This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
mktables: Use Input_file class for always skipped files
[perl5.git] / lib / unicore / mktables
CommitLineData
d73e5302 1#!/usr/bin/perl -w
99870f4d
KW
2
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.
6
9be22ca5
KW
7# There was an attempt when this was first rewritten to make it 5.8
8# compatible, but that has now been abandoned, and newer constructs are used
9# as convenient.
6c68572b 10
cdcef19a 11my $start_time;
98dc9551 12BEGIN { # Get the time the script started running; do it at compilation to
cdcef19a
KW
13 # get it as close as possible
14 $start_time= time;
15}
16
23e33b60 17require 5.010_001;
d73e5302 18use strict;
99870f4d 19use warnings;
cf25bb62 20use Carp;
bd9ebcfd 21use Config;
99870f4d
KW
22use File::Find;
23use File::Path;
d07a55ed 24use File::Spec;
99870f4d 25use Text::Tabs;
6b64c11c 26use re "/aa";
0dee21d6 27use feature 'state';
99870f4d
KW
28
29sub DEBUG () { 0 } # Set to 0 for production; 1 for development
bd9ebcfd 30my $debugging_build = $Config{"ccflags"} =~ /-DDEBUGGING/;
99870f4d 31
74cd47d0
KW
32sub NON_ASCII_PLATFORM { ord("A") != 65 }
33
99870f4d
KW
34##########################################################################
35#
36# mktables -- create the runtime Perl Unicode files (lib/unicore/.../*.pl),
37# from the Unicode database files (lib/unicore/.../*.txt), It also generates
232ed87f 38# a pod file and .t files, depending on option parameters.
99870f4d
KW
39#
40# The structure of this file is:
41# First these introductory comments; then
42# code needed for everywhere, such as debugging stuff; then
43# code to handle input parameters; then
44# data structures likely to be of external interest (some of which depend on
45# the input parameters, so follows them; then
46# more data structures and subroutine and package (class) definitions; then
47# the small actual loop to process the input files and finish up; then
48# a __DATA__ section, for the .t tests
49#
232ed87f
KW
50# This program works on all releases of Unicode so far. The outputs have been
51# scrutinized most intently for release 5.1. The others have been checked for
52# somewhat more than just sanity. It can handle all non-provisional Unicode
53# character properties in those releases.
99870f4d 54#
99870f4d
KW
55# This program is mostly about Unicode character (or code point) properties.
56# A property describes some attribute or quality of a code point, like if it
57# is lowercase or not, its name, what version of Unicode it was first defined
58# in, or what its uppercase equivalent is. Unicode deals with these disparate
59# possibilities by making all properties into mappings from each code point
60# into some corresponding value. In the case of it being lowercase or not,
61# the mapping is either to 'Y' or 'N' (or various synonyms thereof). Each
62# property maps each Unicode code point to a single value, called a "property
232ed87f
KW
63# value". (Some more recently defined properties, map a code point to a set
64# of values.)
99870f4d
KW
65#
66# When using a property in a regular expression, what is desired isn't the
67# mapping of the code point to its property's value, but the reverse (or the
68# mathematical "inverse relation"): starting with the property value, "Does a
69# code point map to it?" These are written in a "compound" form:
70# \p{property=value}, e.g., \p{category=punctuation}. This program generates
71# files containing the lists of code points that map to each such regular
72# expression property value, one file per list
73#
74# There is also a single form shortcut that Perl adds for many of the commonly
75# used properties. This happens for all binary properties, plus script,
76# general_category, and block properties.
77#
78# Thus the outputs of this program are files. There are map files, mostly in
79# the 'To' directory; and there are list files for use in regular expression
80# matching, all in subdirectories of the 'lib' directory, with each
81# subdirectory being named for the property that the lists in it are for.
82# Bookkeeping, test, and documentation files are also generated.
83
84my $matches_directory = 'lib'; # Where match (\p{}) files go.
85my $map_directory = 'To'; # Where map files go.
86
87# DATA STRUCTURES
88#
89# The major data structures of this program are Property, of course, but also
90# Table. There are two kinds of tables, very similar to each other.
91# "Match_Table" is the data structure giving the list of code points that have
92# a particular property value, mentioned above. There is also a "Map_Table"
93# data structure which gives the property's mapping from code point to value.
94# There are two structures because the match tables need to be combined in
95# various ways, such as constructing unions, intersections, complements, etc.,
96# and the map ones don't. And there would be problems, perhaps subtle, if
97# a map table were inadvertently operated on in some of those ways.
98# The use of separate classes with operations defined on one but not the other
99# prevents accidentally confusing the two.
100#
101# At the heart of each table's data structure is a "Range_List", which is just
102# an ordered list of "Ranges", plus ancillary information, and methods to
103# operate on them. A Range is a compact way to store property information.
104# Each range has a starting code point, an ending code point, and a value that
105# is meant to apply to all the code points between the two end points,
106# inclusive. For a map table, this value is the property value for those
107# code points. Two such ranges could be written like this:
108# 0x41 .. 0x5A, 'Upper',
109# 0x61 .. 0x7A, 'Lower'
110#
111# Each range also has a type used as a convenience to classify the values.
112# Most ranges in this program will be Type 0, or normal, but there are some
113# ranges that have a non-zero type. These are used only in map tables, and
114# are for mappings that don't fit into the normal scheme of things. Mappings
115# that require a hash entry to communicate with utf8.c are one example;
116# another example is mappings for charnames.pm to use which indicate a name
232ed87f 117# that is algorithmically determinable from its code point (and the reverse).
99870f4d
KW
118# These are used to significantly compact these tables, instead of listing
119# each one of the tens of thousands individually.
120#
121# In a match table, the value of a range is irrelevant (and hence the type as
122# well, which will always be 0), and arbitrarily set to the null string.
123# Using the example above, there would be two match tables for those two
124# entries, one named Upper would contain the 0x41..0x5A range, and the other
125# named Lower would contain 0x61..0x7A.
126#
127# Actually, there are two types of range lists, "Range_Map" is the one
128# associated with map tables, and "Range_List" with match tables.
232ed87f
KW
129# Again, this is so that methods can be defined on one and not the others so
130# as to prevent operating on them in incorrect ways.
99870f4d
KW
131#
132# Eventually, most tables are written out to files to be read by utf8_heavy.pl
133# in the perl core. All tables could in theory be written, but some are
134# suppressed because there is no current practical use for them. It is easy
135# to change which get written by changing various lists that are near the top
136# of the actual code in this file. The table data structures contain enough
137# ancillary information to allow them to be treated as separate entities for
138# writing, such as the path to each one's file. There is a heading in each
139# map table that gives the format of its entries, and what the map is for all
140# the code points missing from it. (This allows tables to be more compact.)
678f13d5 141#
99870f4d
KW
142# The Property data structure contains one or more tables. All properties
143# contain a map table (except the $perl property which is a
144# pseudo-property containing only match tables), and any properties that
145# are usable in regular expression matches also contain various matching
146# tables, one for each value the property can have. A binary property can
147# have two values, True and False (or Y and N, which are preferred by Unicode
148# terminology). Thus each of these properties will have a map table that
149# takes every code point and maps it to Y or N (but having ranges cuts the
150# number of entries in that table way down), and two match tables, one
151# which has a list of all the code points that map to Y, and one for all the
232ed87f 152# code points that map to N. (For each binary property, a third table is also
99870f4d 153# generated for the pseudo Perl property. It contains the identical code
232ed87f
KW
154# points as the Y table, but can be written in regular expressions, not in the
155# compound form, but in a "single" form like \p{IsUppercase}.) Many
156# properties are binary, but some properties have several possible values,
157# some have many, and properties like Name have a different value for every
158# named code point. Those will not, unless the controlling lists are changed,
159# have their match tables written out. But all the ones which can be used in
160# regular expression \p{} and \P{} constructs will. Prior to 5.14, generally
161# a property would have either its map table or its match tables written but
162# not both. Again, what gets written is controlled by lists which can easily
163# be changed. Starting in 5.14, advantage was taken of this, and all the map
164# tables needed to reconstruct the Unicode db are now written out, while
165# suppressing the Unicode .txt files that contain the data. Our tables are
166# much more compact than the .txt files, so a significant space savings was
167# achieved. Also, tables are not written out that are trivially derivable
168# from tables that do get written. So, there typically is no file containing
169# the code points not matched by a binary property (the table for \P{} versus
170# lowercase \p{}), since you just need to invert the True table to get the
171# False table.
172
173# Properties have a 'Type', like 'binary', or 'string', or 'enum' depending on
174# how many match tables there are and the content of the maps. This 'Type' is
c12f2655
KW
175# different than a range 'Type', so don't get confused by the two concepts
176# having the same name.
678f13d5 177#
99870f4d
KW
178# For information about the Unicode properties, see Unicode's UAX44 document:
179
180my $unicode_reference_url = 'http://www.unicode.org/reports/tr44/';
181
182# As stated earlier, this program will work on any release of Unicode so far.
183# Most obvious problems in earlier data have NOT been corrected except when
be864b6c 184# necessary to make Perl or this program work reasonably, and to keep out
232ed87f
KW
185# potential security issues. For example, no folding information was given in
186# early releases, so this program substitutes lower case instead, just so that
187# a regular expression with the /i option will do something that actually
188# gives the right results in many cases. There are also a couple other
189# corrections for version 1.1.5, commented at the point they are made. As an
190# example of corrections that weren't made (but could be) is this statement
191# from DerivedAge.txt: "The supplementary private use code points and the
192# non-character code points were assigned in version 2.0, but not specifically
193# listed in the UCD until versions 3.0 and 3.1 respectively." (To be precise
194# it was 3.0.1 not 3.0.0) More information on Unicode version glitches is
195# further down in these introductory comments.
99870f4d 196#
232ed87f
KW
197# This program works on all non-provisional properties as of the current
198# Unicode release, though the files for some are suppressed for various
199# reasons. You can change which are output by changing lists in this program.
678f13d5 200#
dc85bd38 201# The old version of mktables emphasized the term "Fuzzy" to mean Unicode's
99870f4d
KW
202# loose matchings rules (from Unicode TR18):
203#
204# The recommended names for UCD properties and property values are in
205# PropertyAliases.txt [Prop] and PropertyValueAliases.txt
206# [PropValue]. There are both abbreviated names and longer, more
207# descriptive names. It is strongly recommended that both names be
208# recognized, and that loose matching of property names be used,
209# whereby the case distinctions, whitespace, hyphens, and underbar
210# are ignored.
232ed87f 211#
99870f4d
KW
212# The program still allows Fuzzy to override its determination of if loose
213# matching should be used, but it isn't currently used, as it is no longer
214# needed; the calculations it makes are good enough.
678f13d5 215#
99870f4d
KW
216# SUMMARY OF HOW IT WORKS:
217#
218# Process arguments
219#
220# A list is constructed containing each input file that is to be processed
221#
222# Each file on the list is processed in a loop, using the associated handler
223# code for each:
224# The PropertyAliases.txt and PropValueAliases.txt files are processed
225# first. These files name the properties and property values.
226# Objects are created of all the property and property value names
227# that the rest of the input should expect, including all synonyms.
228# The other input files give mappings from properties to property
229# values. That is, they list code points and say what the mapping
230# is under the given property. Some files give the mappings for
231# just one property; and some for many. This program goes through
232ed87f
KW
232# each file and populates the properties and their map tables from
233# them. Some properties are listed in more than one file, and
234# Unicode has set up a precedence as to which has priority if there
235# is a conflict. Thus the order of processing matters, and this
236# program handles the conflict possibility by processing the
237# overriding input files last, so that if necessary they replace
238# earlier values.
99870f4d
KW
239# After this is all done, the program creates the property mappings not
240# furnished by Unicode, but derivable from what it does give.
241# The tables of code points that match each property value in each
242# property that is accessible by regular expressions are created.
243# The Perl-defined properties are created and populated. Many of these
244# require data determined from the earlier steps
245# Any Perl-defined synonyms are created, and name clashes between Perl
678f13d5 246# and Unicode are reconciled and warned about.
99870f4d
KW
247# All the properties are written to files
248# Any other files are written, and final warnings issued.
678f13d5 249#
99870f4d
KW
250# For clarity, a number of operators have been overloaded to work on tables:
251# ~ means invert (take all characters not in the set). The more
252# conventional '!' is not used because of the possibility of confusing
253# it with the actual boolean operation.
254# + means union
255# - means subtraction
256# & means intersection
257# The precedence of these is the order listed. Parentheses should be
258# copiously used. These are not a general scheme. The operations aren't
259# defined for a number of things, deliberately, to avoid getting into trouble.
260# Operations are done on references and affect the underlying structures, so
261# that the copy constructors for them have been overloaded to not return a new
262# clone, but the input object itself.
678f13d5 263#
99870f4d
KW
264# The bool operator is deliberately not overloaded to avoid confusion with
265# "should it mean if the object merely exists, or also is non-empty?".
99870f4d
KW
266#
267# WHY CERTAIN DESIGN DECISIONS WERE MADE
678f13d5
KW
268#
269# This program needs to be able to run under miniperl. Therefore, it uses a
270# minimum of other modules, and hence implements some things itself that could
271# be gotten from CPAN
272#
273# This program uses inputs published by the Unicode Consortium. These can
274# change incompatibly between releases without the Perl maintainers realizing
275# it. Therefore this program is now designed to try to flag these. It looks
276# at the directories where the inputs are, and flags any unrecognized files.
277# It keeps track of all the properties in the files it handles, and flags any
278# that it doesn't know how to handle. It also flags any input lines that
279# don't match the expected syntax, among other checks.
280#
281# It is also designed so if a new input file matches one of the known
282# templates, one hopefully just needs to add it to a list to have it
283# processed.
284#
285# As mentioned earlier, some properties are given in more than one file. In
286# particular, the files in the extracted directory are supposedly just
287# reformattings of the others. But they contain information not easily
9be22ca5
KW
288# derivable from the other files, including results for Unihan (which isn't
289# usually available to this program) and for unassigned code points. They
678f13d5
KW
290# also have historically had errors or been incomplete. In an attempt to
291# create the best possible data, this program thus processes them first to
292# glean information missing from the other files; then processes those other
293# files to override any errors in the extracted ones. Much of the design was
294# driven by this need to store things and then possibly override them.
295#
296# It tries to keep fatal errors to a minimum, to generate something usable for
297# testing purposes. It always looks for files that could be inputs, and will
298# warn about any that it doesn't know how to handle (the -q option suppresses
299# the warning).
99870f4d 300#
678f13d5
KW
301# Why is there more than one type of range?
302# This simplified things. There are some very specialized code points that
303# have to be handled specially for output, such as Hangul syllable names.
304# By creating a range type (done late in the development process), it
305# allowed this to be stored with the range, and overridden by other input.
306# Originally these were stored in another data structure, and it became a
307# mess trying to decide if a second file that was for the same property was
308# overriding the earlier one or not.
309#
310# Why are there two kinds of tables, match and map?
311# (And there is a base class shared by the two as well.) As stated above,
312# they actually are for different things. Development proceeded much more
313# smoothly when I (khw) realized the distinction. Map tables are used to
314# give the property value for every code point (actually every code point
315# that doesn't map to a default value). Match tables are used for regular
316# expression matches, and are essentially the inverse mapping. Separating
317# the two allows more specialized methods, and error checks so that one
318# can't just take the intersection of two map tables, for example, as that
319# is nonsensical.
99870f4d 320#
232ed87f
KW
321# What about 'fate' and 'status'. The concept of a table's fate was created
322# late when it became clear that something more was needed. The difference
323# between this and 'status' is unclean, and could be improved if someone
324# wanted to spend the effort.
325#
23e33b60
KW
326# DEBUGGING
327#
678f13d5
KW
328# This program is written so it will run under miniperl. Occasionally changes
329# will cause an error where the backtrace doesn't work well under miniperl.
330# To diagnose the problem, you can instead run it under regular perl, if you
331# have one compiled.
332#
333# There is a good trace facility. To enable it, first sub DEBUG must be set
334# to return true. Then a line like
335#
336# local $to_trace = 1 if main::DEBUG;
337#
232ed87f
KW
338# can be added to enable tracing in its lexical scope (plus dynamic) or until
339# you insert another line:
678f13d5
KW
340#
341# local $to_trace = 0 if main::DEBUG;
342#
232ed87f 343# To actually trace, use a line like "trace $a, @b, %c, ...;
678f13d5
KW
344#
345# Some of the more complex subroutines already have trace statements in them.
346# Permanent trace statements should be like:
347#
348# trace ... if main::DEBUG && $to_trace;
349#
350# If there is just one or a few files that you're debugging, you can easily
351# cause most everything else to be skipped. Change the line
352#
353# my $debug_skip = 0;
354#
355# to 1, and every file whose object is in @input_file_objects and doesn't have
232ed87f
KW
356# a, 'non_skip => 1,' in its constructor will be skipped. However, skipping
357# Jamo.txt or UnicodeData.txt will likely cause fatal errors.
