, except for @input_file_objects
1053 # !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
1054 # This file is machine-generated by $0 from the Unicode
1055 # database, Version $string_version. Any changes made here will be lost!
1058 my $INTERNAL_ONLY=<<"EOF";
1060 # !!!!!!! INTERNAL PERL USE ONLY !!!!!!!
1061 # This file is for internal use by the Perl program only. The format and even
1062 # the name or existence of this file are subject to change without notice.
1063 # Don't use it directly.
1066 my $DEVELOPMENT_ONLY=<<"EOF";
1067 # !!!!!!! DEVELOPMENT USE ONLY !!!!!!!
1068 # This file contains information artificially constrained to code points
1069 # present in Unicode release $string_compare_versions.
1070 # IT CANNOT BE RELIED ON. It is for use during development only and should
1071 # not be used for production.
1075 my $LAST_UNICODE_CODEPOINT_STRING = "10FFFF";
1076 my $LAST_UNICODE_CODEPOINT = hex $LAST_UNICODE_CODEPOINT_STRING;
1077 my $MAX_UNICODE_CODEPOINTS = $LAST_UNICODE_CODEPOINT + 1;
1079 # Matches legal code point. 4-6 hex numbers, If there are 6, the first
1080 # two must be 10; if there are 5, the first must not be a 0. Written this way
1081 # to decrease backtracking
1083 qr/ \b (?: 10[0-9A-F]{4} | [1-9A-F][0-9A-F]{4} | [0-9A-F]{4} ) \b/x;
1085 # This matches the beginning of the line in the Unicode db files that give the
1086 # defaults for code points not listed (i.e., missing) in the file. The code
1087 # depends on this ending with a semi-colon, so it can assume it is a valid
1088 # field when the line is split() by semi-colons
1089 my $missing_defaults_prefix =
1090 qr/^#\s+\@missing:\s+0000\.\.$LAST_UNICODE_CODEPOINT_STRING\s*;/;
1092 # Property types. Unicode has more types, but these are sufficient for our
1094 my $UNKNOWN = -1; # initialized to illegal value
1095 my $NON_STRING = 1; # Either binary or enum
1097 my $ENUM = 3; # Include catalog
1098 my $STRING = 4; # Anything else: string or misc
1100 # Some input files have lines that give default values for code points not
1101 # contained in the file. Sometimes these should be ignored.
1102 my $NO_DEFAULTS = 0; # Must evaluate to false
1103 my $NOT_IGNORED = 1;
1106 # Range types. Each range has a type. Most ranges are type 0, for normal,
1107 # and will appear in the main body of the tables in the output files, but
1108 # there are other types of ranges as well, listed below, that are specially
1109 # handled. There are pseudo-types as well that will never be stored as a
1110 # type, but will affect the calculation of the type.
1112 # 0 is for normal, non-specials
1113 my $MULTI_CP = 1; # Sequence of more than code point
1114 my $HANGUL_SYLLABLE = 2;
1115 my $CP_IN_NAME = 3; # The NAME contains the code point appended to it.
1116 my $NULL = 4; # The map is to the null string; utf8.c can't
1117 # handle these, nor is there an accepted syntax
1118 # for them in \p{} constructs
1119 my $COMPUTE_NO_MULTI_CP = 5; # Pseudo-type; means that ranges that would
1120 # otherwise be $MULTI_CP type are instead type 0
1122 # process_generic_property_file() can accept certain overrides in its input.
1123 # Each of these must begin AND end with $CMD_DELIM.
1124 my $CMD_DELIM = "\a";
1125 my $REPLACE_CMD = 'replace'; # Override the Replace
1126 my $MAP_TYPE_CMD = 'map_type'; # Override the Type
1131 # Values for the Replace argument to add_range.
1132 # $NO # Don't replace; add only the code points not
1134 my $IF_NOT_EQUIVALENT = 1; # Replace only under certain conditions; details in
1135 # the comments at the subroutine definition.
1136 my $UNCONDITIONALLY = 2; # Replace without conditions.
1137 my $MULTIPLE = 4; # Don't replace, but add a duplicate record if
1139 my $CROAK = 5; # Die with an error if is already there
1141 # Flags to give property statuses. The phrases are to remind maintainers that
1142 # if the flag is changed, the indefinite article referring to it in the
1143 # documentation may need to be as well.
1145 my $SUPPRESSED = 'z'; # The character should never actually be seen, since
1147 my $PLACEHOLDER = 'P'; # Implies no pod entry generated
1148 my $DEPRECATED = 'D';
1149 my $a_bold_deprecated = "a 'B<$DEPRECATED>'";
1150 my $A_bold_deprecated = "A 'B<$DEPRECATED>'";
1151 my $DISCOURAGED = 'X';
1152 my $a_bold_discouraged = "an 'B<$DISCOURAGED>'";
1153 my $A_bold_discouraged = "An 'B<$DISCOURAGED>'";
1155 my $a_bold_stricter = "a 'B<$STRICTER>'";
1156 my $A_bold_stricter = "A 'B<$STRICTER>'";
1157 my $STABILIZED = 'S';
1158 my $a_bold_stabilized = "an 'B<$STABILIZED>'";
1159 my $A_bold_stabilized = "An 'B<$STABILIZED>'";
1161 my $a_bold_obsolete = "an 'B<$OBSOLETE>'";
1162 my $A_bold_obsolete = "An 'B<$OBSOLETE>'";
1164 my %status_past_participles = (
1165 $DISCOURAGED => 'discouraged',
1166 $SUPPRESSED => 'should never be generated',
1167 $STABILIZED => 'stabilized',
1168 $OBSOLETE => 'obsolete',
1169 $DEPRECATED => 'deprecated',
1172 # The format of the values of the tables:
1173 my $EMPTY_FORMAT = "";
1174 my $BINARY_FORMAT = 'b';
1175 my $DECIMAL_FORMAT = 'd';
1176 my $FLOAT_FORMAT = 'f';
1177 my $INTEGER_FORMAT = 'i';
1178 my $HEX_FORMAT = 'x';
1179 my $RATIONAL_FORMAT = 'r';
1180 my $STRING_FORMAT = 's';
1181 my $DECOMP_STRING_FORMAT = 'c';
1183 my %map_table_formats = (
1184 $BINARY_FORMAT => 'binary',
1185 $DECIMAL_FORMAT => 'single decimal digit',
1186 $FLOAT_FORMAT => 'floating point number',
1187 $INTEGER_FORMAT => 'integer',
1188 $HEX_FORMAT => 'positive hex whole number; a code point',
1189 $RATIONAL_FORMAT => 'rational: an integer or a fraction',
1190 $STRING_FORMAT => 'string',
1191 $DECOMP_STRING_FORMAT => 'Perl\'s internal (Normalize.pm) decomposition mapping',
1194 # Unicode didn't put such derived files in a separate directory at first.
1195 my $EXTRACTED_DIR = (-d 'extracted') ? 'extracted' : "";
1196 my $EXTRACTED = ($EXTRACTED_DIR) ? "$EXTRACTED_DIR/" : "";
1197 my $AUXILIARY = 'auxiliary';
1199 # Hashes that will eventually go into Heavy.pl for the use of utf8_heavy.pl
1200 my %loose_to_file_of; # loosely maps table names to their respective
1202 my %stricter_to_file_of; # same; but for stricter mapping.
1203 my %nv_floating_to_rational; # maps numeric values floating point numbers to
1204 # their rational equivalent
1205 my %loose_property_name_of; # Loosely maps property names to standard form
1207 # These constants names and values were taken from the Unicode standard,
1208 # version 5.1, section 3.12. They are used in conjunction with Hangul
1209 # syllables. The '_string' versions are so generated tables can retain the
1210 # hex format, which is the more familiar value
1211 my $SBase_string = "0xAC00";
1212 my $SBase = CORE::hex $SBase_string;
1213 my $LBase_string = "0x1100";
1214 my $LBase = CORE::hex $LBase_string;
1215 my $VBase_string = "0x1161";
1216 my $VBase = CORE::hex $VBase_string;
1217 my $TBase_string = "0x11A7";
1218 my $TBase = CORE::hex $TBase_string;
1223 my $NCount = $VCount * $TCount;
1225 # For Hangul syllables; These store the numbers from Jamo.txt in conjunction
1226 # with the above published constants.
1228 my %Jamo_L; # Leading consonants
1229 my %Jamo_V; # Vowels
1230 my %Jamo_T; # Trailing consonants
1232 my @backslash_X_tests; # List of tests read in for testing \X
1233 my @unhandled_properties; # Will contain a list of properties found in
1234 # the input that we didn't process.
1235 my @match_properties; # Properties that have match tables, to be
1237 my @map_properties; # Properties that get map files written
1238 my @named_sequences; # NamedSequences.txt contents.
1239 my %potential_files; # Generated list of all .txt files in the directory
1240 # structure so we can warn if something is being
1242 my @files_actually_output; # List of files we generated.
1243 my @more_Names; # Some code point names are compound; this is used
1244 # to store the extra components of them.
1245 my $MIN_FRACTION_LENGTH = 3; # How many digits of a floating point number at
1246 # the minimum before we consider it equivalent to a
1247 # candidate rational
1248 my $MAX_FLOATING_SLOP = 10 ** - $MIN_FRACTION_LENGTH; # And in floating terms
1250 # These store references to certain commonly used property objects
1257 # Are there conflicting names because of beginning with 'In_', or 'Is_'
1258 my $has_In_conflicts = 0;
1259 my $has_Is_conflicts = 0;
1261 sub internal_file_to_platform ($) {
1262 # Convert our file paths which have '/' separators to those of the
1266 return undef unless defined $file;
1268 return File::Spec->join(split '/', $file);
1271 sub file_exists ($) { # platform independent '-e'. This program internally
1272 # uses slash as a path separator.
1274 return 0 if ! defined $file;
1275 return -e internal_file_to_platform($file);
1279 # Returns the address of the blessed input object.
1280 # It doesn't check for blessedness because that would do a string eval
1281 # every call, and the program is structured so that this is never called
1282 # for a non-blessed object.
1284 no overloading; # If overloaded, numifying below won't work.
1286 # Numifying a ref gives its address.
1287 return pack 'J', $_[0];
1290 # These are used only if $annotate is true.
1291 # The entire range of Unicode characters is examined to populate these
1292 # after all the input has been processed. But most can be skipped, as they
1293 # have the same descriptive phrases, such as being unassigned
1294 my @viacode; # Contains the 1 million character names
1295 my @printable; # boolean: And are those characters printable?
1296 my @annotate_char_type; # Contains a type of those characters, specifically
1297 # for the purposes of annotation.
1298 my $annotate_ranges; # A map of ranges of code points that have the same
1299 # name for the purposes of annoation. They map to the
1300 # upper edge of the range, so that the end point can
1301 # be immediately found. This is used to skip ahead to
1302 # the end of a range, and avoid processing each
1303 # individual code point in it.
1304 my $unassigned_sans_noncharacters; # A Range_List of the unassigned
1305 # characters, but excluding those which are
1306 # also noncharacter code points
1308 # The annotation types are an extension of the regular range types, though
1309 # some of the latter are folded into one. Make the new types negative to
1310 # avoid conflicting with the regular types
1311 my $SURROGATE_TYPE = -1;
1312 my $UNASSIGNED_TYPE = -2;
1313 my $PRIVATE_USE_TYPE = -3;
1314 my $NONCHARACTER_TYPE = -4;
1315 my $CONTROL_TYPE = -5;
1316 my $UNKNOWN_TYPE = -6; # Used only if there is a bug in this program
1318 sub populate_char_info ($) {
1319 # Used only with the $annotate option. Populates the arrays with the
1320 # input code point's info that are needed for outputting more detailed
1321 # comments. If calling context wants a return, it is the end point of
1322 # any contiguous range of characters that share essentially the same info
1325 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
1327 $viacode[$i] = $perl_charname->value_of($i) || "";
1329 # A character is generally printable if Unicode says it is,
1330 # but below we make sure that most Unicode general category 'C' types
1332 $printable[$i] = $print->contains($i);
1334 $annotate_char_type[$i] = $perl_charname->type_of($i) || 0;
1336 # Only these two regular types are treated specially for annotations
1338 $annotate_char_type[$i] = 0 if $annotate_char_type[$i] != $CP_IN_NAME
1339 && $annotate_char_type[$i] != $HANGUL_SYLLABLE;
1341 # Give a generic name to all code points that don't have a real name.
1342 # We output ranges, if applicable, for these. Also calculate the end
1343 # point of the range.
1345 if (! $viacode[$i]) {
1346 if ($gc-> table('Surrogate')->contains($i)) {
1347 $viacode[$i] = 'Surrogate';
1348 $annotate_char_type[$i] = $SURROGATE_TYPE;
1350 $end = $gc->table('Surrogate')->containing_range($i)->end;
1352 elsif ($gc-> table('Private_use')->contains($i)) {
1353 $viacode[$i] = 'Private Use';
1354 $annotate_char_type[$i] = $PRIVATE_USE_TYPE;
1356 $end = $gc->table('Private_Use')->containing_range($i)->end;
1358 elsif (Property::property_ref('Noncharacter_Code_Point')-> table('Y')->
1361 $viacode[$i] = 'Noncharacter';
1362 $annotate_char_type[$i] = $NONCHARACTER_TYPE;
1364 $end = property_ref('Noncharacter_Code_Point')->table('Y')->
1365 containing_range($i)->end;
1367 elsif ($gc-> table('Control')->contains($i)) {
1368 $viacode[$i] = 'Control';
1369 $annotate_char_type[$i] = $CONTROL_TYPE;
1371 $end = 0x81 if $i == 0x80; # Hard-code this one known case
1373 elsif ($gc-> table('Unassigned')->contains($i)) {
1374 $viacode[$i] = 'Unassigned, block=' . $block-> value_of($i);
1375 $annotate_char_type[$i] = $UNASSIGNED_TYPE;
1378 # Because we name the unassigned by the blocks they are in, it
1379 # can't go past the end of that block, and it also can't go past
1380 # the unassigned range it is in. The special table makes sure
1381 # that the non-characters, which are unassigned, are separated
1383 $end = min($block->containing_range($i)->end,
1384 $unassigned_sans_noncharacters-> containing_range($i)->
1388 Carp::my_carp_bug("Can't figure out how to annotate "
1389 . sprintf("U+%04X", $i)
1390 . ". Proceeding anyway.");
1391 $viacode[$i] = 'UNKNOWN';
1392 $annotate_char_type[$i] = $UNKNOWN_TYPE;
1397 # Here, has a name, but if it's one in which the code point number is
1398 # appended to the name, do that.
1399 elsif ($annotate_char_type[$i] == $CP_IN_NAME) {
1400 $viacode[$i] .= sprintf("-%04X", $i);
1401 $end = $perl_charname->containing_range($i)->end;
1404 # And here, has a name, but if it's a hangul syllable one, replace it with
1405 # the correct name from the Unicode algorithm
1406 elsif ($annotate_char_type[$i] == $HANGUL_SYLLABLE) {
1408 my $SIndex = $i - $SBase;
1409 my $L = $LBase + $SIndex / $NCount;
1410 my $V = $VBase + ($SIndex % $NCount) / $TCount;
1411 my $T = $TBase + $SIndex % $TCount;
1412 $viacode[$i] = "HANGUL SYLLABLE $Jamo{$L}$Jamo{$V}";
1413 $viacode[$i] .= $Jamo{$T} if $T != $TBase;
1414 $end = $perl_charname->containing_range($i)->end;
1417 return if ! defined wantarray;
1418 return $i if ! defined $end; # If not a range, return the input
1420 # Save this whole range so can find the end point quickly
1421 $annotate_ranges->add_map($i, $end, $end);
1426 # Commented code below should work on Perl 5.8.
1427 ## This 'require' doesn't necessarily work in miniperl, and even if it does,
1428 ## the native perl version of it (which is what would operate under miniperl)
1429 ## is extremely slow, as it does a string eval every call.
1430 #my $has_fast_scalar_util = $
\18 !~ /miniperl/
1431 # && defined eval "require Scalar::Util";
1434 # # Returns the address of the blessed input object. Uses the XS version if
1435 # # available. It doesn't check for blessedness because that would do a
1436 # # string eval every call, and the program is structured so that this is
1437 # # never called for a non-blessed object.
1439 # return Scalar::Util::refaddr($_[0]) if $has_fast_scalar_util;
1441 # # Check at least that is a ref.
1442 # my $pkg = ref($_[0]) or return undef;
1444 # # Change to a fake package to defeat any overloaded stringify
1445 # bless $_[0], 'main::Fake';
1447 # # Numifying a ref gives its address.
1448 # my $addr = pack 'J', $_[0];
1450 # # Return to original class
1451 # bless $_[0], $pkg;
1458 return $a if $a >= $b;
1465 return $a if $a <= $b;
1469 sub clarify_number ($) {
1470 # This returns the input number with underscores inserted every 3 digits
1471 # in large (5 digits or more) numbers. Input must be entirely digits, not
1475 my $pos = length($number) - 3;
1476 return $number if $pos <= 1;
1478 substr($number, $pos, 0) = '_';
1487 # These routines give a uniform treatment of messages in this program. They
1488 # are placed in the Carp package to cause the stack trace to not include them,
1489 # although an alternative would be to use another package and set @CARP_NOT
1492 our $Verbose = 1 if main::DEBUG; # Useful info when debugging
1494 # This is a work-around suggested by Nicholas Clark to fix a problem with Carp
1495 # and overload trying to load Scalar:Util under miniperl. See
1496 # http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2009-11/msg01057.html
1497 undef $overload::VERSION;
1500 my $message = shift || "";
1501 my $nofold = shift || 0;
1504 $message = main::join_lines($message);
1505 $message =~ s/^$0: *//; # Remove initial program name
1506 $message =~ s/[.;,]+$//; # Remove certain ending punctuation
1507 $message = "\n$0: $message;";
1509 # Fold the message with program name, semi-colon end punctuation
1510 # (which looks good with the message that carp appends to it), and a
1511 # hanging indent for continuation lines.
1512 $message = main::simple_fold($message, "", 4) unless $nofold;
1513 $message =~ s/\n$//; # Remove the trailing nl so what carp
1514 # appends is to the same line
1517 return $message if defined wantarray; # If a caller just wants the msg
1524 # This is called when it is clear that the problem is caused by a bug in
1527 my $message = shift;
1528 $message =~ s/^$0: *//;
1529 $message = my_carp("Bug in $0. Please report it by running perlbug or if that is unavailable, by sending email to perbug\@perl.org:\n$message");
1534 sub carp_too_few_args {
1536 my_carp_bug("Wrong number of arguments: to 'carp_too_few_arguments'. No action taken.");
1540 my $args_ref = shift;
1543 my_carp_bug("Need at least $count arguments to "
1545 . ". Instead got: '"
1546 . join ', ', @$args_ref
1547 . "'. No action taken.");
1551 sub carp_extra_args {
1552 my $args_ref = shift;
1553 my_carp_bug("Too many arguments to 'carp_extra_args': (" . join(', ', @_) . "); Extras ignored.") if @_;
1555 unless (ref $args_ref) {
1556 my_carp_bug("Argument to 'carp_extra_args' ($args_ref) must be a ref. Not checking arguments.");
1559 my ($package, $file, $line) = caller;
1560 my $subroutine = (caller 1)[3];
1563 if (ref $args_ref eq 'HASH') {
1564 foreach my $key (keys %$args_ref) {
1565 $args_ref->{$key} = $UNDEF unless defined $args_ref->{$key};
1567 $list = join ', ', each %{$args_ref};
1569 elsif (ref $args_ref eq 'ARRAY') {
1570 foreach my $arg (@$args_ref) {
1571 $arg = $UNDEF unless defined $arg;
1573 $list = join ', ', @$args_ref;
1576 my_carp_bug("Can't cope with ref "
1578 . " . argument to 'carp_extra_args'. Not checking arguments.");
1582 my_carp_bug("Unrecognized parameters in options: '$list' to $subroutine. Skipped.");
1590 # This program uses the inside-out method for objects, as recommended in
1591 # "Perl Best Practices". This closure aids in generating those. There
1592 # are two routines. setup_package() is called once per package to set
1593 # things up, and then set_access() is called for each hash representing a
1594 # field in the object. These routines arrange for the object to be
1595 # properly destroyed when no longer used, and for standard accessor
1596 # functions to be generated. If you need more complex accessors, just
1597 # write your own and leave those accesses out of the call to set_access().
1598 # More details below.
1600 my %constructor_fields; # fields that are to be used in constructors; see
1603 # The values of this hash will be the package names as keys to other
1604 # hashes containing the name of each field in the package as keys, and
1605 # references to their respective hashes as values.
1609 # Sets up the package, creating standard DESTROY and dump methods
1610 # (unless already defined). The dump method is used in debugging by
1612 # The optional parameters are:
1613 # a) a reference to a hash, that gets populated by later
1614 # set_access() calls with one of the accesses being
1615 # 'constructor'. The caller can then refer to this, but it is
1616 # not otherwise used by these two routines.
1617 # b) a reference to a callback routine to call during destruction
1618 # of the object, before any fields are actually destroyed
1621 my $constructor_ref = delete $args{'Constructor_Fields'};
1622 my $destroy_callback = delete $args{'Destroy_Callback'};
1623 Carp::carp_extra_args(\@_) if main::DEBUG && %args;
1626 my $package = (caller)[0];
1628 $package_fields{$package} = \%fields;
1629 $constructor_fields{$package} = $constructor_ref;
1631 unless ($package->can('DESTROY')) {
1632 my $destroy_name = "${package}::DESTROY";
1635 # Use typeglob to give the anonymous subroutine the name we want
1636 *$destroy_name = sub {
1638 my $addr = do { no overloading; pack 'J', $self; };
1640 $self->$destroy_callback if $destroy_callback;
1641 foreach my $field (keys %{$package_fields{$package}}) {
1642 #print STDERR __LINE__, ": Destroying ", ref $self, " ", sprintf("%04X", $addr), ": ", $field, "\n";
1643 delete $package_fields{$package}{$field}{$addr};
1649 unless ($package->can('dump')) {
1650 my $dump_name = "${package}::dump";
1654 return dump_inside_out($self, $package_fields{$package}, @_);
1661 # Arrange for the input field to be garbage collected when no longer
1662 # needed. Also, creates standard accessor functions for the field
1663 # based on the optional parameters-- none if none of these parameters:
1664 # 'addable' creates an 'add_NAME()' accessor function.
1665 # 'readable' or 'readable_array' creates a 'NAME()' accessor
1667 # 'settable' creates a 'set_NAME()' accessor function.
1668 # 'constructor' doesn't create an accessor function, but adds the
1669 # field to the hash that was previously passed to
1671 # Any of the accesses can be abbreviated down, so that 'a', 'ad',
1672 # 'add' etc. all mean 'addable'.
1673 # The read accessor function will work on both array and scalar
1674 # values. If another accessor in the parameter list is 'a', the read
1675 # access assumes an array. You can also force it to be array access
1676 # by specifying 'readable_array' instead of 'readable'
1678 # A sort-of 'protected' access can be set-up by preceding the addable,
1679 # readable or settable with some initial portion of 'protected_' (but,
1680 # the underscore is required), like 'p_a', 'pro_set', etc. The
1681 # "protection" is only by convention. All that happens is that the
1682 # accessor functions' names begin with an underscore. So instead of
1683 # calling set_foo, the call is _set_foo. (Real protection could be
1684 # accomplished by having a new subroutine, end_package, called at the
1685 # end of each package, and then storing the __LINE__ ranges and
1686 # checking them on every accessor. But that is way overkill.)