678f13d5 358#
b4a0206c 359# To compare the output tables, it may be useful to specify the -annotate
97a8a595
KW
360# flag. (As of this writing, this can't be done on a clean workspace, due to
361# requirements in Text::Tabs used in this option; so first run mktables
362# without this option.) This option adds comment lines to each table, one for
363# each non-algorithmically named character giving, currently its code point,
364# name, and graphic representation if printable (and you have a font that
365# knows about it). This makes it easier to see what the particular code
366# points are in each output table. Non-named code points are annotated with a
367# description of their status, and contiguous ones with the same description
368# will be output as a range rather than individually. Algorithmically named
369# characters are also output as ranges, except when there are just a few
370# contiguous ones.
c4019d52 371#
99870f4d
KW
372# FUTURE ISSUES
373#
374# The program would break if Unicode were to change its names so that
375# interior white space, underscores, or dashes differences were significant
376# within property and property value names.
377#
378# It might be easier to use the xml versions of the UCD if this program ever
379# would need heavy revision, and the ability to handle old versions was not
380# required.
381#
382# There is the potential for name collisions, in that Perl has chosen names
383# that Unicode could decide it also likes. There have been such collisions in
384# the past, with mostly Perl deciding to adopt the Unicode definition of the
385# name. However in the 5.2 Unicode beta testing, there were a number of such
386# collisions, which were withdrawn before the final release, because of Perl's
387# and other's protests. These all involved new properties which began with
388# 'Is'. Based on the protests, Unicode is unlikely to try that again. Also,
389# many of the Perl-defined synonyms, like Any, Word, etc, are listed in a
390# Unicode document, so they are unlikely to be used by Unicode for another
391# purpose. However, they might try something beginning with 'In', or use any
392# of the other Perl-defined properties. This program will warn you of name
393# collisions, and refuse to generate tables with them, but manual intervention
394# will be required in this event. One scheme that could be implemented, if
395# necessary, would be to have this program generate another file, or add a
396# field to mktables.lst that gives the date of first definition of a property.
397# Each new release of Unicode would use that file as a basis for the next
398# iteration. And the Perl synonym addition code could sort based on the age
399# of the property, so older properties get priority, and newer ones that clash
400# would be refused; hence existing code would not be impacted, and some other
401# synonym would have to be used for the new property. This is ugly, and
402# manual intervention would certainly be easier to do in the short run; lets
403# hope it never comes to this.
678f13d5 404#
99870f4d
KW
405# A NOTE ON UNIHAN
406#
b9e57660
KW
407# This program can generate tables from the Unihan database. But that db
408# isn't normally available, so it is marked as optional. Prior to version
409# 5.2, this database was in a single file, Unihan.txt. In 5.2 the database
410# was split into 8 different files, all beginning with the letters 'Unihan'.
411# If you plunk those files down into the directory mktables ($0) is in, this
412# program will read them and automatically create tables for the properties
413# from it that are listed in PropertyAliases.txt and PropValueAliases.txt,
414# plus any you add to the @cjk_properties array and the @cjk_property_values
415# array, being sure to add necessary '# @missings' lines to the latter. For
416# Unicode versions earlier than 5.2, most of the Unihan properties are not
417# listed at all in PropertyAliases nor PropValueAliases. This program assumes
418# for these early releases that you want the properties that are specified in
419# the 5.2 release.
99870f4d
KW
420#
421# You may need to adjust the entries to suit your purposes. setup_unihan(),
422# and filter_unihan_line() are the functions where this is done. This program
423# already does some adjusting to make the lines look more like the rest of the
424# Unicode DB; You can see what that is in filter_unihan_line()
425#
426# There is a bug in the 3.2 data file in which some values for the
427# kPrimaryNumeric property have commas and an unexpected comment. A filter
9be22ca5
KW
428# could be added to correct these; or for a particular installation, the
429# Unihan.txt file could be edited to fix them.
99870f4d 430#
678f13d5
KW
431# HOW TO ADD A FILE TO BE PROCESSED
432#
433# A new file from Unicode needs to have an object constructed for it in
434# @input_file_objects, probably at the end or at the end of the extracted
435# ones. The program should warn you if its name will clash with others on
436# restrictive file systems, like DOS. If so, figure out a better name, and
437# add lines to the README.perl file giving that. If the file is a character
232ed87f 438# property, it should be in the format that Unicode has implicitly
678f13d5
KW
439# standardized for such files for the more recently introduced ones.
440# If so, the Input_file constructor for @input_file_objects can just be the
441# file name and release it first appeared in. If not, then it should be
442# possible to construct an each_line_handler() to massage the line into the
443# standardized form.
444#
445# For non-character properties, more code will be needed. You can look at
446# the existing entries for clues.
447#
448# UNICODE VERSIONS NOTES
449#
450# The Unicode UCD has had a number of errors in it over the versions. And
451# these remain, by policy, in the standard for that version. Therefore it is
452# risky to correct them, because code may be expecting the error. So this
453# program doesn't generally make changes, unless the error breaks the Perl
454# core. As an example, some versions of 2.1.x Jamo.txt have the wrong value
455# for U+1105, which causes real problems for the algorithms for Jamo
456# calculations, so it is changed here.
457#
458# But it isn't so clear cut as to what to do about concepts that are
459# introduced in a later release; should they extend back to earlier releases
460# where the concept just didn't exist? It was easier to do this than to not,
461# so that's what was done. For example, the default value for code points not
462# in the files for various properties was probably undefined until changed by
463# some version. No_Block for blocks is such an example. This program will
464# assign No_Block even in Unicode versions that didn't have it. This has the
465# benefit that code being written doesn't have to special case earlier
466# versions; and the detriment that it doesn't match the Standard precisely for
467# the affected versions.
468#
469# Here are some observations about some of the issues in early versions:
470#
232ed87f
KW
471# Prior to version 3.0, there were 3 character decompositions. These are not
472# handled by Unicode::Normalize, nor will it compile when presented a version
473# that has them. However, you can trivially get it to compile by simply
474# ignoring those decompositions, by changing the croak to a carp. At the time
9be22ca5
KW
475# of this writing, the line (in cpan/Unicode-Normalize/Normalize.pm or
476# cpan/Unicode-Normalize/mkheader) reads
232ed87f
KW
477#
478# croak("Weird Canonical Decomposition of U+$h");
479#
28807e1d 480# Simply comment it out. It will compile, but will not know about any three
9be22ca5 481# character decompositions.
232ed87f
KW
482
483# The number of code points in \p{alpha=True} halved in 2.1.9. It turns out
484# that the reason is that the CJK block starting at 4E00 was removed from
485# PropList, and was not put back in until 3.1.0. The Perl extension (the
486# single property name \p{alpha}) has the correct values. But the compound
487# form is simply not generated until 3.1, as it can be argued that prior to
488# this release, this was not an official property. The comments for
489# filter_old_style_proplist() give more details.
678f13d5
KW
490#
491# Unicode introduced the synonym Space for White_Space in 4.1. Perl has
492# always had a \p{Space}. In release 3.2 only, they are not synonymous. The
493# reason is that 3.2 introduced U+205F=medium math space, which was not
494# classed as white space, but Perl figured out that it should have been. 4.0
495# reclassified it correctly.
496#
497# Another change between 3.2 and 4.0 is the CCC property value ATBL. In 3.2
232ed87f
KW
498# this was erroneously a synonym for 202 (it should be 200). In 4.0, ATB
499# became 202, and ATBL was left with no code points, as all the ones that
500# mapped to 202 stayed mapped to 202. Thus if your program used the numeric
501# name for the class, it would not have been affected, but if it used the
502# mnemonic, it would have been.
678f13d5 503#
9be22ca5 504# \p{Script=Hrkt} (Katakana_Or_Hiragana) came in 4.0.1. Before that, code
678f13d5
KW
505# points which eventually came to have this script property value, instead
506# mapped to "Unknown". But in the next release all these code points were
507# moved to \p{sc=common} instead.
9be22ca5
KW
508
509# The tests furnished by Unicode for testing WordBreak and SentenceBreak
510# generate errors in 5.0 and earlier.
99870f4d
KW
511#
512# The default for missing code points for BidiClass is complicated. Starting
513# in 3.1.1, the derived file DBidiClass.txt handles this, but this program
514# tries to do the best it can for earlier releases. It is done in
515# process_PropertyAliases()
516#
232ed87f
KW
517# In version 2.1.2, the entry in UnicodeData.txt:
518# 0275;LATIN SMALL LETTER BARRED O;Ll;0;L;;;;;N;;;;019F;
519# should instead be
520# 0275;LATIN SMALL LETTER BARRED O;Ll;0;L;;;;;N;;;019F;;019F
521# Without this change, there are casing problems for this character.
522#
7803ad2d
KW
523# Search for $string_compare_versions to see how to compare changes to
524# properties between Unicode versions
525#
99870f4d
KW
526##############################################################################
527
528my $UNDEF = ':UNDEF:'; # String to print out for undefined values in tracing
529 # and errors
530my $MAX_LINE_WIDTH = 78;
531
532# Debugging aid to skip most files so as to not be distracted by them when
533# concentrating on the ones being debugged. Add
534# non_skip => 1,
535# to the constructor for those files you want processed when you set this.
536# Files with a first version number of 0 are special: they are always
c12f2655
KW
537# processed regardless of the state of this flag. Generally, Jamo.txt and
538# UnicodeData.txt must not be skipped if you want this program to not die
539# before normal completion.
99870f4d
KW
540my $debug_skip = 0;
541
e9c4b4f8
KW
542
543# Normally these are suppressed.
544my $write_Unicode_deprecated_tables = 0;
545
99870f4d
KW
546# Set to 1 to enable tracing.
547our $to_trace = 0;
548
549{ # Closure for trace: debugging aid
550 my $print_caller = 1; # ? Include calling subroutine name
551 my $main_with_colon = 'main::';
552 my $main_colon_length = length($main_with_colon);
553
554 sub trace {
555 return unless $to_trace; # Do nothing if global flag not set
556
557 my @input = @_;
558
559 local $DB::trace = 0;
560 $DB::trace = 0; # Quiet 'used only once' message
561
562 my $line_number;
563
564 # Loop looking up the stack to get the first non-trace caller
565 my $caller_line;
566 my $caller_name;
567 my $i = 0;
568 do {
569 $line_number = $caller_line;
570 (my $pkg, my $file, $caller_line, my $caller) = caller $i++;
571 $caller = $main_with_colon unless defined $caller;
572
573 $caller_name = $caller;
574
575 # get rid of pkg
576 $caller_name =~ s/.*:://;
577 if (substr($caller_name, 0, $main_colon_length)
578 eq $main_with_colon)
579 {
580 $caller_name = substr($caller_name, $main_colon_length);
581 }
582
583 } until ($caller_name ne 'trace');
584
585 # If the stack was empty, we were called from the top level
586 $caller_name = 'main' if ($caller_name eq ""
587 || $caller_name eq 'trace');
588
589 my $output = "";
9346f59a 590 #print STDERR __LINE__, ": ", join ", ", @input, "\n";
99870f4d 591 foreach my $string (@input) {
99870f4d
KW
592 if (ref $string eq 'ARRAY' || ref $string eq 'HASH') {
593 $output .= simple_dumper($string);
594 }
595 else {
596 $string = "$string" if ref $string;
597 $string = $UNDEF unless defined $string;
598 chomp $string;
599 $string = '""' if $string eq "";
600 $output .= " " if $output ne ""
601 && $string ne ""
602 && substr($output, -1, 1) ne " "
603 && substr($string, 0, 1) ne " ";
604 $output .= $string;
605 }
606 }
607
99f78760
KW
608 print STDERR sprintf "%4d: ", $line_number if defined $line_number;
609 print STDERR "$caller_name: " if $print_caller;
99870f4d
KW
610 print STDERR $output, "\n";
611 return;
612 }
613}
614
615# This is for a rarely used development feature that allows you to compare two
616# versions of the Unicode standard without having to deal with changes caused
9be22ca5
KW
617# by the code points introduced in the later version. You probably also want
618# to use the -annotate option when using this. Change the 0 to a string
619# containing a SINGLE dotted Unicode release number (e.g. "2.1"). Only code
620# points introduced in that release and earlier will be used; later ones are
621# thrown away. You use the version number of the earliest one you want to
c12f2655
KW
622# compare; then run this program on directory structures containing each
623# release, and compare the outputs. These outputs will therefore include only
624# the code points common to both releases, and you can see the changes caused
625# just by the underlying release semantic changes. For versions earlier than
626# 3.2, you must copy a version of DAge.txt into the directory.
627my $string_compare_versions = DEBUG && 0; # e.g., "2.1";
99870f4d
KW
628my $compare_versions = DEBUG
629 && $string_compare_versions
630 && pack "C*", split /\./, $string_compare_versions;
631
632sub uniques {
633 # Returns non-duplicated input values. From "Perl Best Practices:
634 # Encapsulated Cleverness". p. 455 in first edition.
635
636 my %seen;
0e407844
NC
637 # Arguably this breaks encapsulation, if the goal is to permit multiple
638 # distinct objects to stringify to the same value, and be interchangeable.
639 # However, for this program, no two objects stringify identically, and all
640 # lists passed to this function are either objects or strings. So this
641 # doesn't affect correctness, but it does give a couple of percent speedup.
642 no overloading;
99870f4d
KW
643 return grep { ! $seen{$_}++ } @_;
644}
645
646$0 = File::Spec->canonpath($0);
647
648my $make_test_script = 0; # ? Should we output a test script
6b5ab373 649my $make_norm_test_script = 0; # ? Should we output a normalization test script
99870f4d
KW
650my $write_unchanged_files = 0; # ? Should we update the output files even if
651 # we don't think they have changed
652my $use_directory = ""; # ? Should we chdir somewhere.
653my $pod_directory; # input directory to store the pod file.
654my $pod_file = 'perluniprops';
655my $t_path; # Path to the .t test file
656my $file_list = 'mktables.lst'; # File to store input and output file names.
657 # This is used to speed up the build, by not
658 # executing the main body of the program if
659 # nothing on the list has changed since the
660 # previous build
661my $make_list = 1; # ? Should we write $file_list. Set to always
662 # make a list so that when the pumpking is
663 # preparing a release, s/he won't have to do
664 # special things
665my $glob_list = 0; # ? Should we try to include unknown .txt files
666 # in the input.
bd9ebcfd
KW
667my $output_range_counts = $debugging_build; # ? Should we include the number
668 # of code points in ranges in
669 # the output
558712cf 670my $annotate = 0; # ? Should character names be in the output
9ef2b94f 671
99870f4d
KW
672# Verbosity levels; 0 is quiet
673my $NORMAL_VERBOSITY = 1;
674my $PROGRESS = 2;
675my $VERBOSE = 3;
676
677my $verbosity = $NORMAL_VERBOSITY;
678
0458fbc1
KW
679# Stored in mktables.lst so that if this program is called with different
680# options, will regenerate even if the files otherwise look like they're
681# up-to-date.
682my $command_line_arguments = join " ", @ARGV;
683
99870f4d
KW
684# Process arguments
685while (@ARGV) {
cf25bb62
JH
686 my $arg = shift @ARGV;
687 if ($arg eq '-v') {
99870f4d
KW
688 $verbosity = $VERBOSE;
689 }
690 elsif ($arg eq '-p') {
691 $verbosity = $PROGRESS;
692 $| = 1; # Flush buffers as we go.
693 }
694 elsif ($arg eq '-q') {
695 $verbosity = 0;
696 }
697 elsif ($arg eq '-w') {
698 $write_unchanged_files = 1; # update the files even if havent changed
699 }
700 elsif ($arg eq '-check') {
6ae7e459
YO
701 my $this = shift @ARGV;
702 my $ok = shift @ARGV;
703 if ($this ne $ok) {
704 print "Skipping as check params are not the same.\n";
705 exit(0);
706 }
00a8df5c 707 }
99870f4d
KW
708 elsif ($arg eq '-P' && defined ($pod_directory = shift)) {
709 -d $pod_directory or croak "Directory '$pod_directory' doesn't exist";
710 }
3df51b85
KW
711 elsif ($arg eq '-maketest' || ($arg eq '-T' && defined ($t_path = shift)))
712 {
99870f4d 713 $make_test_script = 1;
99870f4d 714 }
6b5ab373
KW
715 elsif ($arg eq '-makenormtest')
716 {
717 $make_norm_test_script = 1;
718 }
99870f4d
KW
719 elsif ($arg eq '-makelist') {
720 $make_list = 1;
721 }
722 elsif ($arg eq '-C' && defined ($use_directory = shift)) {
723 -d $use_directory or croak "Unknown directory '$use_directory'";
724 }
725 elsif ($arg eq '-L') {
726
727 # Existence not tested until have chdir'd
728 $file_list = shift;
729 }
730 elsif ($arg eq '-globlist') {
731 $glob_list = 1;
732 }
733 elsif ($arg eq '-c') {
734 $output_range_counts = ! $output_range_counts
735 }
b4a0206c 736 elsif ($arg eq '-annotate') {
558712cf 737 $annotate = 1;
bd9ebcfd
KW
738 $debugging_build = 1;
739 $output_range_counts = 1;
9ef2b94f 740 }
99870f4d
KW
741 else {
742 my $with_c = 'with';
743 $with_c .= 'out' if $output_range_counts; # Complements the state
744 croak <<END;
745usage: $0 [-c|-p|-q|-v|-w] [-C dir] [-L filelist] [ -P pod_dir ]
746 [ -T test_file_path ] [-globlist] [-makelist] [-maketest]
747 [-check A B ]
748 -c : Output comments $with_c number of code points in ranges
749 -q : Quiet Mode: Only output serious warnings.