1688 # We create anonymous subroutines as the accessors and then use
1689 # typeglobs to assign them to the proper package and name
1691 my $name = shift; # Name of the field
1692 my $field = shift; # Reference to the inside-out hash containing the
1695 my $package = (caller)[0];
1697 if (! exists $package_fields{$package}) {
1698 croak "$0: Must call 'setup_package' before 'set_access'";
1701 # Stash the field so DESTROY can get it.
1702 $package_fields{$package}{$name} = $field;
1704 # Remaining arguments are the accessors. For each...
1705 foreach my $access (@_) {
1706 my $access = lc $access;
1710 # Match the input as far as it goes.
1711 if ($access =~ /^(p[^_]*)_/) {
1713 if (substr('protected_', 0, length $protected)
1717 # Add 1 for the underscore not included in $protected
1718 $access = substr($access, length($protected) + 1);
1726 if (substr('addable', 0, length $access) eq $access) {
1727 my $subname = "${package}::${protected}add_$name";
1730 # add_ accessor. Don't add if already there, which we
1731 # determine using 'eq' for scalars and '==' otherwise.
1734 return Carp::carp_too_few_args(\@_, 2) if main::DEBUG && @_ < 2;
1737 my $addr = do { no overloading; pack 'J', $self; };
1738 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
1740 return if grep { $value == $_ } @{$field->{$addr}};
1743 return if grep { $value eq $_ } @{$field->{$addr}};
1745 push @{$field->{$addr}}, $value;
1749 elsif (substr('constructor', 0, length $access) eq $access) {
1751 Carp::my_carp_bug("Can't set-up 'protected' constructors")
1754 $constructor_fields{$package}{$name} = $field;
1757 elsif (substr('readable_array', 0, length $access) eq $access) {
1759 # Here has read access. If one of the other parameters for
1760 # access is array, or this one specifies array (by being more
1761 # than just 'readable_'), then create a subroutine that
1762 # assumes the data is an array. Otherwise just a scalar
1763 my $subname = "${package}::${protected}$name";
1764 if (grep { /^a/i } @_
1765 or length($access) > length('readable_'))
1770 Carp::carp_extra_args(\@_) if main::DEBUG && @_ > 1;
1771 my $addr = do { no overloading; pack 'J', $_[0]; };
1772 if (ref $field->{$addr} ne 'ARRAY') {
1773 my $type = ref $field->{$addr};
1774 $type = 'scalar' unless $type;
1775 Carp::my_carp_bug("Trying to read $name as an array when it is a $type. Big problems.");
1778 return scalar @{$field->{$addr}} unless wantarray;
1780 # Make a copy; had problems with caller modifying the
1781 # original otherwise
1782 my @return = @{$field->{$addr}};
1788 # Here not an array value, a simpler function.
1792 Carp::carp_extra_args(\@_) if main::DEBUG && @_ > 1;
1794 return $field->{pack 'J', $_[0]};
1798 elsif (substr('settable', 0, length $access) eq $access) {
1799 my $subname = "${package}::${protected}set_$name";
1804 return Carp::carp_too_few_args(\@_, 2) if @_ < 2;
1805 Carp::carp_extra_args(\@_) if @_ > 2;
1807 # $self is $_[0]; $value is $_[1]
1809 $field->{pack 'J', $_[0]} = $_[1];
1814 Carp::my_carp_bug("Unknown accessor type $access. No accessor set.");
1823 # All input files use this object, which stores various attributes about them,
1824 # and provides for convenient, uniform handling. The run method wraps the
1825 # processing. It handles all the bookkeeping of opening, reading, and closing
1826 # the file, returning only significant input lines.
1828 # Each object gets a handler which processes the body of the file, and is
1829 # called by run(). Most should use the generic, default handler, which has
1830 # code scrubbed to handle things you might not expect. A handler should
1831 # basically be a while(next_line()) {...} loop.
1833 # You can also set up handlers to
1834 # 1) call before the first line is read for pre processing
1835 # 2) call to adjust each line of the input before the main handler gets them
1836 # 3) call upon EOF before the main handler exits its loop
1837 # 4) call at the end for post processing
1839 # $_ is used to store the input line, and is to be filtered by the
1840 # each_line_handler()s. So, if the format of the line is not in the desired
1841 # format for the main handler, these are used to do that adjusting. They can
1842 # be stacked (by enclosing them in an [ anonymous array ] in the constructor,
1843 # so the $_ output of one is used as the input to the next. None of the other
1844 # handlers are stackable, but could easily be changed to be so.
1846 # Most of the handlers can call insert_lines() or insert_adjusted_lines()
1847 # which insert the parameters as lines to be processed before the next input
1848 # file line is read. This allows the EOF handler to flush buffers, for
1849 # example. The difference between the two routines is that the lines inserted
1850 # by insert_lines() are subjected to the each_line_handler()s. (So if you
1851 # called it from such a handler, you would get infinite recursion.) Lines
1852 # inserted by insert_adjusted_lines() go directly to the main handler without
1853 # any adjustments. If the post-processing handler calls any of these, there
1854 # will be no effect. Some error checking for these conditions could be added,
1855 # but it hasn't been done.
1857 # carp_bad_line() should be called to warn of bad input lines, which clears $_
1858 # to prevent further processing of the line. This routine will output the
1859 # message as a warning once, and then keep a count of the lines that have the
1860 # same message, and output that count at the end of the file's processing.
1861 # This keeps the number of messages down to a manageable amount.
1863 # get_missings() should be called to retrieve any @missing input lines.
1864 # Messages will be raised if this isn't done if the options aren't to ignore
1867 sub trace { return main::trace(@_); }
1870 # Keep track of fields that are to be put into the constructor.
1871 my %constructor_fields;
1873 main::setup_package(Constructor_Fields => \%constructor_fields);
1875 my %file; # Input file name, required
1876 main::set_access('file', \%file, qw{ c r });
1878 my %first_released; # Unicode version file was first released in, required
1879 main::set_access('first_released', \%first_released, qw{ c r });
1881 my %handler; # Subroutine to process the input file, defaults to
1882 # 'process_generic_property_file'
1883 main::set_access('handler', \%handler, qw{ c });
1886 # name of property this file is for. defaults to none, meaning not
1887 # applicable, or is otherwise determinable, for example, from each line.
1888 main::set_access('property', \%property, qw{ c });
1891 # If this is true, the file is optional. If not present, no warning is
1892 # output. If it is present, the string given by this parameter is
1893 # evaluated, and if false the file is not processed.
1894 main::set_access('optional', \%optional, 'c', 'r');
1897 # This is used for debugging, to skip processing of all but a few input
1898 # files. Add 'non_skip => 1' to the constructor for those files you want
1899 # processed when you set the $debug_skip global.
1900 main::set_access('non_skip', \%non_skip, 'c');
1903 # This is used to skip processing of this input file semi-permanently.
1904 # It is used for files that we aren't planning to process anytime soon,
1905 # but want to allow to be in the directory and not raise a message that we
1906 # are not handling. Mostly for test files. This is in contrast to the
1907 # non_skip element, which is supposed to be used very temporarily for
1908 # debugging. Sets 'optional' to 1
1909 main::set_access('skip', \%skip, 'c');
1911 my %each_line_handler;
1912 # list of subroutines to look at and filter each non-comment line in the
1913 # file. defaults to none. The subroutines are called in order, each is
1914 # to adjust $_ for the next one, and the final one adjusts it for
1916 main::set_access('each_line_handler', \%each_line_handler, 'c');
1918 my %has_missings_defaults;
1919 # ? Are there lines in the file giving default values for code points
1920 # missing from it?. Defaults to NO_DEFAULTS. Otherwise NOT_IGNORED is
1921 # the norm, but IGNORED means it has such lines, but the handler doesn't
1922 # use them. Having these three states allows us to catch changes to the
1923 # UCD that this program should track
1924 main::set_access('has_missings_defaults',
1925 \%has_missings_defaults, qw{ c r });
1928 # Subroutine to call before doing anything else in the file. If undef, no
1929 # such handler is called.
1930 main::set_access('pre_handler', \%pre_handler, qw{ c });
1933 # Subroutine to call upon getting an EOF on the input file, but before
1934 # that is returned to the main handler. This is to allow buffers to be
1935 # flushed. The handler is expected to call insert_lines() or
1936 # insert_adjusted() with the buffered material
1937 main::set_access('eof_handler', \%eof_handler, qw{ c r });
1940 # Subroutine to call after all the lines of the file are read in and
1941 # processed. If undef, no such handler is called.
1942 main::set_access('post_handler', \%post_handler, qw{ c });
1944 my %progress_message;
1945 # Message to print to display progress in lieu of the standard one
1946 main::set_access('progress_message', \%progress_message, qw{ c });
1949 # cache open file handle, internal. Is undef if file hasn't been
1950 # processed at all, empty if has;
1951 main::set_access('handle', \%handle);
1954 # cache of lines added virtually to the file, internal
1955 main::set_access('added_lines', \%added_lines);
1958 # cache of errors found, internal
1959 main::set_access('errors', \%errors);
1962 # storage of '@missing' defaults lines
1963 main::set_access('missings', \%missings);
1968 my $self = bless \do{ my $anonymous_scalar }, $class;
1969 my $addr = do { no overloading; pack 'J', $self; };
1972 $handler{$addr} = \&main::process_generic_property_file;
1973 $non_skip{$addr} = 0;
1975 $has_missings_defaults{$addr} = $NO_DEFAULTS;
1976 $handle{$addr} = undef;
1977 $added_lines{$addr} = [ ];
1978 $each_line_handler{$addr} = [ ];
1979 $errors{$addr} = { };
1980 $missings{$addr} = [ ];
1982 # Two positional parameters.
1983 return Carp::carp_too_few_args(\@_, 2) if main::DEBUG && @_ < 2;
1984 $file{$addr} = main::internal_file_to_platform(shift);
1985 $first_released{$addr} = shift;
1987 # The rest of the arguments are key => value pairs
1988 # %constructor_fields has been set up earlier to list all possible
1989 # ones. Either set or push, depending on how the default has been set
1992 foreach my $key (keys %args) {
1993 my $argument = $args{$key};
1995 # Note that the fields are the lower case of the constructor keys
1996 my $hash = $constructor_fields{lc $key};
1997 if (! defined $hash) {
1998 Carp::my_carp_bug("Unrecognized parameters '$key => $argument' to new() for $self. Skipped");
2001 if (ref $hash->{$addr} eq 'ARRAY') {
2002 if (ref $argument eq 'ARRAY') {
2003 foreach my $argument (@{$argument}) {
2004 next if ! defined $argument;
2005 push @{$hash->{$addr}}, $argument;
2009 push @{$hash->{$addr}}, $argument if defined $argument;
2013 $hash->{$addr} = $argument;
2018 # If the file has a property for it, it means that the property is not
2019 # listed in the file's entries. So add a handler to the list of line
2020 # handlers to insert the property name into the lines, to provide a
2021 # uniform interface to the final processing subroutine.
2022 # the final code doesn't have to worry about that.
2023 if ($property{$addr}) {
2024 push @{$each_line_handler{$addr}}, \&_insert_property_into_line;
2027 if ($non_skip{$addr} && ! $debug_skip && $verbosity) {
2028 print "Warning: " . __PACKAGE__ . " constructor for $file{$addr} has useless 'non_skip' in it\n";
2031 $optional{$addr} = 1 if $skip{$addr};
2039 qw("") => "_operator_stringify",
2040 "." => \&main::_operator_dot,
2043 sub _operator_stringify {
2046 return __PACKAGE__ . " object for " . $self->file;
2049 # flag to make sure extracted files are processed early
2050 my $seen_non_extracted_non_age = 0;
2053 # Process the input object $self. This opens and closes the file and
2054 # calls all the handlers for it. Currently, this can only be called
2055 # once per file, as it destroy's the EOF handler
2058 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2060 my $addr = do { no overloading; pack 'J', $self; };
2062 my $file = $file{$addr};
2064 # Don't process if not expecting this file (because released later
2065 # than this Unicode version), and isn't there. This means if someone
2066 # copies it into an earlier version's directory, we will go ahead and
2068 return if $first_released{$addr} gt $v_version && ! -e $file;
2070 # If in debugging mode and this file doesn't have the non-skip
2071 # flag set, and isn't one of the critical files, skip it.
2073 && $first_released{$addr} ne v0
2074 && ! $non_skip{$addr})
2076 print "Skipping $file in debugging\n" if $verbosity;
2080 # File could be optional
2081 if ($optional{$addr}) {
2082 return unless -e $file;
2083 my $result = eval $optional{$addr};
2084 if (! defined $result) {
2085 Carp::my_carp_bug("Got '$@' when tried to eval $optional{$addr}. $file Skipped.");
2090 print STDERR "Skipping processing input file '$file' because '$optional{$addr}' is not true\n";
2096 if (! defined $file || ! -e $file) {
2098 # If the file doesn't exist, see if have internal data for it
2099 # (based on first_released being 0).
2100 if ($first_released{$addr} eq v0) {
2101 $handle{$addr} = 'pretend_is_open';
2104 if (! $optional{$addr} # File could be optional
2105 && $v_version ge $first_released{$addr})
2107 print STDERR "Skipping processing input file '$file' because not found\n" if $v_version ge $first_released{$addr};
2114 # Here, the file exists. Some platforms may change the case of
2116 if ($seen_non_extracted_non_age) {
2117 if ($file =~ /$EXTRACTED/i) {
2118 Carp::my_carp_bug(join_lines(<<END
2119 $file should be processed just after the 'Prop...Alias' files, and before
2120 anything not in the $EXTRACTED_DIR directory. Proceeding, but the results may
2121 have subtle problems
2126 elsif ($EXTRACTED_DIR
2127 && $first_released{$addr} ne v0
2128 && $file !~ /$EXTRACTED/i
2129 && lc($file) ne 'dage.txt')
2131 # We don't set this (by the 'if' above) if we have no
2132 # extracted directory, so if running on an early version,
2133 # this test won't work. Not worth worrying about.
2134 $seen_non_extracted_non_age = 1;
2137 # And mark the file as having being processed, and warn if it
2138 # isn't a file we are expecting. As we process the files,
2139 # they are deleted from the hash, so any that remain at the
2140 # end of the program are files that we didn't process.
2141 my $fkey = File::Spec->rel2abs($file);
2142 my $expecting = delete $potential_files{$fkey};
2143 $expecting = delete $potential_files{lc($fkey)} unless defined $expecting;
2144 Carp::my_carp("Was not expecting '$file'.") if
2146 && ! defined $handle{$addr};
2148 # Having deleted from expected files, we can quit if not to do
2149 # anything. Don't print progress unless really want verbosity
2151 print "Skipping $file.\n" if $verbosity >= $VERBOSE;
2155 # Open the file, converting the slashes used in this program
2156 # into the proper form for the OS
2158 if (not open $file_handle, "<", $file) {
2159 Carp::my_carp("Can't open $file. Skipping: $!");
2162 $handle{$addr} = $file_handle; # Cache the open file handle
2165 if ($verbosity >= $PROGRESS) {
2166 if ($progress_message{$addr}) {
2167 print "$progress_message{$addr}\n";
2170 # If using a virtual file, say so.
2171 print "Processing ", (-e $file)
2173 : "substitute $file",
2179 # Call any special handler for before the file.
2180 &{$pre_handler{$addr}}($self) if $pre_handler{$addr};
2182 # Then the main handler
2183 &{$handler{$addr}}($self);
2185 # Then any special post-file handler.
2186 &{$post_handler{$addr}}($self) if $post_handler{$addr};
2188 # If any errors have been accumulated, output the counts (as the first
2189 # error message in each class was output when it was encountered).
2190 if ($errors{$addr}) {
2193 foreach my $error (keys %{$errors{$addr}}) {
2194 $total += $errors{$addr}->{$error};
2195 delete $errors{$addr}->{$error};
2200 = "A total of $total lines had errors in $file. ";
2202 $message .= ($types == 1)
2203 ? '(Only the first one was displayed.)'
2204 : '(Only the first of each type was displayed.)';
2205 Carp::my_carp($message);
2209 if (@{$missings{$addr}}) {
2210 Carp::my_carp_bug("Handler for $file didn't look at all the \@missing lines. Generated tables likely are wrong");
2213 # If a real file handle, close it.
2214 close $handle{$addr} or Carp::my_carp("Can't close $file: $!") if
2216 $handle{$addr} = ""; # Uses empty to indicate that has already seen
2217 # the file, as opposed to undef
2222 # Sets $_ to be the next logical input line, if any. Returns non-zero
2223 # if such a line exists. 'logical' means that any lines that have
2224 # been added via insert_lines() will be returned in $_ before the file
2228 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2230 my $addr = do { no overloading; pack 'J', $self; };
2232 # Here the file is open (or if the handle is not a ref, is an open
2233 # 'virtual' file). Get the next line; any inserted lines get priority
2234 # over the file itself.
2238 while (1) { # Loop until find non-comment, non-empty line
2239 #local $to_trace = 1 if main::DEBUG;
2240 my $inserted_ref = shift @{$added_lines{$addr}};
2241 if (defined $inserted_ref) {
2242 ($adjusted, $_) = @{$inserted_ref};
2243 trace $adjusted, $_ if main::DEBUG && $to_trace;
2244 return 1 if $adjusted;
2247 last if ! ref $handle{$addr}; # Don't read unless is real file
2248 last if ! defined ($_ = readline $handle{$addr});
2251 trace $_ if main::DEBUG && $to_trace;
2253 # See if this line is the comment line that defines what property
2254 # value that code points that are not listed in the file should
2255 # have. The format or existence of these lines is not guaranteed
2256 # by Unicode since they are comments, but the documentation says
2257 # that this was added for machine-readability, so probably won't
2258 # change. This works starting in Unicode Version 5.0. They look
2261 # @missing: 0000..10FFFF; Not_Reordered
2262 # @missing: 0000..10FFFF; Decomposition_Mapping; <code point>
2263 # @missing: 0000..10FFFF; ; NaN
2265 # Save the line for a later get_missings() call.
2266 if (/$missing_defaults_prefix/) {
2267 if ($has_missings_defaults{$addr} == $NO_DEFAULTS) {
2268 $self->carp_bad_line("Unexpected \@missing line. Assuming no missing entries");
2270 elsif ($has_missings_defaults{$addr} == $NOT_IGNORED) {
2271 my @defaults = split /\s* ; \s*/x, $_;
2273 # The first field is the @missing, which ends in a
2274 # semi-colon, so can safely shift.
2277 # Some of these lines may have empty field placeholders
2278 # which get in the way. An example is:
2279 # @missing: 0000..10FFFF; ; NaN
2280 # Remove them. Process starting from the top so the
2281 # splice doesn't affect things still to be looked at.
2282 for (my $i = @defaults - 1; $i >= 0; $i--) {
2283 next if $defaults[$i] ne "";
2284 splice @defaults, $i, 1;
2287 # What's left should be just the property (maybe) and the
2288 # default. Having only one element means it doesn't have
2292 if (@defaults >= 1) {
2293 if (@defaults == 1) {
2294 $default = $defaults[0];
2297 $property = $defaults[0];
2298 $default = $defaults[1];
2304 || ($default =~ /^</
2305 && $default !~ /^<code *point>$/i
2306 && $default !~ /^<none>$/i))
2308 $self->carp_bad_line("Unrecognized \@missing line: $_. Assuming no missing entries");
2312 # If the property is missing from the line, it should
2313 # be the one for the whole file
2314 $property = $property{$addr} if ! defined $property;
2316 # Change <none> to the null string, which is what it
2317 # really means. If the default is the code point
2318 # itself, set it to <code point>, which is what
2319 # Unicode uses (but sometimes they've forgotten the
2321 if ($default =~ /^<none>$/i) {
2324 elsif ($default =~ /^<code *point>$/i) {
2325 $default = $CODE_POINT;
2328 # Store them as a sub-arrays with both components.
2329 push @{$missings{$addr}}, [ $default, $property ];
2333 # There is nothing for the caller to process on this comment
2338 # Remove comments and trailing space, and skip this line if the
2344 # Call any handlers for this line, and skip further processing of
2345 # the line if the handler sets the line to null.
2346 foreach my $sub_ref (@{$each_line_handler{$addr}}) {
2351 # Here the line is ok. return success.
2353 } # End of looping through lines.
2355 # If there is an EOF handler, call it (only once) and if it generates
2356 # more lines to process go back in the loop to handle them.
2357 if ($eof_handler{$addr}) {
2358 &{$eof_handler{$addr}}($self);
2359 $eof_handler{$addr} = ""; # Currently only get one shot at it.
2360 goto LINE if $added_lines{$addr};
2363 # Return failure -- no more lines.
2368 # Not currently used, not fully tested.
2370 # # Non-destructive look-ahead one non-adjusted, non-comment, non-blank
2371 # # record. Not callable from an each_line_handler(), nor does it call
2372 # # an each_line_handler() on the line.
2375 # my $addr = do { no overloading; pack 'J', $self; };
2377 # foreach my $inserted_ref (@{$added_lines{$addr}}) {
2378 # my ($adjusted, $line) = @{$inserted_ref};
2379 # next if $adjusted;
2381 # # Remove comments and trailing space, and return a non-empty
2384 # $line =~ s/\s+$//;
2385 # return $line if $line ne "";
2388 # return if ! ref $handle{$addr}; # Don't read unless is real file
2389 # while (1) { # Loop until find non-comment, non-empty line
2390 # local $to_trace = 1 if main::DEBUG;
2391 # trace $_ if main::DEBUG && $to_trace;
2392 # return if ! defined (my $line = readline $handle{$addr});
2394 # push @{$added_lines{$addr}}, [ 0, $line ];
2397 # $line =~ s/\s+$//;
2398 # return $line if $line ne "";
2406 # Lines can be inserted so that it looks like they were in the input
2407 # file at the place it was when this routine is called. See also
2408 # insert_adjusted_lines(). Lines inserted via this routine go through
2409 # any each_line_handler()
2413 # Each inserted line is an array, with the first element being 0 to
2414 # indicate that this line hasn't been adjusted, and needs to be
2417 push @{$added_lines{pack 'J', $self}}, map { [ 0, $_ ] } @_;
2421 sub insert_adjusted_lines {
2422 # Lines can be inserted so that it looks like they were in the input
2423 # file at the place it was when this routine is called. See also
2424 # insert_lines(). Lines inserted via this routine are already fully
2425 # adjusted, ready to be processed; each_line_handler()s handlers will
2426 # not be called. This means this is not a completely general
2427 # facility, as only the last each_line_handler on the stack should
2428 # call this. It could be made more general, by passing to each of the
2429 # line_handlers their position on the stack, which they would pass on
2430 # to this routine, and that would replace the boolean first element in
2431 # the anonymous array pushed here, so that the next_line routine could
2432 # use that to call only those handlers whose index is after it on the
2433 # stack. But this is overkill for what is needed now.
2436 trace $_[0] if main::DEBUG && $to_trace;
2438 # Each inserted line is an array, with the first element being 1 to
2439 # indicate that this line has been adjusted
2441 push @{$added_lines{pack 'J', $self}}, map { [ 1, $_ ] } @_;
2446 # Returns the stored up @missings lines' values, and clears the list.
2447 # The values are in an array, consisting of the default in the first
2448 # element, and the property in the 2nd. However, since these lines
2449 # can be stacked up, the return is an array of all these arrays.
2452 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2454 my $addr = do { no overloading; pack 'J', $self; };
2456 # If not accepting a list return, just return the first one.
2457 return shift @{$missings{$addr}} unless wantarray;
2459 my @return = @{$missings{$addr}};
2460 undef @{$missings{$addr}};
2464 sub _insert_property_into_line {
2465 # Add a property field to $_, if this file requires it.