750 -p : Set verbosity level to normal plus show progress.
751 -v : Set Verbosity level high: Show progress and non-serious
752 warnings
753 -w : Write files regardless
754 -C dir : Change to this directory before proceeding. All relative paths
755 except those specified by the -P and -T options will be done
756 with respect to this directory.
757 -P dir : Output $pod_file file to directory 'dir'.
3df51b85 758 -T path : Create a test script as 'path'; overrides -maketest
99870f4d
KW
759 -L filelist : Use alternate 'filelist' instead of standard one
760 -globlist : Take as input all non-Test *.txt files in current and sub
761 directories
3df51b85
KW
762 -maketest : Make test script 'TestProp.pl' in current (or -C directory),
763 overrides -T
99870f4d 764 -makelist : Rewrite the file list $file_list based on current setup
b4a0206c 765 -annotate : Output an annotation for each character in the table files;
97a8a595
KW
766 useful for debugging mktables, looking at diffs; but is slow
767 and memory intensive
99870f4d
KW
768 -check A B : Executes $0 only if A and B are the same
769END
770 }
771}
772
773# Stores the most-recently changed file. If none have changed, can skip the
774# build
aeab6150 775my $most_recent = (stat $0)[9]; # Do this before the chdir!
99870f4d
KW
776
777# Change directories now, because need to read 'version' early.
778if ($use_directory) {
3df51b85 779 if ($pod_directory && ! File::Spec->file_name_is_absolute($pod_directory)) {
99870f4d
KW
780 $pod_directory = File::Spec->rel2abs($pod_directory);
781 }
3df51b85 782 if ($t_path && ! File::Spec->file_name_is_absolute($t_path)) {
99870f4d 783 $t_path = File::Spec->rel2abs($t_path);
00a8df5c 784 }
99870f4d 785 chdir $use_directory or croak "Failed to chdir to '$use_directory':$!";
3df51b85 786 if ($pod_directory && File::Spec->file_name_is_absolute($pod_directory)) {
99870f4d 787 $pod_directory = File::Spec->abs2rel($pod_directory);
02b1aeec 788 }
3df51b85 789 if ($t_path && File::Spec->file_name_is_absolute($t_path)) {
99870f4d 790 $t_path = File::Spec->abs2rel($t_path);
02b1aeec 791 }
00a8df5c
YO
792}
793
99870f4d
KW
794# Get Unicode version into regular and v-string. This is done now because
795# various tables below get populated based on it. These tables are populated
796# here to be near the top of the file, and so easily seeable by those needing
797# to modify things.
798open my $VERSION, "<", "version"
799 or croak "$0: can't open required file 'version': $!\n";
800my $string_version = <$VERSION>;
801close $VERSION;
802chomp $string_version;
803my $v_version = pack "C*", split /\./, $string_version; # v string
804
805# The following are the complete names of properties with property values that
806# are known to not match any code points in some versions of Unicode, but that
807# may change in the future so they should be matchable, hence an empty file is
808# generated for them.
a9c9e371
KW
809my @tables_that_may_be_empty;
810push @tables_that_may_be_empty, 'Joining_Type=Left_Joining'
811 if $v_version lt v6.3.0;
99870f4d
KW
812push @tables_that_may_be_empty, 'Script=Common' if $v_version le v4.0.1;
813push @tables_that_may_be_empty, 'Title' if $v_version lt v2.0.0;
814push @tables_that_may_be_empty, 'Script=Katakana_Or_Hiragana'
815 if $v_version ge v4.1.0;
82aed44a
KW
816push @tables_that_may_be_empty, 'Script_Extensions=Katakana_Or_Hiragana'
817 if $v_version ge v6.0.0;
f583b44c
KW
818push @tables_that_may_be_empty, 'Grapheme_Cluster_Break=Prepend'
819 if $v_version ge v6.1.0;
1e958ea9
KW
820push @tables_that_may_be_empty, 'Canonical_Combining_Class=CCC133'
821 if $v_version ge v6.2.0;
99870f4d
KW
822
823# The lists below are hashes, so the key is the item in the list, and the
824# value is the reason why it is in the list. This makes generation of
825# documentation easier.
826
827my %why_suppressed; # No file generated for these.
828
829# Files aren't generated for empty extraneous properties. This is arguable.
830# Extraneous properties generally come about because a property is no longer
831# used in a newer version of Unicode. If we generated a file without code
832# points, programs that used to work on that property will still execute
833# without errors. It just won't ever match (or will always match, with \P{}).
834# This means that the logic is now likely wrong. I (khw) think its better to
835# find this out by getting an error message. Just move them to the table
836# above to change this behavior
837my %why_suppress_if_empty_warn_if_not = (
838
839 # It is the only property that has ever officially been removed from the
840 # Standard. The database never contained any code points for it.
841 'Special_Case_Condition' => 'Obsolete',
842
843 # Apparently never official, but there were code points in some versions of
844 # old-style PropList.txt
845 'Non_Break' => 'Obsolete',
846);
847
848# These would normally go in the warn table just above, but they were changed
849# a long time before this program was written, so warnings about them are
850# moot.
851if ($v_version gt v3.2.0) {
852 push @tables_that_may_be_empty,
853 'Canonical_Combining_Class=Attached_Below_Left'
854}
855
9be22ca5
KW
856# Enum values for to_output_map() method in the Map_Table package. (0 is don't
857# output)
272501f6
KW
858my $EXTERNAL_MAP = 1;
859my $INTERNAL_MAP = 2;
ce712c88 860my $OUTPUT_ADJUSTED = 3;
272501f6 861
fcf1973c
KW
862# To override computed values for writing the map tables for these properties.
863# The default for enum map tables is to write them out, so that the Unicode
864# .txt files can be removed, but all the data to compute any property value
865# for any code point is available in a more compact form.
866my %global_to_output_map = (
867 # Needed by UCD.pm, but don't want to publicize that it exists, so won't
c12f2655
KW
868 # get stuck supporting it if things change. Since it is a STRING
869 # property, it normally would be listed in the pod, but INTERNAL_MAP
870 # suppresses that.
fcf1973c
KW
871 Unicode_1_Name => $INTERNAL_MAP,
872
873 Present_In => 0, # Suppress, as easily computed from Age
74cd47d0
KW
874 Block => (NON_ASCII_PLATFORM) ? 1 : 0, # Suppress, as Blocks.txt is
875 # retained, but needed for
876 # non-ASCII
53d34b6c
KW
877
878 # Suppress, as mapping can be found instead from the
879 # Perl_Decomposition_Mapping file
880 Decomposition_Type => 0,
fcf1973c
KW
881);
882
99870f4d
KW
883# There are several types of obsolete properties defined by Unicode. These
884# must be hand-edited for every new Unicode release.
885my %why_deprecated; # Generates a deprecated warning message if used.
886my %why_stabilized; # Documentation only
887my %why_obsolete; # Documentation only
888
889{ # Closure
8364f9eb 890 my $simple = 'Perl uses the more complete version';
99870f4d
KW
891 my $unihan = 'Unihan properties are by default not enabled in the Perl core. Instead use CPAN: Unicode::Unihan';
892
893 my $other_properties = 'other properties';
894 my $contributory = "Used by Unicode internally for generating $other_properties and not intended to be used stand-alone";
5d294d41 895 my $why_no_expand = "Deprecated by Unicode. These are characters that expand to more than one character in the specified normalization form, but whether they actually take up more bytes or not depends on the encoding being used. For example, a UTF-8 encoded character may expand to a different number of bytes than a UTF-32 encoded character.";
99870f4d
KW
896
897 %why_deprecated = (
5f7264c7 898 'Grapheme_Link' => 'Deprecated by Unicode: Duplicates ccc=vr (Canonical_Combining_Class=Virama)',
99870f4d
KW
899 'Jamo_Short_Name' => $contributory,
900 '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',
901 'Other_Alphabetic' => $contributory,
902 'Other_Default_Ignorable_Code_Point' => $contributory,
903 'Other_Grapheme_Extend' => $contributory,
904 'Other_ID_Continue' => $contributory,
905 'Other_ID_Start' => $contributory,
906 'Other_Lowercase' => $contributory,
907 'Other_Math' => $contributory,
908 'Other_Uppercase' => $contributory,
e22aaf5c
KW
909 'Expands_On_NFC' => $why_no_expand,
910 'Expands_On_NFD' => $why_no_expand,
911 'Expands_On_NFKC' => $why_no_expand,
912 'Expands_On_NFKD' => $why_no_expand,
99870f4d
KW
913 );
914
915 %why_suppressed = (
5f7264c7 916 # There is a lib/unicore/Decomposition.pl (used by Normalize.pm) which
99870f4d
KW
917 # contains the same information, but without the algorithmically
918 # determinable Hangul syllables'. This file is not published, so it's
919 # existence is not noted in the comment.
12fee290 920 'Decomposition_Mapping' => 'Accessible via Unicode::Normalize or prop_invmap() or charprop() in Unicode::UCD::',
99870f4d 921
5f8d1a89
KW
922 # Don't suppress ISO_Comment, as otherwise special handling is needed
923 # to differentiate between it and gc=c, which can be written as 'isc',
924 # which is the same characters as ISO_Comment's short name.
99870f4d 925
12fee290 926 'Name' => "Accessible via \\N{...} or 'use charnames;' or charprop() or prop_invmap() in Unicode::UCD::",
e0b29447 927
12fee290
KW
928 'Simple_Case_Folding' => "$simple. Can access this through casefold(), charprop(), or prop_invmap() in Unicode::UCD",
929 'Simple_Lowercase_Mapping' => "$simple. Can access this through charinfo(), charprop(), or prop_invmap() in Unicode::UCD",
930 'Simple_Titlecase_Mapping' => "$simple. Can access this through charinfo(), charprop(), or prop_invmap() in Unicode::UCD",
931 'Simple_Uppercase_Mapping' => "$simple. Can access this through charinfo(), charprop(), or prop_invmap() in Unicode::UCD",
99870f4d 932
dac6f618 933 FC_NFKC_Closure => 'Deprecated by Unicode, and supplanted in usage by NFKC_Casefold; otherwise not useful',
99870f4d
KW
934 );
935
1704a0ea
KW
936 foreach my $property (
937
938 # The following are suppressed because they were made contributory
939 # or deprecated by Unicode before Perl ever thought about
940 # supporting them.
941 'Jamo_Short_Name',
942 'Grapheme_Link',
943 'Expands_On_NFC',
944 'Expands_On_NFD',
945 'Expands_On_NFKC',
946 'Expands_On_NFKD',
947
948 # The following are suppressed because they have been marked
949 # as deprecated for a sufficient amount of time
950 'Other_Alphabetic',
951 'Other_Default_Ignorable_Code_Point',
952 'Other_Grapheme_Extend',
953 'Other_ID_Continue',
954 'Other_ID_Start',
955 'Other_Lowercase',
956 'Other_Math',
957 'Other_Uppercase',
e22aaf5c 958 ) {
99870f4d
KW
959 $why_suppressed{$property} = $why_deprecated{$property};
960 }
cf25bb62 961
99870f4d
KW
962 # Customize the message for all the 'Other_' properties
963 foreach my $property (keys %why_deprecated) {
964 next if (my $main_property = $property) !~ s/^Other_//;
965 $why_deprecated{$property} =~ s/$other_properties/the $main_property property (which should be used instead)/;
966 }
967}
968
e9c4b4f8
KW
969if ($write_Unicode_deprecated_tables) {
970 foreach my $property (keys %why_suppressed) {
971 delete $why_suppressed{$property} if $property =~
972 / ^ Other | Grapheme /x;
973 }
974}
975
99870f4d
KW
976if ($v_version ge 4.0.0) {
977 $why_stabilized{'Hyphen'} = 'Use the Line_Break property instead; see www.unicode.org/reports/tr14';
5f7264c7
KW
978 if ($v_version ge 6.0.0) {
979 $why_deprecated{'Hyphen'} = 'Supplanted by Line_Break property values; see www.unicode.org/reports/tr14';
980 }
99870f4d 981}
5f7264c7 982if ($v_version ge 5.2.0 && $v_version lt 6.0.0) {
99870f4d 983 $why_obsolete{'ISO_Comment'} = 'Code points for it have been removed';
5f7264c7 984 if ($v_version ge 6.0.0) {
63f74647 985 $why_deprecated{'ISO_Comment'} = 'No longer needed for Unicode\'s internal chart generation; otherwise not useful, and code points for it have been removed';
5f7264c7 986 }
99870f4d
KW
987}
988
989# Probably obsolete forever
990if ($v_version ge v4.1.0) {
82aed44a
KW
991 $why_suppressed{'Script=Katakana_Or_Hiragana'} = 'Obsolete. All code points previously matched by this have been moved to "Script=Common".';
992}
993if ($v_version ge v6.0.0) {
caa75395 994 $why_suppressed{'Script=Katakana_Or_Hiragana'} .= ' Consider instead using "Script_Extensions=Katakana" or "Script_Extensions=Hiragana" (or both)';
2b352efd 995 $why_suppressed{'Script_Extensions=Katakana_Or_Hiragana'} = 'All code points that would be matched by this are matched by either "Script_Extensions=Katakana" or "Script_Extensions=Hiragana"';
99870f4d
KW
996}
997
998# This program can create files for enumerated-like properties, such as
999# 'Numeric_Type'. This file would be the same format as for a string
1000# property, with a mapping from code point to its value, so you could look up,
1001# for example, the script a code point is in. But no one so far wants this
1002# mapping, or they have found another way to get it since this is a new
1003# feature. So no file is generated except if it is in this list.
1004my @output_mapped_properties = split "\n", <<END;
1005END
1006
b9e57660
KW
1007# If you want more Unihan properties than the default, you need to add them to
1008# these arrays. Depending on the property type, @missing lines might have to
1009# be added to the second array. A sample entry would be (including the '#'):
1010# @missing: 0000..10FFFF; cjkAccountingNumeric; NaN
99870f4d 1011my @cjk_properties = split "\n", <<'END';
99870f4d 1012END
99870f4d 1013my @cjk_property_values = split "\n", <<'END';
99870f4d
KW
1014END
1015
1016# The input files don't list every code point. Those not listed are to be
1017# defaulted to some value. Below are hard-coded what those values are for
1018# non-binary properties as of 5.1. Starting in 5.0, there are
caa75395 1019# machine-parsable comment lines in the files that give the defaults; so this
99870f4d
KW
1020# list shouldn't have to be extended. The claim is that all missing entries
1021# for binary properties will default to 'N'. Unicode tried to change that in
1022# 5.2, but the beta period produced enough protest that they backed off.
1023#
1024# The defaults for the fields that appear in UnicodeData.txt in this hash must
1025# be in the form that it expects. The others may be synonyms.