2468 my $addr = do { no overloading; pack 'J', $self; };
2469 my $property = $property{$addr};
2470 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2472 $_ =~ s/(;|$)/; $property$1/;
2477 # Output consistent error messages, using either a generic one, or the
2478 # one given by the optional parameter. To avoid gazillions of the
2479 # same message in case the syntax of a file is way off, this routine
2480 # only outputs the first instance of each message, incrementing a
2481 # count so the totals can be output at the end of the file.
2484 my $message = shift;
2485 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2487 my $addr = do { no overloading; pack 'J', $self; };
2489 $message = 'Unexpected line' unless $message;
2491 # No trailing punctuation so as to fit with our addenda.
2492 $message =~ s/[.:;,]$//;
2494 # If haven't seen this exact message before, output it now. Otherwise
2495 # increment the count of how many times it has occurred
2496 unless ($errors{$addr}->{$message}) {
2497 Carp::my_carp("$message in '$_' in "
2499 . " at line $.. Skipping this line;");
2500 $errors{$addr}->{$message} = 1;
2503 $errors{$addr}->{$message}++;
2506 # Clear the line to prevent any further (meaningful) processing of it.
2513 package Multi_Default;
2515 # Certain properties in early versions of Unicode had more than one possible
2516 # default for code points missing from the files. In these cases, one
2517 # default applies to everything left over after all the others are applied,
2518 # and for each of the others, there is a description of which class of code
2519 # points applies to it. This object helps implement this by storing the
2520 # defaults, and for all but that final default, an eval string that generates
2521 # the class that it applies to.
2526 main::setup_package();
2529 # The defaults structure for the classes
2530 main::set_access('class_defaults', \%class_defaults);
2533 # The default that applies to everything left over.
2534 main::set_access('other_default', \%other_default, 'r');
2538 # The constructor is called with default => eval pairs, terminated by
2539 # the left-over default. e.g.
2540 # Multi_Default->new(
2541 # 'T' => '$gc->table("Mn") + $gc->table("Cf") - 0x200C
2543 # 'R' => 'some other expression that evaluates to code points',
2551 my $self = bless \do{my $anonymous_scalar}, $class;
2552 my $addr = do { no overloading; pack 'J', $self; };
2555 my $default = shift;
2557 $class_defaults{$addr}->{$default} = $eval;
2560 $other_default{$addr} = shift;
2565 sub get_next_defaults {
2566 # Iterates and returns the next class of defaults.
2568 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2570 my $addr = do { no overloading; pack 'J', $self; };
2572 return each %{$class_defaults{$addr}};
2578 # An alias is one of the names that a table goes by. This class defines them
2579 # including some attributes. Everything is currently setup in the
2585 main::setup_package();
2588 main::set_access('name', \%name, 'r');
2591 # Determined by the constructor code if this name should match loosely or
2592 # not. The constructor parameters can override this, but it isn't fully
2593 # implemented, as should have ability to override Unicode one's via
2594 # something like a set_loose_match()
2595 main::set_access('loose_match', \%loose_match, 'r');
2598 # Some aliases should not get their own entries because they are covered
2599 # by a wild-card, and some we want to discourage use of. Binary
2600 main::set_access('make_pod_entry', \%make_pod_entry, 'r');
2603 # Aliases have a status, like deprecated, or even suppressed (which means
2604 # they don't appear in documentation). Enum
2605 main::set_access('status', \%status, 'r');
2608 # Similarly, some aliases should not be considered as usable ones for
2609 # external use, such as file names, or we don't want documentation to
2610 # recommend them. Boolean
2611 main::set_access('externally_ok', \%externally_ok, 'r');
2616 my $self = bless \do { my $anonymous_scalar }, $class;
2617 my $addr = do { no overloading; pack 'J', $self; };
2619 $name{$addr} = shift;
2620 $loose_match{$addr} = shift;
2621 $make_pod_entry{$addr} = shift;
2622 $externally_ok{$addr} = shift;
2623 $status{$addr} = shift;
2625 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2627 # Null names are never ok externally
2628 $externally_ok{$addr} = 0 if $name{$addr} eq "";
2636 # A range is the basic unit for storing code points, and is described in the
2637 # comments at the beginning of the program. Each range has a starting code
2638 # point; an ending code point (not less than the starting one); a value
2639 # that applies to every code point in between the two end-points, inclusive;
2640 # and an enum type that applies to the value. The type is for the user's
2641 # convenience, and has no meaning here, except that a non-zero type is
2642 # considered to not obey the normal Unicode rules for having standard forms.
2644 # The same structure is used for both map and match tables, even though in the
2645 # latter, the value (and hence type) is irrelevant and could be used as a
2646 # comment. In map tables, the value is what all the code points in the range
2647 # map to. Type 0 values have the standardized version of the value stored as
2648 # well, so as to not have to recalculate it a lot.
2650 sub trace { return main::trace(@_); }
2654 main::setup_package();
2657 main::set_access('start', \%start, 'r', 's');
2660 main::set_access('end', \%end, 'r', 's');
2663 main::set_access('value', \%value, 'r');
2666 main::set_access('type', \%type, 'r');
2669 # The value in internal standard form. Defined only if the type is 0.
2670 main::set_access('standard_form', \%standard_form);
2672 # Note that if these fields change, the dump() method should as well
2675 return Carp::carp_too_few_args(\@_, 3) if main::DEBUG && @_ < 3;
2678 my $self = bless \do { my $anonymous_scalar }, $class;
2679 my $addr = do { no overloading; pack 'J', $self; };
2681 $start{$addr} = shift;
2682 $end{$addr} = shift;
2686 my $value = delete $args{'Value'}; # Can be 0
2687 $value = "" unless defined $value;
2688 $value{$addr} = $value;
2690 $type{$addr} = delete $args{'Type'} || 0;
2692 Carp::carp_extra_args(\%args) if main::DEBUG && %args;
2694 if (! $type{$addr}) {
2695 $standard_form{$addr} = main::standardize($value);
2703 qw("") => "_operator_stringify",
2704 "." => \&main::_operator_dot,
2707 sub _operator_stringify {
2709 my $addr = do { no overloading; pack 'J', $self; };
2711 # Output it like '0041..0065 (value)'
2712 my $return = sprintf("%04X", $start{$addr})
2714 . sprintf("%04X", $end{$addr});
2715 my $value = $value{$addr};
2716 my $type = $type{$addr};
2718 $return .= "$value";
2719 $return .= ", Type=$type" if $type != 0;
2726 # The standard form is the value itself if the standard form is
2727 # undefined (that is if the value is special)
2730 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2732 my $addr = do { no overloading; pack 'J', $self; };
2734 return $standard_form{$addr} if defined $standard_form{$addr};
2735 return $value{$addr};
2739 # Human, not machine readable. For machine readable, comment out this
2740 # entire routine and let the standard one take effect.
2743 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2745 my $addr = do { no overloading; pack 'J', $self; };
2747 my $return = $indent
2748 . sprintf("%04X", $start{$addr})
2750 . sprintf("%04X", $end{$addr})
2751 . " '$value{$addr}';";
2752 if (! defined $standard_form{$addr}) {
2753 $return .= "(type=$type{$addr})";
2755 elsif ($standard_form{$addr} ne $value{$addr}) {
2756 $return .= "(standard '$standard_form{$addr}')";
2762 package _Range_List_Base;
2764 # Base class for range lists. A range list is simply an ordered list of
2765 # ranges, so that the ranges with the lowest starting numbers are first in it.
2767 # When a new range is added that is adjacent to an existing range that has the
2768 # same value and type, it merges with it to form a larger range.
2770 # Ranges generally do not overlap, except that there can be multiple entries
2771 # of single code point ranges. This is because of NameAliases.txt.
2773 # In this program, there is a standard value such that if two different
2774 # values, have the same standard value, they are considered equivalent. This
2775 # value was chosen so that it gives correct results on Unicode data
2777 # There are a number of methods to manipulate range lists, and some operators
2778 # are overloaded to handle them.
2780 sub trace { return main::trace(@_); }
2786 main::setup_package();
2789 # The list of ranges
2790 main::set_access('ranges', \%ranges, 'readable_array');
2793 # The highest code point in the list. This was originally a method, but
2794 # actual measurements said it was used a lot.
2795 main::set_access('max', \%max, 'r');
2797 my %each_range_iterator;
2798 # Iterator position for each_range()
2799 main::set_access('each_range_iterator', \%each_range_iterator);
2802 # Name of parent this is attached to, if any. Solely for better error
2804 main::set_access('owner_name_of', \%owner_name_of, 'p_r');
2806 my %_search_ranges_cache;
2807 # A cache of the previous result from _search_ranges(), for better
2809 main::set_access('_search_ranges_cache', \%_search_ranges_cache);
2815 # Optional initialization data for the range list.
2816 my $initialize = delete $args{'Initialize'};
2820 # Use _union() to initialize. _union() returns an object of this
2821 # class, which means that it will call this constructor recursively.
2822 # But it won't have this $initialize parameter so that it won't
2823 # infinitely loop on this.
2824 return _union($class, $initialize, %args) if defined $initialize;
2826 $self = bless \do { my $anonymous_scalar }, $class;
2827 my $addr = do { no overloading; pack 'J', $self; };
2829 # Optional parent object, only for debug info.
2830 $owner_name_of{$addr} = delete $args{'Owner'};
2831 $owner_name_of{$addr} = "" if ! defined $owner_name_of{$addr};
2833 # Stringify, in case it is an object.
2834 $owner_name_of{$addr} = "$owner_name_of{$addr}";
2836 # This is used only for error messages, and so a colon is added
2837 $owner_name_of{$addr} .= ": " if $owner_name_of{$addr} ne "";
2839 Carp::carp_extra_args(\%args) if main::DEBUG && %args;
2841 # Max is initialized to a negative value that isn't adjacent to 0,
2845 $_search_ranges_cache{$addr} = 0;
2846 $ranges{$addr} = [];
2853 qw("") => "_operator_stringify",
2854 "." => \&main::_operator_dot,
2857 sub _operator_stringify {
2859 my $addr = do { no overloading; pack 'J', $self; };
2861 return "Range_List attached to '$owner_name_of{$addr}'"
2862 if $owner_name_of{$addr};
2863 return "anonymous Range_List " . \$self;
2867 # Returns the union of the input code points. It can be called as
2868 # either a constructor or a method. If called as a method, the result
2869 # will be a new() instance of the calling object, containing the union
2870 # of that object with the other parameter's code points; if called as
2871 # a constructor, the first parameter gives the class the new object
2872 # should be, and the second parameter gives the code points to go into
2874 # In either case, there are two parameters looked at by this routine;
2875 # any additional parameters are passed to the new() constructor.
2877 # The code points can come in the form of some object that contains
2878 # ranges, and has a conventionally named method to access them; or
2879 # they can be an array of individual code points (as integers); or
2880 # just a single code point.
2882 # If they are ranges, this routine doesn't make any effort to preserve
2883 # the range values of one input over the other. Therefore this base
2884 # class should not allow _union to be called from other than
2885 # initialization code, so as to prevent two tables from being added
2886 # together where the range values matter. The general form of this
2887 # routine therefore belongs in a derived class, but it was moved here
2888 # to avoid duplication of code. The failure to overload this in this
2889 # class keeps it safe.
2893 my @args; # Arguments to pass to the constructor
2897 # If a method call, will start the union with the object itself, and
2898 # the class of the new object will be the same as self.
2905 # Add the other required parameter.
2907 # Rest of parameters are passed on to the constructor
2909 # Accumulate all records from both lists.
2911 for my $arg (@args) {
2912 #local $to_trace = 0 if main::DEBUG;
2913 trace "argument = $arg" if main::DEBUG && $to_trace;
2914 if (! defined $arg) {
2916 if (defined $self) {
2918 $message .= $owner_name_of{pack 'J', $self};
2920 Carp::my_carp_bug($message .= "Undefined argument to _union. No union done.");
2923 $arg = [ $arg ] if ! ref $arg;
2924 my $type = ref $arg;
2925 if ($type eq 'ARRAY') {
2926 foreach my $element (@$arg) {
2927 push @records, Range->new($element, $element);
2930 elsif ($arg->isa('Range')) {
2931 push @records, $arg;
2933 elsif ($arg->can('ranges')) {
2934 push @records, $arg->ranges;
2938 if (defined $self) {
2940 $message .= $owner_name_of{pack 'J', $self};
2942 Carp::my_carp_bug($message . "Cannot take the union of a $type. No union done.");
2947 # Sort with the range containing the lowest ordinal first, but if
2948 # two ranges start at the same code point, sort with the bigger range
2949 # of the two first, because it takes fewer cycles.
2950 @records = sort { ($a->start <=> $b->start)
2952 # if b is shorter than a, b->end will be
2953 # less than a->end, and we want to select
2954 # a, so want to return -1
2955 ($b->end <=> $a->end)
2958 my $new = $class->new(@_);
2960 # Fold in records so long as they add new information.
2961 for my $set (@records) {
2962 my $start = $set->start;
2963 my $end = $set->end;
2964 my $value = $set->value;
2965 if ($start > $new->max) {
2966 $new->_add_delete('+', $start, $end, $value);
2968 elsif ($end > $new->max) {
2969 $new->_add_delete('+', $new->max +1, $end, $value);
2976 sub range_count { # Return the number of ranges in the range list
2978 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2981 return scalar @{$ranges{pack 'J', $self}};
2985 # Returns the minimum code point currently in the range list, or if
2986 # the range list is empty, 2 beyond the max possible. This is a
2987 # method because used so rarely, that not worth saving between calls,
2988 # and having to worry about changing it as ranges are added and
2992 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2994 my $addr = do { no overloading; pack 'J', $self; };
2996 # If the range list is empty, return a large value that isn't adjacent
2997 # to any that could be in the range list, for simpler tests
2998 return $LAST_UNICODE_CODEPOINT + 2 unless scalar @{$ranges{$addr}};
2999 return $ranges{$addr}->[0]->start;
3003 # Boolean: Is argument in the range list? If so returns $i such that:
3004 # range[$i]->end < $codepoint <= range[$i+1]->end
3005 # which is one beyond what you want; this is so that the 0th range
3006 # doesn't return false
3008 my $codepoint = shift;
3009 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3011 my $i = $self->_search_ranges($codepoint);
3012 return 0 unless defined $i;
3014 # The search returns $i, such that
3015 # range[$i-1]->end < $codepoint <= range[$i]->end
3016 # So is in the table if and only iff it is at least the start position
3019 return 0 if $ranges{pack 'J', $self}->[$i]->start > $codepoint;
3023 sub containing_range {
3024 # Returns the range object that contains the code point, undef if none
3027 my $codepoint = shift;
3028 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3030 my $i = $self->contains($codepoint);
3033 # contains() returns 1 beyond where we should look
3035 return $ranges{pack 'J', $self}->[$i-1];
3039 # Returns the value associated with the code point, undef if none
3042 my $codepoint = shift;
3043 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3045 my $range = $self->containing_range($codepoint);
3046 return unless defined $range;
3048 return $range->value;
3052 # Returns the type of the range containing the code point, undef if
3053 # the code point is not in the table
3056 my $codepoint = shift;
3057 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3059 my $range = $self->containing_range($codepoint);
3060 return unless defined $range;
3062 return $range->type;
3065 sub _search_ranges {
3066 # Find the range in the list which contains a code point, or where it
3067 # should go if were to add it. That is, it returns $i, such that:
3068 # range[$i-1]->end < $codepoint <= range[$i]->end
3069 # Returns undef if no such $i is possible (e.g. at end of table), or
3070 # if there is an error.
3073 my $code_point = shift;
3074 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3076 my $addr = do { no overloading; pack 'J', $self; };
3078 return if $code_point > $max{$addr};
3079 my $r = $ranges{$addr}; # The current list of ranges
3080 my $range_list_size = scalar @$r;
3083 use integer; # want integer division
3085 # Use the cached result as the starting guess for this one, because,
3086 # an experiment on 5.1 showed that 90% of the time the cache was the
3087 # same as the result on the next call (and 7% it was one less).
3088 $i = $_search_ranges_cache{$addr};
3089 $i = 0 if $i >= $range_list_size; # Reset if no longer valid (prob.
3090 # from an intervening deletion
3091 #local $to_trace = 1 if main::DEBUG;
3092 trace "previous \$i is still valid: $i" if main::DEBUG && $to_trace && $code_point <= $r->[$i]->end && ($i == 0 || $r->[$i-1]->end < $code_point);
3093 return $i if $code_point <= $r->[$i]->end
3094 && ($i == 0 || $r->[$i-1]->end < $code_point);
3096 # Here the cache doesn't yield the correct $i. Try adding 1.
3097 if ($i < $range_list_size - 1
3098 && $r->[$i]->end < $code_point &&
3099 $code_point <= $r->[$i+1]->end)
3102 trace "next \$i is correct: $i" if main::DEBUG && $to_trace;
3103 $_search_ranges_cache{$addr} = $i;
3107 # Here, adding 1 also didn't work. We do a binary search to
3108 # find the correct position, starting with current $i
3110 my $upper = $range_list_size - 1;
3112 trace "top of loop i=$i:", sprintf("%04X", $r->[$lower]->start), "[$lower] .. ", sprintf("%04X", $r->[$i]->start), "[$i] .. ", sprintf("%04X", $r->[$upper]->start), "[$upper]" if main::DEBUG && $to_trace;
3114 if ($code_point <= $r->[$i]->end) {
3116 # Here we have met the upper constraint. We can quit if we
3117 # also meet the lower one.
3118 last if $i == 0 || $r->[$i-1]->end < $code_point;
3120 $upper = $i; # Still too high.
3125 # Here, $r[$i]->end < $code_point, so look higher up.
3129 # Split search domain in half to try again.
3130 my $temp = ($upper + $lower) / 2;
3132 # No point in continuing unless $i changes for next time
3136 # We can't reach the highest element because of the averaging.
3137 # So if one below the upper edge, force it there and try one
3139 if ($i == $range_list_size - 2) {
3141 trace "Forcing to upper edge" if main::DEBUG && $to_trace;
3142 $i = $range_list_size - 1;
3144 # Change $lower as well so if fails next time through,
3145 # taking the average will yield the same $i, and we will
3146 # quit with the error message just below.
3150 Carp::my_carp_bug("$owner_name_of{$addr}Can't find where the range ought to go. No action taken.");
3154 } # End of while loop
3156 if (main::DEBUG && $to_trace) {
3157 trace 'i-1=[', $i-1, ']', $r->[$i-1] if $i;
3158 trace "i= [ $i ]", $r->[$i];
3159 trace 'i+1=[', $i+1, ']', $r->[$i+1] if $i < $range_list_size - 1;
3162 # Here we have found the offset. Cache it as a starting point for the
3164 $_search_ranges_cache{$addr} = $i;
3169 # Add, replace or delete ranges to or from a list. The $type
3170 # parameter gives which:
3171 # '+' => insert or replace a range, returning a list of any changed
3173 # '-' => delete a range, returning a list of any deleted ranges.
3175 # The next three parameters give respectively the start, end, and
3176 # value associated with the range. 'value' should be null unless the
3179 # The range list is kept sorted so that the range with the lowest
3180 # starting position is first in the list, and generally, adjacent
3181 # ranges with the same values are merged into a single larger one (see
3182 # exceptions below).
3184 # There are more parameters; all are key => value pairs:
3185 # Type gives the type of the value. It is only valid for '+'.
3186 # All ranges have types; if this parameter is omitted, 0 is
3187 # assumed. Ranges with type 0 are assumed to obey the
3188 # Unicode rules for casing, etc; ranges with other types are
3189 # not. Otherwise, the type is arbitrary, for the caller's
3190 # convenience, and looked at only by this routine to keep
3191 # adjacent ranges of different types from being merged into
3192 # a single larger range, and when Replace =>
3193 # $IF_NOT_EQUIVALENT is specified (see just below).
3194 # Replace determines what to do if the range list already contains
3195 # ranges which coincide with all or portions of the input
3196 # range. It is only valid for '+':
3197 # => $NO means that the new value is not to replace
3198 # any existing ones, but any empty gaps of the
3199 # range list coinciding with the input range
3200 # will be filled in with the new value.
3201 # => $UNCONDITIONALLY means to replace the existing values with
3202 # this one unconditionally. However, if the
3203 # new and old values are identical, the
3204 # replacement is skipped to save cycles
3205 # => $IF_NOT_EQUIVALENT means to replace the existing values
3206 # with this one if they are not equivalent.
3207 # Ranges are equivalent if their types are the
3208 # same, and they are the same string; or if
3209 # both are type 0 ranges, if their Unicode
3210 # standard forms are identical. In this last
3211 # case, the routine chooses the more "modern"
3212 # one to use. This is because some of the
3213 # older files are formatted with values that
3214 # are, for example, ALL CAPs, whereas the
3215 # derived files have a more modern style,
3216 # which looks better. By looking for this
3217 # style when the pre-existing and replacement
3218 # standard forms are the same, we can move to
3220 # => $MULTIPLE means that if this range duplicates an
3221 # existing one, but has a different value,
3222 # don't replace the existing one, but insert
3223 # this, one so that the same range can occur
3224 # multiple times. They are stored LIFO, so
3225 # that the final one inserted is the first one
3226 # returned in an ordered search of the table.
3227 # => anything else is the same as => $IF_NOT_EQUIVALENT
3229 # "same value" means identical for non-type-0 ranges, and it means
3230 # having the same standard forms for type-0 ranges.
3232 return Carp::carp_too_few_args(\@_, 5) if main::DEBUG && @_ < 5;
3235 my $operation = shift; # '+' for add/replace; '-' for delete;
3242 $value = "" if not defined $value; # warning: $value can be "0"
3244 my $replace = delete $args{'Replace'};
3245 $replace = $IF_NOT_EQUIVALENT unless defined $replace;
3247 my $type = delete $args{'Type'};
3248 $type = 0 unless defined $type;
3250 Carp::carp_extra_args(\%args) if main::DEBUG && %args;
3252 my $addr = do { no overloading; pack 'J', $self; };
3254 if ($operation ne '+' && $operation ne '-') {
3255 Carp::my_carp_bug("$owner_name_of{$addr}First parameter to _add_delete must be '+' or '-'. No action taken.");
3258 unless (defined $start && defined $end) {
3259 Carp::my_carp_bug("$owner_name_of{$addr}Undefined start and/or end to _add_delete. No action taken.");
3262 unless ($end >= $start) {
3263 Carp::my_carp_bug("$owner_name_of{$addr}End of range (" . sprintf("%04X", $end) . ") must not be before start (" . sprintf("%04X", $start) . "). No action taken.");
3266 #local $to_trace = 1 if main::DEBUG;
3268 if ($operation eq '-') {
3269 if ($replace != $IF_NOT_EQUIVALENT) {
3270 Carp::my_carp_bug("$owner_name_of{$addr}Replace => \$IF_NOT_EQUIVALENT is required when deleting a range from a range list. Assuming Replace => \$IF_NOT_EQUIVALENT.");
3271 $replace = $IF_NOT_EQUIVALENT;
3274 Carp::my_carp_bug("$owner_name_of{$addr}Type => 0 is required when deleting a range from a range list. Assuming Type => 0.");
3278 Carp::my_carp_bug("$owner_name_of{$addr}Value => \"\" is required when deleting a range from a range list. Assuming Value => \"\".");
3283 my $r = $ranges{$addr}; # The current list of ranges
3284 my $range_list_size = scalar @$r; # And its size
3285 my $max = $max{$addr}; # The current high code point in
3286 # the list of ranges
3288 # Do a special case requiring fewer machine cycles when the new range
3289 # starts after the current highest point. The Unicode input data is
3290 # structured so this is common.