1026my $CODE_POINT = '<code point>';
1027my %default_mapping = (
1028 Age => "Unassigned",
1029 # Bidi_Class => Complicated; set in code
1030 Bidi_Mirroring_Glyph => "",
1031 Block => 'No_Block',
1032 Canonical_Combining_Class => 0,
1033 Case_Folding => $CODE_POINT,
1034 Decomposition_Mapping => $CODE_POINT,
1035 Decomposition_Type => 'None',
1036 East_Asian_Width => "Neutral",
1037 FC_NFKC_Closure => $CODE_POINT,
ec2cd995 1038 General_Category => ($v_version le 6.3.0) ? 'Cn' : 'Unassigned',
99870f4d
KW
1039 Grapheme_Cluster_Break => 'Other',
1040 Hangul_Syllable_Type => 'NA',
1041 ISO_Comment => "",
1042 Jamo_Short_Name => "",
1043 Joining_Group => "No_Joining_Group",
1044 # Joining_Type => Complicated; set in code
1045 kIICore => 'N', # Is converted to binary
1046 #Line_Break => Complicated; set in code
1047 Lowercase_Mapping => $CODE_POINT,
1048 Name => "",
1049 Name_Alias => "",
1050 NFC_QC => 'Yes',
1051 NFD_QC => 'Yes',
1052 NFKC_QC => 'Yes',
1053 NFKD_QC => 'Yes',
1054 Numeric_Type => 'None',
1055 Numeric_Value => 'NaN',
1056 Script => ($v_version le 4.1.0) ? 'Common' : 'Unknown',
1057 Sentence_Break => 'Other',
1058 Simple_Case_Folding => $CODE_POINT,
1059 Simple_Lowercase_Mapping => $CODE_POINT,
1060 Simple_Titlecase_Mapping => $CODE_POINT,
1061 Simple_Uppercase_Mapping => $CODE_POINT,
1062 Titlecase_Mapping => $CODE_POINT,
1063 Unicode_1_Name => "",
1064 Unicode_Radical_Stroke => "",
1065 Uppercase_Mapping => $CODE_POINT,
1066 Word_Break => 'Other',
1067);
1068
678f13d5 1069### End of externally interesting definitions, except for @input_file_objects
99870f4d
KW
1070
1071my $HEADER=<<"EOF";
1072# !!!!!!! DO NOT EDIT THIS FILE !!!!!!!
3df51b85
KW
1073# This file is machine-generated by $0 from the Unicode
1074# database, Version $string_version. Any changes made here will be lost!
cf25bb62
JH
1075EOF
1076
126c3d4e 1077my $INTERNAL_ONLY_HEADER = <<"EOF";
99870f4d
KW
1078
1079# !!!!!!! INTERNAL PERL USE ONLY !!!!!!!
fac53429
KW
1080# This file is for internal use by core Perl only. The format and even the
1081# name or existence of this file are subject to change without notice. Don't
d9ae3878
KW
1082# use it directly. Use Unicode::UCD to access the Unicode character data
1083# base.
99870f4d
KW
1084EOF
1085
1086my $DEVELOPMENT_ONLY=<<"EOF";
1087# !!!!!!! DEVELOPMENT USE ONLY !!!!!!!
1088# This file contains information artificially constrained to code points
1089# present in Unicode release $string_compare_versions.
1090# IT CANNOT BE RELIED ON. It is for use during development only and should
23e33b60 1091# not be used for production.
b6922eda
KW
1092
1093EOF
1094
346a20cf
KW
1095my $MAX_UNICODE_CODEPOINT_STRING = ($v_version ge v2.0.0)
1096 ? "10FFFF"
1097 : "FFFF";
6189eadc
KW
1098my $MAX_UNICODE_CODEPOINT = hex $MAX_UNICODE_CODEPOINT_STRING;
1099my $MAX_UNICODE_CODEPOINTS = $MAX_UNICODE_CODEPOINT + 1;
99870f4d 1100
2d88a86a
KW
1101# We work with above-Unicode code points, up to UV_MAX. But when you get
1102# that high, above IV_MAX, some operations don't work, and you can easily get
1103# overflow. Therefore for internal use, we use a much smaller number,
1104# translating it to UV_MAX only for output. The exact number is immaterial
1105# (all Unicode code points are treated exactly the same), but the algorithm
1106# requires it to be at least 2 * $MAX_UNICODE_CODEPOINTS + 1;
1107my $MAX_WORKING_CODEPOINTS= $MAX_UNICODE_CODEPOINT * 8;
1108my $MAX_WORKING_CODEPOINT = $MAX_WORKING_CODEPOINTS - 1;
1109my $MAX_WORKING_CODEPOINT_STRING = sprintf("%X", $MAX_WORKING_CODEPOINT);
1110
1111my $MAX_PLATFORM_CODEPOINT = ~0;
1112
99870f4d
KW
1113# Matches legal code point. 4-6 hex numbers, If there are 6, the first
1114# two must be 10; if there are 5, the first must not be a 0. Written this way
92199589
KW
1115# to decrease backtracking. The first regex allows the code point to be at
1116# the end of a word, but to work properly, the word shouldn't end with a valid
1117# hex character. The second one won't match a code point at the end of a
1118# word, and doesn't have the run-on issue
8c32d378
KW
1119my $run_on_code_point_re =
1120 qr/ (?: 10[0-9A-F]{4} | [1-9A-F][0-9A-F]{4} | [0-9A-F]{4} ) \b/x;
1121my $code_point_re = qr/\b$run_on_code_point_re/;
99870f4d
KW
1122
1123# This matches the beginning of the line in the Unicode db files that give the
1124# defaults for code points not listed (i.e., missing) in the file. The code
1125# depends on this ending with a semi-colon, so it can assume it is a valid
1126# field when the line is split() by semi-colons
346a20cf 1127my $missing_defaults_prefix = qr/^#\s+\@missing:\s+0000\.\.10FFFF\s*;/;
99870f4d
KW
1128
1129# Property types. Unicode has more types, but these are sufficient for our
1130# purposes.
1131my $UNKNOWN = -1; # initialized to illegal value
1132my $NON_STRING = 1; # Either binary or enum
1133my $BINARY = 2;
06f26c45
KW
1134my $FORCED_BINARY = 3; # Not a binary property, but, besides its normal
1135 # tables, additional true and false tables are
1136 # generated so that false is anything matching the
1137 # default value, and true is everything else.
1138my $ENUM = 4; # Include catalog
1139my $STRING = 5; # Anything else: string or misc
99870f4d
KW
1140
1141# Some input files have lines that give default values for code points not
1142# contained in the file. Sometimes these should be ignored.
1143my $NO_DEFAULTS = 0; # Must evaluate to false
1144my $NOT_IGNORED = 1;
1145my $IGNORED = 2;
1146
1147# Range types. Each range has a type. Most ranges are type 0, for normal,
1148# and will appear in the main body of the tables in the output files, but
1149# there are other types of ranges as well, listed below, that are specially
1150# handled. There are pseudo-types as well that will never be stored as a
1151# type, but will affect the calculation of the type.
1152
1153# 0 is for normal, non-specials
1154my $MULTI_CP = 1; # Sequence of more than code point
1155my $HANGUL_SYLLABLE = 2;
1156my $CP_IN_NAME = 3; # The NAME contains the code point appended to it.
1157my $NULL = 4; # The map is to the null string; utf8.c can't
1158 # handle these, nor is there an accepted syntax
1159 # for them in \p{} constructs
f86864ac 1160my $COMPUTE_NO_MULTI_CP = 5; # Pseudo-type; means that ranges that would
99870f4d
KW
1161 # otherwise be $MULTI_CP type are instead type 0
1162
1163# process_generic_property_file() can accept certain overrides in its input.
1164# Each of these must begin AND end with $CMD_DELIM.
1165my $CMD_DELIM = "\a";
1166my $REPLACE_CMD = 'replace'; # Override the Replace
1167my $MAP_TYPE_CMD = 'map_type'; # Override the Type
1168
1169my $NO = 0;
1170my $YES = 1;
1171
1172# Values for the Replace argument to add_range.
1173# $NO # Don't replace; add only the code points not
1174 # already present.
1175my $IF_NOT_EQUIVALENT = 1; # Replace only under certain conditions; details in
1176 # the comments at the subroutine definition.
1177my $UNCONDITIONALLY = 2; # Replace without conditions.
9470941f 1178my $MULTIPLE_BEFORE = 4; # Don't replace, but add a duplicate record if
99870f4d 1179 # already there
7f4b1e25
KW
1180my $MULTIPLE_AFTER = 5; # Don't replace, but add a duplicate record if
1181 # already there
1182my $CROAK = 6; # Die with an error if is already there
99870f4d
KW
1183
1184# Flags to give property statuses. The phrases are to remind maintainers that
1185# if the flag is changed, the indefinite article referring to it in the
1186# documentation may need to be as well.
1187my $NORMAL = "";
99870f4d
KW
1188my $DEPRECATED = 'D';
1189my $a_bold_deprecated = "a 'B<$DEPRECATED>'";
1190my $A_bold_deprecated = "A 'B<$DEPRECATED>'";
1191my $DISCOURAGED = 'X';
1192my $a_bold_discouraged = "an 'B<$DISCOURAGED>'";
1193my $A_bold_discouraged = "An 'B<$DISCOURAGED>'";
1194my $STRICTER = 'T';
1195my $a_bold_stricter = "a 'B<$STRICTER>'";
1196my $A_bold_stricter = "A 'B<$STRICTER>'";
1197my $STABILIZED = 'S';
1198my $a_bold_stabilized = "an 'B<$STABILIZED>'";
1199my $A_bold_stabilized = "An 'B<$STABILIZED>'";
1200my $OBSOLETE = 'O';
1201my $a_bold_obsolete = "an 'B<$OBSOLETE>'";
1202my $A_bold_obsolete = "An 'B<$OBSOLETE>'";
1203
5c6d9947
KW
1204# Aliases can also have an extra status:
1205my $INTERNAL_ALIAS = 'P';
1206
99870f4d
KW
1207my %status_past_participles = (
1208 $DISCOURAGED => 'discouraged',
99870f4d
KW
1209 $STABILIZED => 'stabilized',
1210 $OBSOLETE => 'obsolete',
37e2e78e 1211 $DEPRECATED => 'deprecated',
5c6d9947 1212 $INTERNAL_ALIAS => 'reserved for Perl core internal use only',
99870f4d
KW
1213);
1214
395dfc19
KW
1215# Table fates. These are somewhat ordered, so that fates < $MAP_PROXIED should be
1216# externally documented.
301ba948 1217my $ORDINARY = 0; # The normal fate.
395dfc19
KW
1218my $MAP_PROXIED = 1; # The map table for the property isn't written out,
1219 # but there is a file written that can be used to
1220 # reconstruct this table
3cdaf629 1221my $INTERNAL_ONLY = 2; # The file for this table is written out, but it is
301ba948 1222 # for Perl's internal use only
277b7b16
KW
1223my $LEGACY_ONLY = 3; # Like $INTERNAL_ONLY, but not actually used by Perl.
1224 # Is for backwards compatibility for applications that
1225 # read the file directly, so it's format is
1226 # unchangeable.
1227my $SUPPRESSED = 4; # The file for this table is not written out, and as a
3cdaf629
KW
1228 # result, we don't bother to do many computations on
1229 # it.
277b7b16 1230my $PLACEHOLDER = 5; # Like $SUPPRESSED, but we go through all the
3cdaf629
KW
1231 # computations anyway, as the values are needed for
1232 # things to work. This happens when we have Perl
1233 # extensions that depend on Unicode tables that
1234 # wouldn't normally be in a given Unicode version.
301ba948 1235
f5817e0a
KW
1236# The format of the values of the tables:
1237my $EMPTY_FORMAT = "";
99870f4d
KW
1238my $BINARY_FORMAT = 'b';
1239my $DECIMAL_FORMAT = 'd';
1240my $FLOAT_FORMAT = 'f';
1241my $INTEGER_FORMAT = 'i';
1242my $HEX_FORMAT = 'x';
1243my $RATIONAL_FORMAT = 'r';
1244my $STRING_FORMAT = 's';
d11155ec 1245my $ADJUST_FORMAT = 'a';
24303724 1246my $HEX_ADJUST_FORMAT = 'ax';
a14f3cb1 1247my $DECOMP_STRING_FORMAT = 'c';
c3ff2976 1248my $STRING_WHITE_SPACE_LIST = 'sw';
99870f4d
KW
1249
1250my %map_table_formats = (
1251 $BINARY_FORMAT => 'binary',
1252 $DECIMAL_FORMAT => 'single decimal digit',
1253 $FLOAT_FORMAT => 'floating point number',
1254 $INTEGER_FORMAT => 'integer',
add63c13 1255 $HEX_FORMAT => 'non-negative hex whole number; a code point',
99870f4d 1256 $RATIONAL_FORMAT => 'rational: an integer or a fraction',
1a9d544b 1257 $STRING_FORMAT => 'string',
d11155ec 1258 $ADJUST_FORMAT => 'some entries need adjustment',
24303724 1259 $HEX_ADJUST_FORMAT => 'mapped value in hex; some entries need adjustment',
92f9d56c 1260 $DECOMP_STRING_FORMAT => 'Perl\'s internal (Normalize.pm) decomposition mapping',
c3ff2976 1261 $STRING_WHITE_SPACE_LIST => 'string, but some elements are interpreted as a list; white space occurs only as list item separators'
99870f4d
KW
1262);
1263
1264# Unicode didn't put such derived files in a separate directory at first.
1265my $EXTRACTED_DIR = (-d 'extracted') ? 'extracted' : "";
1266my $EXTRACTED = ($EXTRACTED_DIR) ? "$EXTRACTED_DIR/" : "";
1267my $AUXILIARY = 'auxiliary';
1268
3854b4b8
KW
1269# Hashes and arrays that will eventually go into Heavy.pl for the use of
1270# utf8_heavy.pl and into UCD.pl for the use of UCD.pm
99870f4d
KW
1271my %loose_to_file_of; # loosely maps table names to their respective
1272 # files
1273my %stricter_to_file_of; # same; but for stricter mapping.
315bfd4e 1274my %loose_property_to_file_of; # Maps a loose property name to its map file
e1dc048d 1275my %strict_property_to_file_of; # Same, but strict
3854b4b8
KW
1276my @inline_definitions = "V0"; # Each element gives a definition of a unique
1277 # inversion list. When a definition is inlined,
1278 # its value in the hash it's in (one of the two
1279 # defined just above) will include an index into
1280 # this array. The 0th element is initialized to
9be22ca5 1281 # the definition for a zero length inversion list
89cf10cc
KW
1282my %file_to_swash_name; # Maps the file name to its corresponding key name
1283 # in the hash %utf8::SwashInfo
99870f4d
KW
1284my %nv_floating_to_rational; # maps numeric values floating point numbers to
1285 # their rational equivalent
c12f2655
KW
1286my %loose_property_name_of; # Loosely maps (non_string) property names to
1287 # standard form
e1dc048d
KW
1288my %strict_property_name_of; # Strictly maps (non_string) property names to
1289 # standard form
86a52d1e 1290my %string_property_loose_to_name; # Same, for string properties.
c15fda25
KW
1291my %loose_defaults; # keys are of form "prop=value", where 'prop' is
1292 # the property name in standard loose form, and
1293 # 'value' is the default value for that property,
1294 # also in standard loose form.
9e4a1e86
KW
1295my %loose_to_standard_value; # loosely maps table names to the canonical
1296 # alias for them
2df7880f
KW
1297my %ambiguous_names; # keys are alias names (in standard form) that
1298 # have more than one possible meaning.
d963b40d
KW
1299my %combination_property; # keys are alias names (in standard form) that
1300 # have both a map table, and a binary one that
1301 # yields true for all non-null maps.
5d1df013
KW
1302my %prop_aliases; # Keys are standard property name; values are each
1303 # one's aliases
1e863613
KW
1304my %prop_value_aliases; # Keys of top level are standard property name;
1305 # values are keys to another hash, Each one is
1306 # one of the property's values, in standard form.
1307 # The values are that prop-val's aliases.
b9e57660 1308my %skipped_files; # List of files that we skip
2df7880f 1309my %ucd_pod; # Holds entries that will go into the UCD section of the pod
99870f4d 1310
d867ccfb
KW
1311# Most properties are immune to caseless matching, otherwise you would get
1312# nonsensical results, as properties are a function of a code point, not
1313# everything that is caselessly equivalent to that code point. For example,
1314# Changes_When_Case_Folded('s') should be false, whereas caselessly it would
1315# be true because 's' and 'S' are equivalent caselessly. However,
1316# traditionally, [:upper:] and [:lower:] are equivalent caselessly, so we
1317# extend that concept to those very few properties that are like this. Each
1318# such property will match the full range caselessly. They are hard-coded in
1319# the program; it's not worth trying to make it general as it's extremely
1320# unlikely that they will ever change.
1321my %caseless_equivalent_to;
1322
99870f4d
KW
1323# These constants names and values were taken from the Unicode standard,
1324# version 5.1, section 3.12. They are used in conjunction with Hangul
6e5a209b
KW
1325# syllables. The '_string' versions are so generated tables can retain the
1326# hex format, which is the more familiar value
1327my $SBase_string = "0xAC00";
1328my $SBase = CORE::hex $SBase_string;
1329my $LBase_string = "0x1100";
1330my $LBase = CORE::hex $LBase_string;
1331my $VBase_string = "0x1161";
1332my $VBase = CORE::hex $VBase_string;
1333my $TBase_string = "0x11A7";
1334my $TBase = CORE::hex $TBase_string;
99870f4d
KW
1335my $SCount = 11172;
1336my $LCount = 19;
1337my $VCount = 21;
1338my $TCount = 28;
1339my $NCount = $VCount * $TCount;
1340
1341# For Hangul syllables; These store the numbers from Jamo.txt in conjunction
1342# with the above published constants.
1343my %Jamo;
1344my %Jamo_L; # Leading consonants
1345my %Jamo_V; # Vowels
1346my %Jamo_T; # Trailing consonants
1347
bb1dd3da
KW
1348# For code points whose name contains its ordinal as a '-ABCD' suffix.
1349# The key is the base name of the code point, and the value is an
1350# array giving all the ranges that use this base name. Each range
1351# is actually a hash giving the 'low' and 'high' values of it.
1352my %names_ending_in_code_point;
1353my %loose_names_ending_in_code_point; # Same as above, but has blanks, dashes
1354 # removed from the names
1355# Inverse mapping. The list of ranges that have these kinds of
1356# names. Each element contains the low, high, and base names in an
1357# anonymous hash.
1358my @code_points_ending_in_code_point;
1359
6b5ab373
KW
1360# To hold Unicode's normalization test suite
1361my @normalization_tests;
1362
bb1dd3da
KW
1363# Boolean: does this Unicode version have the hangul syllables, and are we
1364# writing out a table for them?
1365my $has_hangul_syllables = 0;
1366
1367# Does this Unicode version have code points whose names end in their
1368# respective code points, and are we writing out a table for them? 0 for no;
1369# otherwise points to first property that a table is needed for them, so that
1370# if multiple tables are needed, we don't create duplicates
1371my $needing_code_points_ending_in_code_point = 0;
1372
37e2e78e 1373my @backslash_X_tests; # List of tests read in for testing \X
06ae2722 1374my @SB_tests; # List of tests read in for testing \b{sb}
ae3bb8ea 1375my @WB_tests; # List of tests read in for testing \b{wb}
99870f4d
KW
1376my @unhandled_properties; # Will contain a list of properties found in
1377 # the input that we didn't process.
f86864ac 1378my @match_properties; # Properties that have match tables, to be
99870f4d
KW
1379 # listed in the pod
1380my @map_properties; # Properties that get map files written
1381my @named_sequences; # NamedSequences.txt contents.
1382my %potential_files; # Generated list of all .txt files in the directory
1383 # structure so we can warn if something is being
1384 # ignored.
1385my @files_actually_output; # List of files we generated.
1386my @more_Names; # Some code point names are compound; this is used
1387 # to store the extra components of them.
1388my $MIN_FRACTION_LENGTH = 3; # How many digits of a floating point number at
1389 # the minimum before we consider it equivalent to a
1390 # candidate rational
1391my $MAX_FLOATING_SLOP = 10 ** - $MIN_FRACTION_LENGTH; # And in floating terms
1392
1393# These store references to certain commonly used property objects
3c88a801 1394my $ccc;
99870f4d
KW
1395my $gc;
1396my $perl;
1397my $block;
3e20195b
KW
1398my $perl_charname;
1399my $print;
ac7dbdb6 1400my $All;
bc0c431d 1401my $Assigned; # All assigned characters in this Unicode release
359523e2 1402my $script;
99870f4d
KW
1403
1404# Are there conflicting names because of beginning with 'In_', or 'Is_'
1405my $has_In_conflicts = 0;
1406my $has_Is_conflicts = 0;
1407
1408sub internal_file_to_platform ($) {
1409 # Convert our file paths which have '/' separators to those of the
1410 # platform.
1411
1412 my $file = shift;
1413 return undef unless defined $file;
1414
1415 return File::Spec->join(split '/', $file);
d07a55ed 1416}
5beb625e 1417
99870f4d
KW
1418sub file_exists ($) { # platform independent '-e'. This program internally
1419 # uses slash as a path separator.
1420 my $file = shift;
1421 return 0 if ! defined $file;
1422 return -e internal_file_to_platform($file);
1423}
5beb625e 1424
99870f4d 1425sub objaddr($) {
23e33b60
KW
1426 # Returns the address of the blessed input object.
1427 # It doesn't check for blessedness because that would do a string eval
1428 # every call, and the program is structured so that this is never called
1429 # for a non-blessed object.
99870f4d 1430
23e33b60 1431 no overloading; # If overloaded, numifying below won't work.
99870f4d
KW
1432
1433 # Numifying a ref gives its address.
051df77b 1434 return pack 'J', $_[0];
99870f4d
KW
1435}
1436
558712cf 1437# These are used only if $annotate is true.
c4019d52
KW
1438# The entire range of Unicode characters is examined to populate these
1439# after all the input has been processed. But most can be skipped, as they
1440# have the same descriptive phrases, such as being unassigned
1441my @viacode; # Contains the 1 million character names
1442my @printable; # boolean: And are those characters printable?
1443my @annotate_char_type; # Contains a type of those characters, specifically
1444 # for the purposes of annotation.
1445my $annotate_ranges; # A map of ranges of code points that have the same
98dc9551 1446 # name for the purposes of annotation. They map to the
c4019d52
KW
1447 # upper edge of the range, so that the end point can
1448 # be immediately found. This is used to skip ahead to
1449 # the end of a range, and avoid processing each
1450 # individual code point in it.
1451my $unassigned_sans_noncharacters; # A Range_List of the unassigned
1452 # characters, but excluding those which are
1453 # also noncharacter code points
1454
1455# The annotation types are an extension of the regular range types, though
1456# some of the latter are folded into one. Make the new types negative to
1457# avoid conflicting with the regular types
1458my $SURROGATE_TYPE = -1;
1459my $UNASSIGNED_TYPE = -2;
1460my $PRIVATE_USE_TYPE = -3;
1461my $NONCHARACTER_TYPE = -4;
1462my $CONTROL_TYPE = -5;
2d88a86a
KW
1463my $ABOVE_UNICODE_TYPE = -6;
1464my $UNKNOWN_TYPE = -7; # Used only if there is a bug in this program
c4019d52
KW
1465
1466sub populate_char_info ($) {
558712cf 1467 # Used only with the $annotate option. Populates the arrays with the
c4019d52
KW
1468 # input code point's info that are needed for outputting more detailed
1469 # comments. If calling context wants a return, it is the end point of
1470 # any contiguous range of characters that share essentially the same info
1471
1472 my $i = shift;
1473 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
1474
1475 $viacode[$i] = $perl_charname->value_of($i) || "";
1476
1477 # A character is generally printable if Unicode says it is,
1478 # but below we make sure that most Unicode general category 'C' types
1479 # aren't.
1480 $printable[$i] = $print->contains($i);
1481
1482 $annotate_char_type[$i] = $perl_charname->type_of($i) || 0;
1483
1484 # Only these two regular types are treated specially for annotations
1485 # purposes
1486 $annotate_char_type[$i] = 0 if $annotate_char_type[$i] != $CP_IN_NAME
1487 && $annotate_char_type[$i] != $HANGUL_SYLLABLE;
1488
1489 # Give a generic name to all code points that don't have a real name.
1490 # We output ranges, if applicable, for these. Also calculate the end
1491 # point of the range.
1492 my $end;
1493 if (! $viacode[$i]) {
1d025d66 1494 my $nonchar;
2d88a86a
KW
1495 if ($i > $MAX_UNICODE_CODEPOINT) {
1496 $viacode[$i] = 'Above-Unicode';
1497 $annotate_char_type[$i] = $ABOVE_UNICODE_TYPE;
1498 $printable[$i] = 0;
1499 $end = $MAX_WORKING_CODEPOINT;
1500 }
1501 elsif ($gc-> table('Private_use')->contains($i)) {
c4019d52
KW
1502 $viacode[$i] = 'Private Use';
1503 $annotate_char_type[$i] = $PRIVATE_USE_TYPE;
1504 $printable[$i] = 0;
1505 $end = $gc->table('Private_Use')->containing_range($i)->end;
1506 }
1d025d66
KW
1507 elsif ((defined ($nonchar =
1508 Property::property_ref('Noncharacter_Code_Point'))
1509 && $nonchar->table('Y')->contains($i)))
c4019d52
KW
1510 {
1511 $viacode[$i] = 'Noncharacter';
1512 $annotate_char_type[$i] = $NONCHARACTER_TYPE;
1513 $printable[$i] = 0;
1514 $end = property_ref('Noncharacter_Code_Point')->table('Y')->
1515 containing_range($i)->end;
1516 }
1517 elsif ($gc-> table('Control')->contains($i)) {
c71dea7f 1518 $viacode[$i] = property_ref('Name_Alias')->value_of($i) || 'Control';
c4019d52
KW
1519 $annotate_char_type[$i] = $CONTROL_TYPE;
1520 $printable[$i] = 0;
c4019d52
KW
1521 }
1522 elsif ($gc-> table('Unassigned')->contains($i)) {
c4019d52
KW
1523 $annotate_char_type[$i] = $UNASSIGNED_TYPE;
1524 $printable[$i] = 0;
1d025d66
KW
1525 if ($v_version lt v2.0.0) { # No blocks in earliest releases
1526 $viacode[$i] = 'Unassigned';
1527 $end = $gc-> table('Unassigned')->containing_range($i)->end;
1528 }
1529 else {
1530 $viacode[$i] = 'Unassigned, block=' . $block-> value_of($i);
c4019d52 1531
bf06c733
KW
1532 # Because we name the unassigned by the blocks they are in, it
1533 # can't go past the end of that block, and it also can't go
1534 # past the unassigned range it is in. The special table makes
1535 # sure that the non-characters, which are unassigned, are
1536 # separated out.
1537 $end = min($block->containing_range($i)->end,
1538 $unassigned_sans_noncharacters->
1539 containing_range($i)->end);
1d025d66
KW
1540 }
1541 }
35e2b416 1542 elsif ($perl->table('_Perl_Surrogate')->contains($i)) {
1d025d66
KW
1543 $viacode[$i] = 'Surrogate';
1544 $annotate_char_type[$i] = $SURROGATE_TYPE;
1545 $printable[$i] = 0;
1546 $end = $gc->table('Surrogate')->containing_range($i)->end;
13ca76ff
KW
1547 }
1548 else {
1549 Carp::my_carp_bug("Can't figure out how to annotate "
1550 . sprintf("U+%04X", $i)
1551 . ". Proceeding anyway.");
c4019d52
KW
1552 $viacode[$i] = 'UNKNOWN';
1553 $annotate_char_type[$i] = $UNKNOWN_TYPE;
1554 $printable[$i] = 0;
1555 }
1556 }
1557
1558 # Here, has a name, but if it's one in which the code point number is
1559 # appended to the name, do that.
1560 elsif ($annotate_char_type[$i] == $CP_IN_NAME) {
1561 $viacode[$i] .= sprintf("-%04X", $i);
1562 $end = $perl_charname->containing_range($i)->end;
1563 }
1564
1565 # And here, has a name, but if it's a hangul syllable one, replace it with
1566 # the correct name from the Unicode algorithm
1567 elsif ($annotate_char_type[$i] == $HANGUL_SYLLABLE) {
1568 use integer;
1569 my $SIndex = $i - $SBase;
1570 my $L = $LBase + $SIndex / $NCount;
1571 my $V = $VBase + ($SIndex % $NCount) / $TCount;
1572 my $T = $TBase + $SIndex % $TCount;
1573 $viacode[$i] = "HANGUL SYLLABLE $Jamo{$L}$Jamo{$V}";
1574 $viacode[$i] .= $Jamo{$T} if $T != $TBase;
1575 $end = $perl_charname->containing_range($i)->end;
1576 }
1577
1578 return if ! defined wantarray;
1579 return $i if ! defined $end; # If not a range, return the input
1580
1581 # Save this whole range so can find the end point quickly
1582 $annotate_ranges->add_map($i, $end, $end);
1583
1584 return $end;
1585}
1586
23e33b60
KW
1587# Commented code below should work on Perl 5.8.
1588## This 'require' doesn't necessarily work in miniperl, and even if it does,
1589## the native perl version of it (which is what would operate under miniperl)
1590## is extremely slow, as it does a string eval every call.
7e017d6d 1591#my $has_fast_scalar_util = $^X !~ /miniperl/
23e33b60
KW
1592# && defined eval "require Scalar::Util";
1593#
1594#sub objaddr($) {
1595# # Returns the address of the blessed input object. Uses the XS version if
1596# # available. It doesn't check for blessedness because that would do a
1597# # string eval every call, and the program is structured so that this is
1598# # never called for a non-blessed object.
1599#
1600# return Scalar::Util::refaddr($_[0]) if $has_fast_scalar_util;
1601#
1602# # Check at least that is a ref.
1603# my $pkg = ref($_[0]) or return undef;
1604#
1605# # Change to a fake package to defeat any overloaded stringify
1606# bless $_[0], 'main::Fake';
1607#
1608# # Numifying a ref gives its address.
051df77b 1609# my $addr = pack 'J', $_[0];
23e33b60
KW
1610#
1611# # Return to original class
1612# bless $_[0], $pkg;
1613# return $addr;
1614#}
1615
99870f4d
KW
1616sub max ($$) {
1617 my $a = shift;
1618 my $b = shift;
1619 return $a if $a >= $b;
1620 return $b;
1621}
1622
1623sub min ($$) {
1624 my $a = shift;
1625 my $b = shift;
1626 return $a if $a <= $b;
1627 return $b;
1628}
1629
1630sub clarify_number ($) {
1631 # This returns the input number with underscores inserted every 3 digits
1632 # in large (5 digits or more) numbers. Input must be entirely digits, not
1633 # checked.
1634
1635 my $number = shift;
1636 my $pos = length($number) - 3;
1637 return $number if $pos <= 1;
1638 while ($pos > 0) {
1639 substr($number, $pos, 0) = '_';
1640 $pos -= 3;
5beb625e 1641 }
99870f4d 1642 return $number;
99598c8c
JH
1643}
1644
731cb813
KW
1645sub clarify_code_point_count ($) {
1646 # This is like clarify_number(), but the input is assumed to be a count of
1647 # code points, rather than a generic number.
1648
2d88a86a
KW
1649 my $append = "";
1650
1651 my $number = shift;
1652 if ($number > $MAX_UNICODE_CODEPOINTS) {
1653 $number -= ($MAX_WORKING_CODEPOINTS - $MAX_UNICODE_CODEPOINTS);
1654 return "All above-Unicode code points" if $number == 0;
1655 $append = " + all above-Unicode code points";
1656 }
1657 return clarify_number($number) . $append;
731cb813 1658}
12ac2576 1659
99870f4d 1660package Carp;
7ebf06b3 1661
99870f4d
KW
1662# These routines give a uniform treatment of messages in this program. They
1663# are placed in the Carp package to cause the stack trace to not include them,
1664# although an alternative would be to use another package and set @CARP_NOT
1665# for it.
12ac2576 1666
99870f4d 1667our $Verbose = 1 if main::DEBUG; # Useful info when debugging
12ac2576 1668
99f78760
KW
1669# This is a work-around suggested by Nicholas Clark to fix a problem with Carp
1670# and overload trying to load Scalar:Util under miniperl. See
1671# http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2009-11/msg01057.html
1672undef $overload::VERSION;
1673
99870f4d
KW
1674sub my_carp {
1675 my $message = shift || "";
1676 my $nofold = shift || 0;
7ebf06b3 1677
99870f4d
KW
1678 if ($message) {
1679 $message = main::join_lines($message);
1680 $message =~ s/^$0: *//; # Remove initial program name
1681 $message =~ s/[.;,]+$//; # Remove certain ending punctuation
1682 $message = "\n$0: $message;";
12ac2576 1683
99870f4d
KW
1684 # Fold the message with program name, semi-colon end punctuation
1685 # (which looks good with the message that carp appends to it), and a
1686 # hanging indent for continuation lines.
1687 $message = main::simple_fold($message, "", 4) unless $nofold;
1688 $message =~ s/\n$//; # Remove the trailing nl so what carp
1689 # appends is to the same line
1690 }
12ac2576 1691
99870f4d 1692 return $message if defined wantarray; # If a caller just wants the msg
12ac2576 1693
99870f4d
KW
1694 carp $message;
1695 return;
1696}
7ebf06b3 1697
99870f4d
KW
1698sub my_carp_bug {
1699 # This is called when it is clear that the problem is caused by a bug in
1700 # this program.