3291 if ($start > $max) {
3293 trace "$owner_name_of{$addr} $operation", sprintf("%04X", $start) . '..' . sprintf("%04X", $end) . " ($value) type=$type" if main::DEBUG && $to_trace;
3294 return if $operation eq '-'; # Deleting a non-existing range is a
3297 # If the new range doesn't logically extend the current final one
3298 # in the range list, create a new range at the end of the range
3299 # list. (max cleverly is initialized to a negative number not
3300 # adjacent to 0 if the range list is empty, so even adding a range
3301 # to an empty range list starting at 0 will have this 'if'
3303 if ($start > $max + 1 # non-adjacent means can't extend.
3304 || @{$r}[-1]->value ne $value # values differ, can't extend.
3305 || @{$r}[-1]->type != $type # types differ, can't extend.
3307 push @$r, Range->new($start, $end,
3313 # Here, the new range starts just after the current highest in
3314 # the range list, and they have the same type and value.
3315 # Extend the current range to incorporate the new one.
3316 @{$r}[-1]->set_end($end);
3319 # This becomes the new maximum.
3324 #local $to_trace = 0 if main::DEBUG;
3326 trace "$owner_name_of{$addr} $operation", sprintf("%04X", $start) . '..' . sprintf("%04X", $end) . " ($value) replace=$replace" if main::DEBUG && $to_trace;
3328 # Here, the input range isn't after the whole rest of the range list.
3329 # Most likely 'splice' will be needed. The rest of the routine finds
3330 # the needed splice parameters, and if necessary, does the splice.
3331 # First, find the offset parameter needed by the splice function for
3332 # the input range. Note that the input range may span multiple
3333 # existing ones, but we'll worry about that later. For now, just find
3334 # the beginning. If the input range is to be inserted starting in a
3335 # position not currently in the range list, it must (obviously) come
3336 # just after the range below it, and just before the range above it.
3337 # Slightly less obviously, it will occupy the position currently
3338 # occupied by the range that is to come after it. More formally, we
3339 # are looking for the position, $i, in the array of ranges, such that:
3341 # r[$i-1]->start <= r[$i-1]->end < $start < r[$i]->start <= r[$i]->end
3343 # (The ordered relationships within existing ranges are also shown in
3344 # the equation above). However, if the start of the input range is
3345 # within an existing range, the splice offset should point to that
3346 # existing range's position in the list; that is $i satisfies a
3347 # somewhat different equation, namely:
3349 #r[$i-1]->start <= r[$i-1]->end < r[$i]->start <= $start <= r[$i]->end
3351 # More briefly, $start can come before or after r[$i]->start, and at
3352 # this point, we don't know which it will be. However, these
3353 # two equations share these constraints:
3355 # r[$i-1]->end < $start <= r[$i]->end
3357 # And that is good enough to find $i.
3359 my $i = $self->_search_ranges($start);
3361 Carp::my_carp_bug("Searching $self for range beginning with $start unexpectedly returned undefined. Operation '$operation' not performed");
3365 # The search function returns $i such that:
3367 # r[$i-1]->end < $start <= r[$i]->end
3369 # That means that $i points to the first range in the range list
3370 # that could possibly be affected by this operation. We still don't
3371 # know if the start of the input range is within r[$i], or if it
3372 # points to empty space between r[$i-1] and r[$i].
3373 trace "[$i] is the beginning splice point. Existing range there is ", $r->[$i] if main::DEBUG && $to_trace;
3375 # Special case the insertion of data that is not to replace any
3377 if ($replace == $NO) { # If $NO, has to be operation '+'
3378 #local $to_trace = 1 if main::DEBUG;
3379 trace "Doesn't replace" if main::DEBUG && $to_trace;
3381 # Here, the new range is to take effect only on those code points
3382 # that aren't already in an existing range. This can be done by
3383 # looking through the existing range list and finding the gaps in
3384 # the ranges that this new range affects, and then calling this
3385 # function recursively on each of those gaps, leaving untouched
3386 # anything already in the list. Gather up a list of the changed
3387 # gaps first so that changes to the internal state as new ranges
3388 # are added won't be a problem.
3391 # First, if the starting point of the input range is outside an
3392 # existing one, there is a gap from there to the beginning of the
3393 # existing range -- add a span to fill the part that this new
3395 if ($start < $r->[$i]->start) {
3396 push @gap_list, Range->new($start,
3398 $r->[$i]->start - 1),
3400 trace "gap before $r->[$i] [$i], will add", $gap_list[-1] if main::DEBUG && $to_trace;
3403 # Then look through the range list for other gaps until we reach
3404 # the highest range affected by the input one.
3406 for ($j = $i+1; $j < $range_list_size; $j++) {
3407 trace "j=[$j]", $r->[$j] if main::DEBUG && $to_trace;
3408 last if $end < $r->[$j]->start;
3410 # If there is a gap between when this range starts and the
3411 # previous one ends, add a span to fill it. Note that just
3412 # because there are two ranges doesn't mean there is a
3413 # non-zero gap between them. It could be that they have
3414 # different values or types
3415 if ($r->[$j-1]->end + 1 != $r->[$j]->start) {
3417 Range->new($r->[$j-1]->end + 1,
3418 $r->[$j]->start - 1,
3420 trace "gap between $r->[$j-1] and $r->[$j] [$j], will add: $gap_list[-1]" if main::DEBUG && $to_trace;
3424 # Here, we have either found an existing range in the range list,
3425 # beyond the area affected by the input one, or we fell off the
3426 # end of the loop because the input range affects the whole rest
3427 # of the range list. In either case, $j is 1 higher than the
3428 # highest affected range. If $j == $i, it means that there are no
3429 # affected ranges, that the entire insertion is in the gap between
3430 # r[$i-1], and r[$i], which we already have taken care of before
3432 # On the other hand, if there are affected ranges, it might be
3433 # that there is a gap that needs filling after the final such
3434 # range to the end of the input range
3435 if ($r->[$j-1]->end < $end) {
3436 push @gap_list, Range->new(main::max($start,
3437 $r->[$j-1]->end + 1),
3440 trace "gap after $r->[$j-1], will add $gap_list[-1]" if main::DEBUG && $to_trace;
3443 # Call recursively to fill in all the gaps.
3444 foreach my $gap (@gap_list) {
3445 $self->_add_delete($operation,
3455 # Here, we have taken care of the case where $replace is $NO.
3456 # Remember that here, r[$i-1]->end < $start <= r[$i]->end
3457 # If inserting a multiple record, this is where it goes, before the
3458 # first (if any) existing one. This implies an insertion, and no
3459 # change to any existing ranges. Note that $i can be -1 if this new
3460 # range doesn't actually duplicate any existing, and comes at the
3461 # beginning of the list.
3462 if ($replace == $MULTIPLE) {
3464 if ($start != $end) {
3465 Carp::my_carp_bug("$owner_name_of{$addr}Can't cope with adding a multiple record when the range ($start..$end) contains more than one code point. No action taken.");
3469 # Don't add an exact duplicate, as it isn't really a multiple
3470 if ($end >= $r->[$i]->start) {
3471 if ($r->[$i]->start != $r->[$i]->end) {
3472 Carp::my_carp_bug("$owner_name_of{$addr}Can't cope with adding a multiple record when the other range ($r->[$i]) contains more than one code point. No action taken.");
3475 return if $value eq $r->[$i]->value && $type eq $r->[$i]->type;
3478 trace "Adding multiple record at $i with $start..$end, $value" if main::DEBUG && $to_trace;
3479 my @return = splice @$r,
3486 if (main::DEBUG && $to_trace) {
3487 trace "After splice:";
3488 trace 'i-2=[', $i-2, ']', $r->[$i-2] if $i >= 2;
3489 trace 'i-1=[', $i-1, ']', $r->[$i-1] if $i >= 1;
3490 trace "i =[", $i, "]", $r->[$i] if $i >= 0;
3491 trace 'i+1=[', $i+1, ']', $r->[$i+1] if $i < @$r - 1;
3492 trace 'i+2=[', $i+2, ']', $r->[$i+2] if $i < @$r - 2;
3493 trace 'i+3=[', $i+3, ']', $r->[$i+3] if $i < @$r - 3;
3498 # Here, we have taken care of $NO and $MULTIPLE replaces. This leaves
3499 # delete, insert, and replace either unconditionally or if not
3500 # equivalent. $i still points to the first potential affected range.
3501 # Now find the highest range affected, which will determine the length
3502 # parameter to splice. (The input range can span multiple existing
3503 # ones.) If this isn't a deletion, while we are looking through the
3504 # range list, see also if this is a replacement rather than a clean
3505 # insertion; that is if it will change the values of at least one
3506 # existing range. Start off assuming it is an insert, until find it
3508 my $clean_insert = $operation eq '+';
3509 my $j; # This will point to the highest affected range
3511 # For non-zero types, the standard form is the value itself;
3512 my $standard_form = ($type) ? $value : main::standardize($value);
3514 for ($j = $i; $j < $range_list_size; $j++) {
3515 trace "Looking for highest affected range; the one at $j is ", $r->[$j] if main::DEBUG && $to_trace;
3517 # If find a range that it doesn't overlap into, we can stop
3519 last if $end < $r->[$j]->start;
3521 # Here, overlaps the range at $j. If the values don't match,
3522 # and so far we think this is a clean insertion, it becomes a
3523 # non-clean insertion, i.e., a 'change' or 'replace' instead.
3524 if ($clean_insert) {
3525 if ($r->[$j]->standard_form ne $standard_form) {
3527 if ($replace == $CROAK) {
3528 main::croak("The range to add "
3529 . sprintf("%04X", $start)
3531 . sprintf("%04X", $end)
3532 . " with value '$value' overlaps an existing range $r->[$j]");
3537 # Here, the two values are essentially the same. If the
3538 # two are actually identical, replacing wouldn't change
3539 # anything so skip it.
3540 my $pre_existing = $r->[$j]->value;
3541 if ($pre_existing ne $value) {
3543 # Here the new and old standardized values are the
3544 # same, but the non-standardized values aren't. If
3545 # replacing unconditionally, then replace
3546 if( $replace == $UNCONDITIONALLY) {
3551 # Here, are replacing conditionally. Decide to
3552 # replace or not based on which appears to look
3553 # the "nicest". If one is mixed case and the
3554 # other isn't, choose the mixed case one.
3555 my $new_mixed = $value =~ /[A-Z]/
3556 && $value =~ /[a-z]/;
3557 my $old_mixed = $pre_existing =~ /[A-Z]/
3558 && $pre_existing =~ /[a-z]/;
3560 if ($old_mixed != $new_mixed) {
3561 $clean_insert = 0 if $new_mixed;
3562 if (main::DEBUG && $to_trace) {
3563 if ($clean_insert) {
3564 trace "Retaining $pre_existing over $value";
3567 trace "Replacing $pre_existing with $value";
3573 # Here casing wasn't different between the two.
3574 # If one has hyphens or underscores and the
3575 # other doesn't, choose the one with the
3577 my $new_punct = $value =~ /[-_]/;
3578 my $old_punct = $pre_existing =~ /[-_]/;
3580 if ($old_punct != $new_punct) {
3581 $clean_insert = 0 if $new_punct;
3582 if (main::DEBUG && $to_trace) {
3583 if ($clean_insert) {
3584 trace "Retaining $pre_existing over $value";
3587 trace "Replacing $pre_existing with $value";
3590 } # else existing one is just as "good";
3591 # retain it to save cycles.
3597 } # End of loop looking for highest affected range.
3599 # Here, $j points to one beyond the highest range that this insertion
3600 # affects (hence to beyond the range list if that range is the final
3601 # one in the range list).
3603 # The splice length is all the affected ranges. Get it before
3604 # subtracting, for efficiency, so we don't have to later add 1.