7ebf06b3 1701
99870f4d
KW
1702 my $message = shift;
1703 $message =~ s/^$0: *//;
1704 $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");
1705 carp $message;
1706 return;
1707}
7ebf06b3 1708
99870f4d
KW
1709sub carp_too_few_args {
1710 if (@_ != 2) {
1711 my_carp_bug("Wrong number of arguments: to 'carp_too_few_arguments'. No action taken.");
1712 return;
12ac2576 1713 }
7ebf06b3 1714
99870f4d
KW
1715 my $args_ref = shift;
1716 my $count = shift;
7ebf06b3 1717
99870f4d
KW
1718 my_carp_bug("Need at least $count arguments to "
1719 . (caller 1)[3]
1720 . ". Instead got: '"
1721 . join ', ', @$args_ref
1722 . "'. No action taken.");
1723 return;
12ac2576
JP
1724}
1725
99870f4d
KW
1726sub carp_extra_args {
1727 my $args_ref = shift;
1728 my_carp_bug("Too many arguments to 'carp_extra_args': (" . join(', ', @_) . "); Extras ignored.") if @_;
12ac2576 1729
99870f4d
KW
1730 unless (ref $args_ref) {
1731 my_carp_bug("Argument to 'carp_extra_args' ($args_ref) must be a ref. Not checking arguments.");
1732 return;
1733 }
1734 my ($package, $file, $line) = caller;
1735 my $subroutine = (caller 1)[3];
cf25bb62 1736
99870f4d
KW
1737 my $list;
1738 if (ref $args_ref eq 'HASH') {
1739 foreach my $key (keys %$args_ref) {
1740 $args_ref->{$key} = $UNDEF unless defined $args_ref->{$key};
cf25bb62 1741 }
99870f4d 1742 $list = join ', ', each %{$args_ref};
cf25bb62 1743 }
99870f4d
KW
1744 elsif (ref $args_ref eq 'ARRAY') {
1745 foreach my $arg (@$args_ref) {
1746 $arg = $UNDEF unless defined $arg;
1747 }
1748 $list = join ', ', @$args_ref;
1749 }
1750 else {
1751 my_carp_bug("Can't cope with ref "
1752 . ref($args_ref)
1753 . " . argument to 'carp_extra_args'. Not checking arguments.");
1754 return;
1755 }
1756
1757 my_carp_bug("Unrecognized parameters in options: '$list' to $subroutine. Skipped.");
1758 return;
d73e5302
JH
1759}
1760
99870f4d
KW
1761package main;
1762
1763{ # Closure
1764
1765 # This program uses the inside-out method for objects, as recommended in
1b6b3fa9
KW
1766 # "Perl Best Practices". (This is the best solution still, since this has
1767 # to run under miniperl.) This closure aids in generating those. There
99870f4d
KW
1768 # are two routines. setup_package() is called once per package to set
1769 # things up, and then set_access() is called for each hash representing a
1770 # field in the object. These routines arrange for the object to be
1771 # properly destroyed when no longer used, and for standard accessor
1772 # functions to be generated. If you need more complex accessors, just
1773 # write your own and leave those accesses out of the call to set_access().
1774 # More details below.
1775
1776 my %constructor_fields; # fields that are to be used in constructors; see
1777 # below
1778
1779 # The values of this hash will be the package names as keys to other
1780 # hashes containing the name of each field in the package as keys, and
1781 # references to their respective hashes as values.
1782 my %package_fields;
1783
1784 sub setup_package {
1785 # Sets up the package, creating standard DESTROY and dump methods
1786 # (unless already defined). The dump method is used in debugging by
1787 # simple_dumper().
1788 # The optional parameters are:
1789 # a) a reference to a hash, that gets populated by later
1790 # set_access() calls with one of the accesses being
1791 # 'constructor'. The caller can then refer to this, but it is
1792 # not otherwise used by these two routines.
1793 # b) a reference to a callback routine to call during destruction
1794 # of the object, before any fields are actually destroyed
1795
1796 my %args = @_;
1797 my $constructor_ref = delete $args{'Constructor_Fields'};
1798 my $destroy_callback = delete $args{'Destroy_Callback'};
1799 Carp::carp_extra_args(\@_) if main::DEBUG && %args;
1800
1801 my %fields;
1802 my $package = (caller)[0];
1803
1804 $package_fields{$package} = \%fields;
1805 $constructor_fields{$package} = $constructor_ref;
1806
1807 unless ($package->can('DESTROY')) {
1808 my $destroy_name = "${package}::DESTROY";
1809 no strict "refs";
1810
1811 # Use typeglob to give the anonymous subroutine the name we want
1812 *$destroy_name = sub {
1813 my $self = shift;
ffe43484 1814 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
1815
1816 $self->$destroy_callback if $destroy_callback;
1817 foreach my $field (keys %{$package_fields{$package}}) {
1818 #print STDERR __LINE__, ": Destroying ", ref $self, " ", sprintf("%04X", $addr), ": ", $field, "\n";
1819 delete $package_fields{$package}{$field}{$addr};
1820 }
1821 return;
1822 }
1823 }
1824
1825 unless ($package->can('dump')) {
1826 my $dump_name = "${package}::dump";
1827 no strict "refs";
1828 *$dump_name = sub {
1829 my $self = shift;
1830 return dump_inside_out($self, $package_fields{$package}, @_);
1831 }
1832 }
1833 return;
1834 }
1835
1836 sub set_access {
1837 # Arrange for the input field to be garbage collected when no longer
1838 # needed. Also, creates standard accessor functions for the field
1839 # based on the optional parameters-- none if none of these parameters:
1840 # 'addable' creates an 'add_NAME()' accessor function.
1841 # 'readable' or 'readable_array' creates a 'NAME()' accessor
1842 # function.
1843 # 'settable' creates a 'set_NAME()' accessor function.
1844 # 'constructor' doesn't create an accessor function, but adds the
1845 # field to the hash that was previously passed to
1846 # setup_package();
1847 # Any of the accesses can be abbreviated down, so that 'a', 'ad',
1848 # 'add' etc. all mean 'addable'.
1849 # The read accessor function will work on both array and scalar
1850 # values. If another accessor in the parameter list is 'a', the read
1851 # access assumes an array. You can also force it to be array access
1852 # by specifying 'readable_array' instead of 'readable'
1853 #
1854 # A sort-of 'protected' access can be set-up by preceding the addable,
1855 # readable or settable with some initial portion of 'protected_' (but,
1856 # the underscore is required), like 'p_a', 'pro_set', etc. The
1857 # "protection" is only by convention. All that happens is that the
1858 # accessor functions' names begin with an underscore. So instead of
1859 # calling set_foo, the call is _set_foo. (Real protection could be
c1739a4a 1860 # accomplished by having a new subroutine, end_package, called at the
99870f4d
KW
1861 # end of each package, and then storing the __LINE__ ranges and
1862 # checking them on every accessor. But that is way overkill.)
1863
1864 # We create anonymous subroutines as the accessors and then use
1865 # typeglobs to assign them to the proper package and name
1866
1867 my $name = shift; # Name of the field
1868 my $field = shift; # Reference to the inside-out hash containing the
1869 # field
1870
1871 my $package = (caller)[0];
1872
1873 if (! exists $package_fields{$package}) {
1874 croak "$0: Must call 'setup_package' before 'set_access'";
1875 }
d73e5302 1876
99870f4d
KW
1877 # Stash the field so DESTROY can get it.
1878 $package_fields{$package}{$name} = $field;
cf25bb62 1879
99870f4d
KW
1880 # Remaining arguments are the accessors. For each...
1881 foreach my $access (@_) {
1882 my $access = lc $access;
cf25bb62 1883
99870f4d 1884 my $protected = "";
cf25bb62 1885
99870f4d
KW
1886 # Match the input as far as it goes.
1887 if ($access =~ /^(p[^_]*)_/) {
1888 $protected = $1;
1889 if (substr('protected_', 0, length $protected)
1890 eq $protected)
1891 {
1892
1893 # Add 1 for the underscore not included in $protected
1894 $access = substr($access, length($protected) + 1);
1895 $protected = '_';
1896 }
1897 else {
1898 $protected = "";
1899 }
1900 }
1901
1902 if (substr('addable', 0, length $access) eq $access) {
1903 my $subname = "${package}::${protected}add_$name";
1904 no strict "refs";
1905
1906 # add_ accessor. Don't add if already there, which we
1907 # determine using 'eq' for scalars and '==' otherwise.
1908 *$subname = sub {
1909 use strict "refs";
1910 return Carp::carp_too_few_args(\@_, 2) if main::DEBUG && @_ < 2;
1911 my $self = shift;
1912 my $value = shift;
ffe43484 1913 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
1914 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
1915 if (ref $value) {
f998e60c 1916 return if grep { $value == $_ } @{$field->{$addr}};
99870f4d
KW
1917 }
1918 else {
f998e60c 1919 return if grep { $value eq $_ } @{$field->{$addr}};
99870f4d 1920 }
f998e60c 1921 push @{$field->{$addr}}, $value;
99870f4d
KW
1922 return;
1923 }
1924 }
1925 elsif (substr('constructor', 0, length $access) eq $access) {
1926 if ($protected) {
1927 Carp::my_carp_bug("Can't set-up 'protected' constructors")
1928 }
1929 else {
1930 $constructor_fields{$package}{$name} = $field;
1931 }
1932 }
1933 elsif (substr('readable_array', 0, length $access) eq $access) {
1934
1935 # Here has read access. If one of the other parameters for
1936 # access is array, or this one specifies array (by being more
1937 # than just 'readable_'), then create a subroutine that
1938 # assumes the data is an array. Otherwise just a scalar
1939 my $subname = "${package}::${protected}$name";
1940 if (grep { /^a/i } @_
1941 or length($access) > length('readable_'))
1942 {
1943 no strict "refs";
1944 *$subname = sub {
1945 use strict "refs";
23e33b60 1946 Carp::carp_extra_args(\@_) if main::DEBUG && @_ > 1;
ffe43484 1947 my $addr = do { no overloading; pack 'J', $_[0]; };
99870f4d
KW
1948 if (ref $field->{$addr} ne 'ARRAY') {
1949 my $type = ref $field->{$addr};
1950 $type = 'scalar' unless $type;
1951 Carp::my_carp_bug("Trying to read $name as an array when it is a $type. Big problems.");
1952 return;
1953 }
1954 return scalar @{$field->{$addr}} unless wantarray;
1955
1956 # Make a copy; had problems with caller modifying the
1957 # original otherwise
1958 my @return = @{$field->{$addr}};
1959 return @return;
1960 }
1961 }
1962 else {
1963
1964 # Here not an array value, a simpler function.
1965 no strict "refs";
1966 *$subname = sub {
1967 use strict "refs";
23e33b60 1968 Carp::carp_extra_args(\@_) if main::DEBUG && @_ > 1;
f998e60c 1969 no overloading;
051df77b 1970 return $field->{pack 'J', $_[0]};
99870f4d
KW
1971 }
1972 }
1973 }
1974 elsif (substr('settable', 0, length $access) eq $access) {
1975 my $subname = "${package}::${protected}set_$name";
1976 no strict "refs";
1977 *$subname = sub {
1978 use strict "refs";
23e33b60
KW
1979 if (main::DEBUG) {
1980 return Carp::carp_too_few_args(\@_, 2) if @_ < 2;
1981 Carp::carp_extra_args(\@_) if @_ > 2;
1982 }
1983 # $self is $_[0]; $value is $_[1]
f998e60c 1984 no overloading;
051df77b 1985 $field->{pack 'J', $_[0]} = $_[1];
99870f4d
KW
1986 return;
1987 }
1988 }
1989 else {
1990 Carp::my_carp_bug("Unknown accessor type $access. No accessor set.");
1991 }
cf25bb62 1992 }
99870f4d 1993 return;
cf25bb62 1994 }
99870f4d
KW
1995}
1996
1997package Input_file;
1998
1999# All input files use this object, which stores various attributes about them,
2000# and provides for convenient, uniform handling. The run method wraps the
2001# processing. It handles all the bookkeeping of opening, reading, and closing
2002# the file, returning only significant input lines.
2003#
2004# Each object gets a handler which processes the body of the file, and is
74cd47d0
KW
2005# called by run(). All character property files must use the generic,
2006# default handler, which has code scrubbed to handle things you might not
2007# expect, including automatic EBCDIC handling. For files that don't deal with
2008# mapping code points to a property value, such as test files,
2009# PropertyAliases, PropValueAliases, and named sequences, you can override the
2010# handler to be a custom one. Such a handler should basically be a
2011# while(next_line()) {...} loop.
99870f4d
KW
2012#
2013# You can also set up handlers to
40045c2d 2014# 0) call during object construction time, after everything else is done
537124e4 2015# 1) call before the first line is read, for pre processing
83b68635
KW
2016# 2) call to adjust each line of the input before the main handler gets
2017# them. This can be automatically generated, if appropriately simple
2018# enough, by specifiying a Properties parameter in the constructor.
99870f4d 2019# 3) call upon EOF before the main handler exits its loop
537124e4 2020# 4) call at the end, for post processing
99870f4d
KW
2021#
2022# $_ is used to store the input line, and is to be filtered by the
2023# each_line_handler()s. So, if the format of the line is not in the desired
2024# format for the main handler, these are used to do that adjusting. They can
2025# be stacked (by enclosing them in an [ anonymous array ] in the constructor,
3848e5b7
KW
2026# so the $_ output of one is used as the input to the next. The eof handler
2027# is also stackable, but none of the others are, but could easily be changed
2028# to be so.
99870f4d
KW
2029#
2030# Most of the handlers can call insert_lines() or insert_adjusted_lines()
2031# which insert the parameters as lines to be processed before the next input
3848e5b7 2032# file line is read. This allows the EOF handler(s) to flush buffers, for
99870f4d
KW
2033# example. The difference between the two routines is that the lines inserted
2034# by insert_lines() are subjected to the each_line_handler()s. (So if you
9be22ca5
KW
2035# called it from such a handler, you would get infinite recursion without some
2036# mechanism to prevent that.) Lines inserted by insert_adjusted_lines() go
2037# directly to the main handler without any adjustments. If the
2038# post-processing handler calls any of these, there will be no effect. Some
2039# error checking for these conditions could be added, but it hasn't been done.
99870f4d
KW
2040#
2041# carp_bad_line() should be called to warn of bad input lines, which clears $_
2042# to prevent further processing of the line. This routine will output the
2043# message as a warning once, and then keep a count of the lines that have the
2044# same message, and output that count at the end of the file's processing.
2045# This keeps the number of messages down to a manageable amount.
2046#
2047# get_missings() should be called to retrieve any @missing input lines.
2048# Messages will be raised if this isn't done if the options aren't to ignore
2049# missings.
2050
2051sub trace { return main::trace(@_); }
2052
99870f4d
KW
2053{ # Closure
2054 # Keep track of fields that are to be put into the constructor.
2055 my %constructor_fields;
2056
2057 main::setup_package(Constructor_Fields => \%constructor_fields);
2058
2059 my %file; # Input file name, required
2060 main::set_access('file', \%file, qw{ c r });
2061
2062 my %first_released; # Unicode version file was first released in, required
2063 main::set_access('first_released', \%first_released, qw{ c r });
2064
2065 my %handler; # Subroutine to process the input file, defaults to
2066 # 'process_generic_property_file'
2067 main::set_access('handler', \%handler, qw{ c });
2068
2069 my %property;
2070 # name of property this file is for. defaults to none, meaning not
2071 # applicable, or is otherwise determinable, for example, from each line.
696609bf 2072 main::set_access('property', \%property, qw{ c r });
99870f4d
KW
2073
2074 my %optional;
3274d896
KW
2075 # This is either an unsigned number, or a list of property names. In the
2076 # former case, if it is non-zero, it means the file is optional, so if the
2077 # file is absent, no warning about that is output. In the latter case, it
2078 # is a list of properties that the file (exclusively) defines. If the
2079 # file is present, tables for those properties will be produced; if
2080 # absent, none will, even if they are listed elsewhere (namely
2081 # PropertyAliases.txt and PropValueAliases.txt) as being in this release,
2082 # and no warnings will be raised about them not being available. (And no
2083 # warning about the file itself will be raised.)
2084 main::set_access('optional', \%optional, qw{ c readable_array } );
99870f4d
KW
2085
2086 my %non_skip;
2087 # This is used for debugging, to skip processing of all but a few input
2088 # files. Add 'non_skip => 1' to the constructor for those files you want
2089 # processed when you set the $debug_skip global.
2090 main::set_access('non_skip', \%non_skip, 'c');
2091
37e2e78e 2092 my %skip;
0dee21d6
KW
2093 # This is used to skip processing of this input file (semi-) permanently.
2094 # The value should be the reason the file is being skipped. It is used
2095 # for files that we aren't planning to process anytime soon, but want to
2096 # allow to be in the directory and be checked for their names not
2097 # conflicting with any other files on a DOS 8.3 name filesystem, but to
2098 # not otherwise be processed, and to not raise a warning about not being
2099 # handled. In the constructor call, any value that evaluates to a numeric
2100 # 0 or undef means don't skip. Any other value is a string giving the
2101 # reason it is being skippped, and this will appear in generated pod.
2102 # However, an empty string reason will suppress the pod entry.
2103 # Internally, calls that evaluate to numeric 0 are changed into undef to
2104 # distinguish them from an empty string call.
2105 main::set_access('skip', \%skip, 'c', 'r');
37e2e78e 2106
99870f4d
KW
2107 my %each_line_handler;
2108 # list of subroutines to look at and filter each non-comment line in the
2109 # file. defaults to none. The subroutines are called in order, each is
2110 # to adjust $_ for the next one, and the final one adjusts it for
2111 # 'handler'
2112 main::set_access('each_line_handler', \%each_line_handler, 'c');
2113
83b68635
KW
2114 my %properties; # Optional ordered list of the properties that occur in each
2115 # meaningful line of the input file. If present, an appropriate
2116 # each_line_handler() is automatically generated and pushed onto the stack
2117 # of such handlers. This is useful when a file contains multiple
2118 # proerties per line, but no other special considerations are necessary.
2119 # The special value "<ignored>" means to discard the corresponding input
2120 # field.
2121 # Any @missing lines in the file should also match this syntax; no such
2122 # files exist as of 6.3. But if it happens in a future release, the code
2123 # could be expanded to properly parse them.
2124 main::set_access('properties', \%properties, qw{ c r });
2125
99870f4d
KW
2126 my %has_missings_defaults;
2127 # ? Are there lines in the file giving default values for code points
2128 # missing from it?. Defaults to NO_DEFAULTS. Otherwise NOT_IGNORED is
2129 # the norm, but IGNORED means it has such lines, but the handler doesn't
2130 # use them. Having these three states allows us to catch changes to the
83b68635
KW
2131 # UCD that this program should track. XXX This could be expanded to
2132 # specify the syntax for such lines, like %properties above.
99870f4d
KW
2133 main::set_access('has_missings_defaults',
2134 \%has_missings_defaults, qw{ c r });
2135
40045c2d
KW
2136 my %construction_time_handler;
2137 # Subroutine to call at the end of the new method. If undef, no such
2138 # handler is called.
2139 main::set_access('construction_time_handler',
2140 \%construction_time_handler, qw{ c });
2141
99870f4d
KW
2142 my %pre_handler;
2143 # Subroutine to call before doing anything else in the file. If undef, no
2144 # such handler is called.
2145 main::set_access('pre_handler', \%pre_handler, qw{ c });
2146
2147 my %eof_handler;
3848e5b7 2148 # Subroutines to call upon getting an EOF on the input file, but before
99870f4d
KW
2149 # that is returned to the main handler. This is to allow buffers to be
2150 # flushed. The handler is expected to call insert_lines() or
2151 # insert_adjusted() with the buffered material
3848e5b7 2152 main::set_access('eof_handler', \%eof_handler, qw{ c });
99870f4d
KW
2153
2154 my %post_handler;
2155 # Subroutine to call after all the lines of the file are read in and
9be22ca5
KW
2156 # processed. If undef, no such handler is called. Note that this cannot
2157 # add lines to be processed; instead use eof_handler
99870f4d
KW
2158 main::set_access('post_handler', \%post_handler, qw{ c });
2159
2160 my %progress_message;
2161 # Message to print to display progress in lieu of the standard one
2162 main::set_access('progress_message', \%progress_message, qw{ c });
2163
2164 my %handle;
2165 # cache open file handle, internal. Is undef if file hasn't been
2166 # processed at all, empty if has;
2167 main::set_access('handle', \%handle);
2168
2169 my %added_lines;
2170 # cache of lines added virtually to the file, internal
2171 main::set_access('added_lines', \%added_lines);
2172
74cd47d0
KW
2173 my %remapped_lines;
2174 # cache of lines added virtually to the file, internal
2175 main::set_access('remapped_lines', \%remapped_lines);
2176
99870f4d
KW
2177 my %errors;
2178 # cache of errors found, internal
2179 main::set_access('errors', \%errors);
2180
2181 my %missings;
2182 # storage of '@missing' defaults lines
2183 main::set_access('missings', \%missings);
2184
0dee21d6
KW
2185 my %required_even_in_debug_skip;
2186 # debug_skip is used to speed up compilation during debugging by skipping
2187 # processing files that are not needed for the task at hand. However,
2188 # some files pretty much can never be skipped, and this is used to specify
2189 # that this is one of them. In order to skip this file, the call to the
2190 # constructor must be edited to comment out this parameter.
2191 main::set_access('required_even_in_debug_skip',
2192 \%required_even_in_debug_skip, 'c');
2193
d8062573
KW
2194 my %withdrawn;
2195 # Some files get removed from the Unicode DB. This is a version object
2196 # giving the first release without this file.
2197 main::set_access('withdrawn', \%withdrawn, 'c');
2198
0dee21d6
KW
2199 my %in_this_release;
2200 # Calculated value from %first_released and %withdrawn. Are we compiling
2201 # a Unicode release which includes this file?
2202 main::set_access('in_this_release', \%in_this_release);
2203
74cd47d0
KW
2204 sub _next_line;
2205 sub _next_line_with_remapped_range;
2206
99870f4d
KW
2207 sub new {
2208 my $class = shift;
2209
2210 my $self = bless \do{ my $anonymous_scalar }, $class;
ffe43484 2211 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
2212
2213 # Set defaults
2214 $handler{$addr} = \&main::process_generic_property_file;
2215 $non_skip{$addr} = 0;
0dee21d6 2216 $skip{$addr} = undef;
99870f4d
KW
2217 $has_missings_defaults{$addr} = $NO_DEFAULTS;
2218 $handle{$addr} = undef;
2219 $added_lines{$addr} = [ ];
74cd47d0 2220 $remapped_lines{$addr} = [ ];
99870f4d 2221 $each_line_handler{$addr} = [ ];
3848e5b7 2222 $eof_handler{$addr} = [ ];
99870f4d
KW
2223 $errors{$addr} = { };
2224 $missings{$addr} = [ ];
3274d896 2225 $optional{$addr} = [ ];
99870f4d
KW
2226
2227 # Two positional parameters.
99f78760 2228 return Carp::carp_too_few_args(\@_, 2) if main::DEBUG && @_ < 2;
99870f4d
KW
2229 $file{$addr} = main::internal_file_to_platform(shift);
2230 $first_released{$addr} = shift;
2231
2232 # The rest of the arguments are key => value pairs
2233 # %constructor_fields has been set up earlier to list all possible
2234 # ones. Either set or push, depending on how the default has been set
2235 # up just above.
2236 my %args = @_;
2237 foreach my $key (keys %args) {
2238 my $argument = $args{$key};
2239
2240 # Note that the fields are the lower case of the constructor keys
2241 my $hash = $constructor_fields{lc $key};
2242 if (! defined $hash) {
2243 Carp::my_carp_bug("Unrecognized parameters '$key => $argument' to new() for $self. Skipped");
2244 next;
2245 }
2246 if (ref $hash->{$addr} eq 'ARRAY') {
2247 if (ref $argument eq 'ARRAY') {
2248 foreach my $argument (@{$argument}) {
2249 next if ! defined $argument;
2250 push @{$hash->{$addr}}, $argument;
2251 }
2252 }
2253 else {
2254 push @{$hash->{$addr}}, $argument if defined $argument;
2255 }
2256 }
2257 else {
2258 $hash->{$addr} = $argument;
2259 }
2260 delete $args{$key};
2261 };
2262
0dee21d6
KW
2263 $non_skip{$addr} = 1 if $required_even_in_debug_skip{$addr};
2264
2265 # Convert 0 (meaning don't skip) to undef
2266 undef $skip{$addr} unless $skip{$addr};
2267
3274d896
KW
2268 # Handle the case where this file is optional
2269 my $pod_message_for_non_existent_optional = "";
2270 if ($optional{$addr}->@*) {
2271
2272 # First element is the pod message
2273 $pod_message_for_non_existent_optional
2274 = shift $optional{$addr}->@*;
2275 # Convert a 0 'Optional' argument to an empty list to make later
2276 # code more concise.
2277 if ( $optional{$addr}->@*
2278 && $optional{$addr}->@* == 1
2279 && $optional{$addr}[0] ne ""
2280 && $optional{$addr}[0] !~ /\D/
2281 && $optional{$addr}[0] == 0)
2282 {
2283 $optional{$addr} = [ ];
2284 }
2285 else { # But if the only element doesn't evaluate to 0, make sure
2286 # that this file is indeed considered optional below.
2287 unshift $optional{$addr}->@*, 1;
2288 }
2289 }
2290
0dee21d6
KW
2291 my $progress;
2292
2293 if ($first_released{$addr} le $v_version) {
2294 $progress = $file{$addr};
99870f4d
KW
2295 }
2296
0dee21d6
KW
2297 my $file = $file{$addr};
2298 $progress_message{$addr} = "Processing $progress"
2299 unless $progress_message{$addr};
2300
d8062573
KW
2301 # A file should be there if it is within the window of versions for
2302 # which Unicode supplies it
2303 if ($withdrawn{$addr} && $withdrawn{$addr} le $v_version) {
2304 $in_this_release{$addr} = 0;
2305 $skip{$addr} = "";
2306 }
2307 else {
0dee21d6
KW
2308 $in_this_release{$addr} = $first_released{$addr} le $v_version;
2309
2310 # Check that the file for this object exists
2311 if (! main::file_exists($file))
2312 {
2313 # Here there is nothing available for this release. This is
2314 # fine if we aren't expecting anything in this release.
2315 if (! $in_this_release{$addr}) {
2316 $skip{$addr} = ""; # Don't remark since we expected
2317 # nothing and got nothing
2318 }
3274d896 2319 elsif ($optional{$addr}->@*) {
0dee21d6 2320
3274d896
KW
2321 # Here the file is optional in this release; Use the
2322 # passed in text to document this case in the pod.
2323 $skip{$addr} = $pod_message_for_non_existent_optional;
0dee21d6
KW
2324 }
2325 elsif ( $in_this_release{$addr}
2326 && ! defined $skip{$addr}
2327 && defined $file)
2328 { # Doesn't exist but should.
2329 $skip{$addr} = "'$file' not found. Possibly Big problems";
2330 Carp::my_carp($skip{$addr});
2331 }
2332 }
2333 elsif ($debug_skip && ! defined $skip{$addr} && ! $non_skip{$addr})
2334 {
2335
2336 # The file exists; if not skipped for another reason, and we are
2337 # skipping most everything during debugging builds, use that as
2338 # the skip reason.
2339 $skip{$addr} = '$debug_skip is on'
2340 }
d8062573 2341 }
0dee21d6
KW
2342
2343 if ( ! $debug_skip
2344 && $non_skip{$addr}
2345 && ! $required_even_in_debug_skip{$addr}
2346 && $verbosity)
2347 {
2348 print "Warning: " . __PACKAGE__ . " constructor for $file has useless 'non_skip' in it\n";
a3a8c5f0 2349 }
99870f4d 2350
0dee21d6 2351 # Here, we have figured out if we will be skipping this file or not.
3274d896
KW
2352 # If so, we add any single property it defines to any passed in
2353 # optional property list. These will be dealt with at run time.
0dee21d6 2354 if (defined $skip{$addr}) {
3274d896
KW
2355 if ($property{$addr}) {
2356 push $optional{$addr}->@*, $property{$addr};
2357 }
2358 } # Otherwise, are going to process the file.
0dee21d6
KW
2359 elsif ($property{$addr}) {
2360
2361 # If the file has a property defined in the constructor for it, it
2362 # means that the property is not listed in the file's entries. So
2363 # add a handler (to the list of line handlers) to insert the
2364 # property name into the lines, to provide a uniform interface to
2365 # the final processing subroutine.
2366 push @{$each_line_handler{$addr}}, \&_insert_property_into_line;
09ca89ce 2367 }
83b68635
KW
2368 elsif ($properties{$addr}) {
2369
9be22ca5
KW
2370 # Similarly, there may be more than one property represented on
2371 # each line, with no clue but the constructor input what those
2372 # might be. Add a handler for each line in the input so that it
2373 # creates a separate input line for each property in those input
2374 # lines, thus making them suitable to handle generically.
83b68635
KW
2375
2376 push @{$each_line_handler{$addr}},
2377 sub {
2378 my $file = shift;
2379 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2380
2381 my @fields = split /\s*;\s*/, $_, -1;
2382
2383 if (@fields - 1 > @{$properties{$addr}}) {
2384 $file->carp_bad_line('Extra fields');
2385 $_ = "";
2386 return;
2387 }
2388 my $range = shift @fields; # 0th element is always the
2389 # range
2390
2391 # The next fields in the input line correspond
2392 # respectively to the stored properties.
2393 for my $i (0 .. @{$properties{$addr}} - 1) {
2394 my $property_name = $properties{$addr}[$i];
2395 next if $property_name eq '<ignored>';
2396 $file->insert_adjusted_lines(
2397 "$range; $property_name; $fields[$i]");
2398 }
2399 $_ = "";
2400
2401 return;
2402 };
2403 }
37e2e78e 2404
9be22ca5 2405 { # On non-ascii platforms, we use a special pre-handler
74cd47d0
KW
2406 no strict;
2407 no warnings 'once';
2408 *next_line = (main::NON_ASCII_PLATFORM)
2409 ? *_next_line_with_remapped_range
2410 : *_next_line;
2411 }
2412
40045c2d
KW
2413 &{$construction_time_handler{$addr}}($self)
2414 if $construction_time_handler{$addr};
2415
99870f4d 2416 return $self;
d73e5302
JH
2417 }
2418
cf25bb62 2419
99870f4d
KW
2420 use overload
2421 fallback => 0,
2422 qw("") => "_operator_stringify",
2423 "." => \&main::_operator_dot,
1285127e 2424 ".=" => \&main::_operator_dot_equal,
99870f4d 2425 ;
cf25bb62 2426
99870f4d
KW
2427 sub _operator_stringify {
2428 my $self = shift;
cf25bb62 2429
99870f4d 2430 return __PACKAGE__ . " object for " . $self->file;
d73e5302 2431 }
d73e5302 2432
99870f4d
KW
2433 sub run {
2434 # Process the input object $self. This opens and closes the file and
2435 # calls all the handlers for it. Currently, this can only be called
3848e5b7 2436 # once per file, as it destroy's the EOF handlers
d73e5302 2437
0dee21d6
KW
2438 # flag to make sure extracted files are processed early
2439 state $seen_non_extracted_non_age = 0;
2440
99870f4d
KW
2441 my $self = shift;
2442 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
b6922eda 2443
ffe43484 2444 my $addr = do { no overloading; pack 'J', $self; };
b6922eda 2445
99870f4d 2446 my $file = $file{$addr};
d73e5302 2447
0dee21d6
KW
2448 if (! $file) {
2449 $handle{$addr} = 'pretend_is_open';
99870f4d
KW
2450 }
2451 else {
99870f4d 2452 if ($seen_non_extracted_non_age) {
9be22ca5
KW
2453 if ($file =~ /$EXTRACTED/i) # Some platforms may change the
2454 # case of the file's name
2455 {
1675ea0d 2456 Carp::my_carp_bug(main::join_lines(<<END
99f78760 2457$file should be processed just after the 'Prop...Alias' files, and before
99870f4d
KW
2458anything not in the $EXTRACTED_DIR directory. Proceeding, but the results may
2459have subtle problems
2460END
2461 ));
2462 }
2463 }
2464 elsif ($EXTRACTED_DIR
0dee21d6
KW
2465
2466 # We only do this check for generic property files
2467 && $handler{$addr} == \&main::process_generic_property_file
2468
517956bf
CB
2469 && $file !~ /$EXTRACTED/i
2470 && lc($file) ne 'dage.txt')
99870f4d
KW
2471 {
2472 # We don't set this (by the 'if' above) if we have no
2473 # extracted directory, so if running on an early version,
2474 # this test won't work. Not worth worrying about.
2475 $seen_non_extracted_non_age = 1;
2476 }
2477
9be22ca5 2478 # Mark the file as having being processed, and warn if it
99870f4d
KW
2479 # isn't a file we are expecting. As we process the files,
2480 # they are deleted from the hash, so any that remain at the
2481 # end of the program are files that we didn't process.
517956bf 2482 my $fkey = File::Spec->rel2abs($file);
0dee21d6 2483 my $exists = delete $potential_files{lc($fkey)};
faf3cf6b 2484
0dee21d6
KW
2485 Carp::my_carp("Was not expecting '$file'.")
2486 if $exists && ! $in_this_release{$addr};
2487
2488 # We may be skipping this file ...