3605 my $length = $j - $i;
3607 $j--; # $j now points to the highest affected range.
3608 trace "Final affected range is $j: $r->[$j]" if main::DEBUG && $to_trace;
3610 # Here, have taken care of $NO and $MULTIPLE replaces.
3611 # $j points to the highest affected range. But it can be < $i or even
3612 # -1. These happen only if the insertion is entirely in the gap
3613 # between r[$i-1] and r[$i]. Here's why: j < i means that the j loop
3614 # above exited first time through with $end < $r->[$i]->start. (And
3615 # then we subtracted one from j) This implies also that $start <
3616 # $r->[$i]->start, but we know from above that $r->[$i-1]->end <
3617 # $start, so the entire input range is in the gap.
3620 # Here the entire input range is in the gap before $i.
3622 if (main::DEBUG && $to_trace) {
3624 trace "Entire range is between $r->[$i-1] and $r->[$i]";
3627 trace "Entire range is before $r->[$i]";
3630 return if $operation ne '+'; # Deletion of a non-existent range is
3635 # Here part of the input range is not in the gap before $i. Thus,
3636 # there is at least one affected one, and $j points to the highest
3639 # At this point, here is the situation:
3640 # This is not an insertion of a multiple, nor of tentative ($NO)
3642 # $i points to the first element in the current range list that
3643 # may be affected by this operation. In fact, we know
3644 # that the range at $i is affected because we are in
3645 # the else branch of this 'if'
3646 # $j points to the highest affected range.
3648 # r[$i-1]->end < $start <= r[$i]->end
3650 # r[$i-1]->end < $start <= $end <= r[$j]->end
3653 # $clean_insert is a boolean which is set true if and only if
3654 # this is a "clean insertion", i.e., not a change nor a
3655 # deletion (multiple was handled above).
3657 # We now have enough information to decide if this call is a no-op
3658 # or not. It is a no-op if this is an insertion of already
3661 if (main::DEBUG && $to_trace && $clean_insert
3663 && $start >= $r->[$i]->start)
3667 return if $clean_insert
3668 && $i == $j # more than one affected range => not no-op
3670 # Here, r[$i-1]->end < $start <= $end <= r[$i]->end
3671 # Further, $start and/or $end is >= r[$i]->start
3672 # The test below hence guarantees that
3673 # r[$i]->start < $start <= $end <= r[$i]->end
3674 # This means the input range is contained entirely in
3675 # the one at $i, so is a no-op
3676 && $start >= $r->[$i]->start;
3679 # Here, we know that some action will have to be taken. We have
3680 # calculated the offset and length (though adjustments may be needed)
3681 # for the splice. Now start constructing the replacement list.
3683 my $splice_start = $i;
3688 # See if should extend any adjacent ranges.
3689 if ($operation eq '-') { # Don't extend deletions
3690 $extends_below = $extends_above = 0;
3692 else { # Here, should extend any adjacent ranges. See if there are
3694 $extends_below = ($i > 0
3695 # can't extend unless adjacent
3696 && $r->[$i-1]->end == $start -1
3697 # can't extend unless are same standard value
3698 && $r->[$i-1]->standard_form eq $standard_form
3699 # can't extend unless share type
3700 && $r->[$i-1]->type == $type);
3701 $extends_above = ($j+1 < $range_list_size
3702 && $r->[$j+1]->start == $end +1
3703 && $r->[$j+1]->standard_form eq $standard_form
3704 && $r->[$j+1]->type == $type);
3706 if ($extends_below && $extends_above) { # Adds to both
3707 $splice_start--; # start replace at element below
3708 $length += 2; # will replace on both sides
3709 trace "Extends both below and above ranges" if main::DEBUG && $to_trace;
3711 # The result will fill in any gap, replacing both sides, and
3712 # create one large range.
3713 @replacement = Range->new($r->[$i-1]->start,
3720 # Here we know that the result won't just be the conglomeration of
3721 # a new range with both its adjacent neighbors. But it could
3722 # extend one of them.
3724 if ($extends_below) {
3726 # Here the new element adds to the one below, but not to the
3727 # one above. If inserting, and only to that one range, can
3728 # just change its ending to include the new one.
3729 if ($length == 0 && $clean_insert) {
3730 $r->[$i-1]->set_end($end);
3731 trace "inserted range extends range to below so it is now $r->[$i-1]" if main::DEBUG && $to_trace;
3735 trace "Changing inserted range to start at ", sprintf("%04X", $r->[$i-1]->start), " instead of ", sprintf("%04X", $start) if main::DEBUG && $to_trace;
3736 $splice_start--; # start replace at element below
3737 $length++; # will replace the element below
3738 $start = $r->[$i-1]->start;
3741 elsif ($extends_above) {
3743 # Here the new element adds to the one above, but not below.
3744 # Mirror the code above
3745 if ($length == 0 && $clean_insert) {
3746 $r->[$j+1]->set_start($start);
3747 trace "inserted range extends range to above so it is now $r->[$j+1]" if main::DEBUG && $to_trace;
3751 trace "Changing inserted range to end at ", sprintf("%04X", $r->[$j+1]->end), " instead of ", sprintf("%04X", $end) if main::DEBUG && $to_trace;
3752 $length++; # will replace the element above
3753 $end = $r->[$j+1]->end;
3757 trace "Range at $i is $r->[$i]" if main::DEBUG && $to_trace;
3759 # Finally, here we know there will have to be a splice.
3760 # If the change or delete affects only the highest portion of the
3761 # first affected range, the range will have to be split. The
3762 # splice will remove the whole range, but will replace it by a new
3763 # range containing just the unaffected part. So, in this case,
3764 # add to the replacement list just this unaffected portion.
3765 if (! $extends_below
3766 && $start > $r->[$i]->start && $start <= $r->[$i]->end)
3769 Range->new($r->[$i]->start,
3771 Value => $r->[$i]->value,
3772 Type => $r->[$i]->type);
3775 # In the case of an insert or change, but not a delete, we have to
3776 # put in the new stuff; this comes next.
3777 if ($operation eq '+') {
3778 push @replacement, Range->new($start,
3784 trace "Range at $j is $r->[$j]" if main::DEBUG && $to_trace && $j != $i;
3785 #trace "$end >=", $r->[$j]->start, " && $end <", $r->[$j]->end if main::DEBUG && $to_trace;
3787 # And finally, if we're changing or deleting only a portion of the
3788 # highest affected range, it must be split, as the lowest one was.
3789 if (! $extends_above
3790 && $j >= 0 # Remember that j can be -1 if before first
3792 && $end >= $r->[$j]->start
3793 && $end < $r->[$j]->end)
3796 Range->new($end + 1,
3798 Value => $r->[$j]->value,
3799 Type => $r->[$j]->type);
3803 # And do the splice, as calculated above
3804 if (main::DEBUG && $to_trace) {
3805 trace "replacing $length element(s) at $i with ";
3806 foreach my $replacement (@replacement) {
3807 trace " $replacement";
3809 trace "Before splice:";
3810 trace 'i-2=[', $i-2, ']', $r->[$i-2] if $i >= 2;
3811 trace 'i-1=[', $i-1, ']', $r->[$i-1] if $i >= 1;
3812 trace "i =[", $i, "]", $r->[$i];
3813 trace 'i+1=[', $i+1, ']', $r->[$i+1] if $i < @$r - 1;
3814 trace 'i+2=[', $i+2, ']', $r->[$i+2] if $i < @$r - 2;
3817 my @return = splice @$r, $splice_start, $length, @replacement;
3819 if (main::DEBUG && $to_trace) {
3820 trace "After splice:";
3821 trace 'i-2=[', $i-2, ']', $r->[$i-2] if $i >= 2;
3822 trace 'i-1=[', $i-1, ']', $r->[$i-1] if $i >= 1;
3823 trace "i =[", $i, "]", $r->[$i];
3824 trace 'i+1=[', $i+1, ']', $r->[$i+1] if $i < @$r - 1;
3825 trace 'i+2=[', $i+2, ']', $r->[$i+2] if $i < @$r - 2;
3826 trace "removed ", @return if @return;
3829 # An actual deletion could have changed the maximum in the list.
3830 # There was no deletion if the splice didn't return something, but
3831 # otherwise recalculate it. This is done too rarely to worry about
3833 if ($operation eq '-' && @return) {
3834 $max{$addr} = $r->[-1]->end;
3839 sub reset_each_range { # reset the iterator for each_range();
3841 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3844 undef $each_range_iterator{pack 'J', $self};
3849 # Iterate over each range in a range list. Results are undefined if
3850 # the range list is changed during the iteration.
3853 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3855 my $addr = do { no overloading; pack 'J', $self; };
3857 return if $self->is_empty;
3859 $each_range_iterator{$addr} = -1
3860 if ! defined $each_range_iterator{$addr};
3861 $each_range_iterator{$addr}++;
3862 return $ranges{$addr}->[$each_range_iterator{$addr}]
3863 if $each_range_iterator{$addr} < @{$ranges{$addr}};
3864 undef $each_range_iterator{$addr};
3868 sub count { # Returns count of code points in range list
3870 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3872 my $addr = do { no overloading; pack 'J', $self; };
3875 foreach my $range (@{$ranges{$addr}}) {
3876 $count += $range->end - $range->start + 1;
3881 sub delete_range { # Delete a range
3886 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3888 return $self->_add_delete('-', $start, $end, "");
3891 sub is_empty { # Returns boolean as to if a range list is empty
3893 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3896 return scalar @{$ranges{pack 'J', $self}} == 0;
3900 # Quickly returns a scalar suitable for separating tables into
3901 # buckets, i.e. it is a hash function of the contents of a table, so
3902 # there are relatively few conflicts.
3905 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3907 my $addr = do { no overloading; pack 'J', $self; };
3909 # These are quickly computable. Return looks like 'min..max;count'
3910 return $self->min . "..$max{$addr};" . scalar @{$ranges{$addr}};
3912 } # End closure for _Range_List_Base
3915 use base '_Range_List_Base';
3917 # A Range_List is a range list for match tables; i.e. the range values are
3918 # not significant. Thus a number of operations can be safely added to it,
3919 # such as inversion, intersection. Note that union is also an unsafe
3920 # operation when range values are cared about, and that method is in the base
3921 # class, not here. But things are set up so that that method is callable only
3922 # during initialization. Only in this derived class, is there an operation
3923 # that combines two tables. A Range_Map can thus be used to initialize a
3924 # Range_List, and its mappings will be in the list, but are not significant to
3927 sub trace { return main::trace(@_); }
3933 '+' => sub { my $self = shift;
3936 return $self->_union($other)
3938 '&' => sub { my $self = shift;
3941 return $self->_intersect($other, 0);
3948 # Returns a new Range_List that gives all code points not in $self.
3952 my $new = Range_List->new;
3954 # Go through each range in the table, finding the gaps between them
3955 my $max = -1; # Set so no gap before range beginning at 0
3956 for my $range ($self->ranges) {
3957 my $start = $range->start;
3958 my $end = $range->end;
3960 # If there is a gap before this range, the inverse will contain
3962 if ($start > $max + 1) {
3963 $new->add_range($max + 1, $start - 1);
3968 # And finally, add the gap from the end of the table to the max
3969 # possible code point
3970 if ($max < $LAST_UNICODE_CODEPOINT) {
3971 $new->add_range($max + 1, $LAST_UNICODE_CODEPOINT);
3977 # Returns a new Range_List with the argument deleted from it. The
3978 # argument can be a single code point, a range, or something that has
3979 # a range, with the _range_list() method on it returning them
3983 my $reversed = shift;
3984 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3987 Carp::my_carp_bug("Can't cope with a "
3989 . " being the second parameter in a '-'. Subtraction ignored.");
3993 my $new = Range_List->new(Initialize => $self);
3995 if (! ref $other) { # Single code point
3996 $new->delete_range($other, $other);
3998 elsif ($other->isa('Range')) {
3999 $new->delete_range($other->start, $other->end);