2489 if (defined $skip{$addr}) {
2490
2491 # If the file isn't supposed to be in this release, there is
2492 # nothing to do
2493 if ($in_this_release{$addr}) {
2494
2495 # But otherwise, we may print a message
2496 if ($debug_skip) {
2497 print STDERR "Skipping input file '$file'",
2498 " because '$skip{$addr}'\n";
2499 }
2500
2501 # And add it to the list of skipped files, which is later
2502 # used to make the pod
2503 $skipped_files{$file} = $skip{$addr};
3274d896
KW
2504
2505 # The 'optional' list contains properties that are also to
2506 # be skipped along with the file. (There may also be
2507 # digits which are just placeholders to make sure it isn't
2508 # an empty list
2509 foreach my $property ($optional{$addr}->@*) {
2510 next unless $property =~ /\D/;
2511 my $prop_object = main::property_ref($property);
2512 next unless defined $prop_object;
2513 $prop_object->set_fate($SUPPRESSED, $skip{$addr});
2514 }
0dee21d6 2515 }
99870f4d 2516
37e2e78e
KW
2517 return;
2518 }
2519
9be22ca5
KW
2520 # Here, we are going to process the file. Open it, converting the
2521 # slashes used in this program into the proper form for the OS
99870f4d
KW
2522 my $file_handle;
2523 if (not open $file_handle, "<", $file) {
2524 Carp::my_carp("Can't open $file. Skipping: $!");
d3664eb9 2525 return;
99870f4d
KW
2526 }
2527 $handle{$addr} = $file_handle; # Cache the open file handle
9e65c3f4 2528
0dee21d6
KW
2529 # If possible, make sure that the file is the correct version.
2530 # (This data isn't available on early Unicode releases or in
2531 # UnicodeData.txt.) We don't do this check if we are using a
2532 # substitute file instead of the official one (though the code
2533 # could be extended to do so).
2534 if ($in_this_release{$addr}
2535 && lc($file) ne 'unicodedata.txt')
2536 {
96f226dc 2537 if ($file !~ /^Unihan/i) {
0dee21d6
KW
2538
2539 # The non-Unihan files started getting version numbers in
2540 # 3.2, but some files in 4.0 are unchanged from 3.2, and
2541 # marked as 3.2. 4.0.1 is the first version where there
2542 # are no files marked as being from less than 4.0, though
2543 # some are marked as 4.0. In versions after that, the
2544 # numbers are correct.
2545 if ($v_version ge v4.0.1) {
2546 $_ = <$file_handle>; # The version number is in the
2547 # very first line
2548 if ($_ !~ / - $string_version \. /x) {
2549 chomp;
2550 $_ =~ s/^#\s*//;
2551
2552 # 4.0.1 had some valid files that weren't updated.
2553 if (! ($v_version eq v4.0.1 && $_ =~ /4\.0\.0/)) {
2554 die Carp::my_carp("File '$file' is version "
2555 . "'$_'. It should be "
2556 . "version $string_version");
2557 }
2558 }
cafe9cf0 2559 }
96f226dc 2560 }
0dee21d6
KW
2561 elsif ($v_version ge v6.0.0) { # Unihan
2562
2563 # Unihan files didn't get accurate version numbers until
2564 # 6.0. The version is somewhere in the first comment
2565 # block
96f226dc
KW
2566 while (<$file_handle>) {
2567 if ($_ !~ /^#/) {
0dee21d6
KW
2568 Carp::my_carp_bug("Could not find the expected "
2569 . "version info in file '$file'");
96f226dc
KW
2570 last;
2571 }
2572 chomp;
2573 $_ =~ s/^#\s*//;
2574 next if $_ !~ / version: /x;
2575 last if $_ =~ /$string_version/;
0dee21d6
KW
2576 die Carp::my_carp("File '$file' is version "
2577 . "'$_'. It should be "
2578 . "version $string_version");
96f226dc
KW
2579 }
2580 }
9e65c3f4 2581 }
99870f4d
KW
2582 }
2583
0dee21d6 2584 print "$progress_message{$addr}\n" if $verbosity >= $PROGRESS;
99870f4d
KW
2585
2586 # Call any special handler for before the file.
2587 &{$pre_handler{$addr}}($self) if $pre_handler{$addr};
2588
2589 # Then the main handler
2590 &{$handler{$addr}}($self);
2591
2592 # Then any special post-file handler.
2593 &{$post_handler{$addr}}($self) if $post_handler{$addr};
2594
2595 # If any errors have been accumulated, output the counts (as the first
2596 # error message in each class was output when it was encountered).
2597 if ($errors{$addr}) {
2598 my $total = 0;
2599 my $types = 0;
2600 foreach my $error (keys %{$errors{$addr}}) {
2601 $total += $errors{$addr}->{$error};
2602 delete $errors{$addr}->{$error};
2603 $types++;
2604 }
2605 if ($total > 1) {
2606 my $message
2607 = "A total of $total lines had errors in $file. ";
2608
2609 $message .= ($types == 1)
2610 ? '(Only the first one was displayed.)'
2611 : '(Only the first of each type was displayed.)';
2612 Carp::my_carp($message);
2613 }
2614 }
2615
2616 if (@{$missings{$addr}}) {
2617 Carp::my_carp_bug("Handler for $file didn't look at all the \@missing lines. Generated tables likely are wrong");
2618 }
2619
2620 # If a real file handle, close it.
2621 close $handle{$addr} or Carp::my_carp("Can't close $file: $!") if
2622 ref $handle{$addr};
2623 $handle{$addr} = ""; # Uses empty to indicate that has already seen
2624 # the file, as opposed to undef
2625 return;
2626 }
2627
74cd47d0 2628 sub _next_line {
99870f4d
KW
2629 # Sets $_ to be the next logical input line, if any. Returns non-zero
2630 # if such a line exists. 'logical' means that any lines that have
2631 # been added via insert_lines() will be returned in $_ before the file
2632 # is read again.
2633
2634 my $self = shift;
2635 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2636
ffe43484 2637 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
2638
2639 # Here the file is open (or if the handle is not a ref, is an open
2640 # 'virtual' file). Get the next line; any inserted lines get priority
2641 # over the file itself.
2642 my $adjusted;
2643
2644 LINE:
2645 while (1) { # Loop until find non-comment, non-empty line
2646 #local $to_trace = 1 if main::DEBUG;
2647 my $inserted_ref = shift @{$added_lines{$addr}};
2648 if (defined $inserted_ref) {
2649 ($adjusted, $_) = @{$inserted_ref};
2650 trace $adjusted, $_ if main::DEBUG && $to_trace;
2651 return 1 if $adjusted;
2652 }
2653 else {
2654 last if ! ref $handle{$addr}; # Don't read unless is real file
2655 last if ! defined ($_ = readline $handle{$addr});
2656 }
2657 chomp;
2658 trace $_ if main::DEBUG && $to_trace;
2659
2660 # See if this line is the comment line that defines what property
2661 # value that code points that are not listed in the file should
2662 # have. The format or existence of these lines is not guaranteed
2663 # by Unicode since they are comments, but the documentation says
2664 # that this was added for machine-readability, so probably won't
2665 # change. This works starting in Unicode Version 5.0. They look
2666 # like:
2667 #
2668 # @missing: 0000..10FFFF; Not_Reordered
2669 # @missing: 0000..10FFFF; Decomposition_Mapping; <code point>
2670 # @missing: 0000..10FFFF; ; NaN
2671 #
2672 # Save the line for a later get_missings() call.
2673 if (/$missing_defaults_prefix/) {
2674 if ($has_missings_defaults{$addr} == $NO_DEFAULTS) {
2675 $self->carp_bad_line("Unexpected \@missing line. Assuming no missing entries");
2676 }
2677 elsif ($has_missings_defaults{$addr} == $NOT_IGNORED) {
2678 my @defaults = split /\s* ; \s*/x, $_;
2679
2680 # The first field is the @missing, which ends in a
2681 # semi-colon, so can safely shift.
2682 shift @defaults;
2683
2684 # Some of these lines may have empty field placeholders
2685 # which get in the way. An example is:
2686 # @missing: 0000..10FFFF; ; NaN
2687 # Remove them. Process starting from the top so the
2688 # splice doesn't affect things still to be looked at.
2689 for (my $i = @defaults - 1; $i >= 0; $i--) {
2690 next if $defaults[$i] ne "";
2691 splice @defaults, $i, 1;
2692 }
2693
2694 # What's left should be just the property (maybe) and the
2695 # default. Having only one element means it doesn't have
2696 # the property.
2697 my $default;
2698 my $property;
2699 if (@defaults >= 1) {
2700 if (@defaults == 1) {
2701 $default = $defaults[0];
2702 }
2703 else {
2704 $property = $defaults[0];
2705 $default = $defaults[1];
2706 }
2707 }
2708
2709 if (@defaults < 1
2710 || @defaults > 2
2711 || ($default =~ /^</
2712 && $default !~ /^<code *point>$/i
09f8d0ac
KW
2713 && $default !~ /^<none>$/i
2714 && $default !~ /^<script>$/i))
99870f4d
KW
2715 {
2716 $self->carp_bad_line("Unrecognized \@missing line: $_. Assuming no missing entries");
2717 }
2718 else {
2719
2720 # If the property is missing from the line, it should
2721 # be the one for the whole file
2722 $property = $property{$addr} if ! defined $property;
2723
2724 # Change <none> to the null string, which is what it
2725 # really means. If the default is the code point
2726 # itself, set it to <code point>, which is what
2727 # Unicode uses (but sometimes they've forgotten the
2728 # space)
2729 if ($default =~ /^<none>$/i) {
2730 $default = "";
2731 }
2732 elsif ($default =~ /^<code *point>$/i) {
2733 $default = $CODE_POINT;
2734 }
09f8d0ac
KW
2735 elsif ($default =~ /^<script>$/i) {
2736
2737 # Special case this one. Currently is from
2738 # ScriptExtensions.txt, and means for all unlisted
2739 # code points, use their Script property values.
2740 # For the code points not listed in that file, the
2741 # default value is 'Unknown'.
2742 $default = "Unknown";
2743 }
99870f4d
KW
2744
2745 # Store them as a sub-arrays with both components.
2746 push @{$missings{$addr}}, [ $default, $property ];
2747 }
2748 }
2749
2750 # There is nothing for the caller to process on this comment
2751 # line.
2752 next;
2753 }
2754
2755 # Remove comments and trailing space, and skip this line if the
2756 # result is empty
2757 s/#.*//;
2758 s/\s+$//;
2759 next if /^$/;
2760
2761 # Call any handlers for this line, and skip further processing of
2762 # the line if the handler sets the line to null.
2763 foreach my $sub_ref (@{$each_line_handler{$addr}}) {
2764 &{$sub_ref}($self);
2765 next LINE if /^$/;
2766 }
2767
2768 # Here the line is ok. return success.
2769 return 1;
2770 } # End of looping through lines.
2771
3848e5b7 2772 # If there are EOF handlers, call each (only once) and if it generates
99870f4d 2773 # more lines to process go back in the loop to handle them.
3848e5b7
KW
2774 while ($eof_handler{$addr}->@*) {
2775 &{$eof_handler{$addr}[0]}($self);
2776 shift $eof_handler{$addr}->@*; # Currently only get one shot at it.
99870f4d
KW
2777 goto LINE if $added_lines{$addr};
2778 }
2779
2780 # Return failure -- no more lines.
2781 return 0;
2782
2783 }
2784
74cd47d0
KW
2785 sub _next_line_with_remapped_range {
2786 my $self = shift;
2787 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2788
2789 # like _next_line(), but for use on non-ASCII platforms. It sets $_
2790 # to be the next logical input line, if any. Returns non-zero if such
2791 # a line exists. 'logical' means that any lines that have been added
2792 # via insert_lines() will be returned in $_ before the file is read
2793 # again.
2794 #
2795 # The difference from _next_line() is that this remaps the Unicode
2796 # code points in the input to those of the native platform. Each
2797 # input line contains a single code point, or a single contiguous
2798 # range of them This routine splits each range into its individual
2799 # code points and caches them. It returns the cached values,
2800 # translated into their native equivalents, one at a time, for each
2801 # call, before reading the next line. Since native values can only be
2802 # a single byte wide, no translation is needed for code points above
2803 # 0xFF, and ranges that are entirely above that number are not split.
2804 # If an input line contains the range 254-1000, it would be split into
2805 # three elements: 254, 255, and 256-1000. (The downstream table
2806 # insertion code will sort and coalesce the individual code points
2807 # into appropriate ranges.)
2808
2809 my $addr = do { no overloading; pack 'J', $self; };
2810
2811 while (1) {
2812
2813 # Look in cache before reading the next line. Return any cached
2814 # value, translated
2815 my $inserted = shift @{$remapped_lines{$addr}};
2816 if (defined $inserted) {
2817 trace $inserted if main::DEBUG && $to_trace;
2818 $_ = $inserted =~ s/^ ( \d+ ) /sprintf("%04X", utf8::unicode_to_native($1))/xer;
2819 trace $_ if main::DEBUG && $to_trace;
2820 return 1;
2821 }
2822
2823 # Get the next line.
2824 return 0 unless _next_line($self);
2825
2826 # If there is a special handler for it, return the line,
2827 # untranslated. This should happen only for files that are
2828 # special, not being code-point related, such as property names.
2829 return 1 if $handler{$addr}
2830 != \&main::process_generic_property_file;
2831
2832 my ($range, $property_name, $map, @remainder)
2833 = split /\s*;\s*/, $_, -1; # -1 => retain trailing null fields
2834
2835 if (@remainder
2836 || ! defined $property_name
2837 || $range !~ /^ ($code_point_re) (?:\.\. ($code_point_re) )? $/x)
2838 {
2839 Carp::my_carp_bug("Unrecognized input line '$_'. Ignored");
2840 }
2841
2842 my $low = hex $1;
2843 my $high = (defined $2) ? hex $2 : $low;
2844
2845 # If the input maps the range to another code point, remap the
2846 # target if it is between 0 and 255.
2847 my $tail;
2848 if (defined $map) {
2849 $map =~ s/\b 00 ( [0-9A-F]{2} ) \b/sprintf("%04X", utf8::unicode_to_native(hex $1))/gxe;
2850 $tail = "$property_name; $map";
2851 $_ = "$range; $tail";
2852 }
2853 else {
2854 $tail = $property_name;
2855 }
2856
2857 # If entire range is above 255, just return it, unchanged (except
2858 # any mapped-to code point, already changed above)
2859 return 1 if $low > 255;
2860
2861 # Cache an entry for every code point < 255. For those in the
2862 # range above 255, return a dummy entry for just that portion of
2863 # the range. Note that this will be out-of-order, but that is not
2864 # a problem.
2865 foreach my $code_point ($low .. $high) {
2866 if ($code_point > 255) {
2867 $_ = sprintf "%04X..%04X; $tail", $code_point, $high;
2868 return 1;
2869 }
2870 push @{$remapped_lines{$addr}}, "$code_point; $tail";
2871 }
2872 } # End of looping through lines.
2873
2874 # NOTREACHED
2875 }
2876
99870f4d
KW
2877# Not currently used, not fully tested.
2878# sub peek {
2879# # Non-destructive look-ahead one non-adjusted, non-comment, non-blank
2880# # record. Not callable from an each_line_handler(), nor does it call
2881# # an each_line_handler() on the line.
2882#
2883# my $self = shift;
ffe43484 2884# my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
2885#
2886# foreach my $inserted_ref (@{$added_lines{$addr}}) {
2887# my ($adjusted, $line) = @{$inserted_ref};
2888# next if $adjusted;
2889#
2890# # Remove comments and trailing space, and return a non-empty
2891# # resulting line
2892# $line =~ s/#.*//;
2893# $line =~ s/\s+$//;
2894# return $line if $line ne "";
2895# }
2896#
2897# return if ! ref $handle{$addr}; # Don't read unless is real file
2898# while (1) { # Loop until find non-comment, non-empty line
2899# local $to_trace = 1 if main::DEBUG;
2900# trace $_ if main::DEBUG && $to_trace;
2901# return if ! defined (my $line = readline $handle{$addr});
2902# chomp $line;
2903# push @{$added_lines{$addr}}, [ 0, $line ];
2904#
2905# $line =~ s/#.*//;
2906# $line =~ s/\s+$//;
2907# return $line if $line ne "";
2908# }
2909#
2910# return;
2911# }
2912
2913
2914 sub insert_lines {
2915 # Lines can be inserted so that it looks like they were in the input
2916 # file at the place it was when this routine is called. See also
2917 # insert_adjusted_lines(). Lines inserted via this routine go through
2918 # any each_line_handler()
2919
2920 my $self = shift;
2921
2922 # Each inserted line is an array, with the first element being 0 to
2923 # indicate that this line hasn't been adjusted, and needs to be
2924 # processed.
f998e60c 2925 no overloading;
051df77b 2926 push @{$added_lines{pack 'J', $self}}, map { [ 0, $_ ] } @_;
99870f4d
KW
2927 return;
2928 }
2929
2930 sub insert_adjusted_lines {
2931 # Lines can be inserted so that it looks like they were in the input
2932 # file at the place it was when this routine is called. See also
2933 # insert_lines(). Lines inserted via this routine are already fully
2934 # adjusted, ready to be processed; each_line_handler()s handlers will
2935 # not be called. This means this is not a completely general
2936 # facility, as only the last each_line_handler on the stack should
2937 # call this. It could be made more general, by passing to each of the
2938 # line_handlers their position on the stack, which they would pass on
2939 # to this routine, and that would replace the boolean first element in
2940 # the anonymous array pushed here, so that the next_line routine could
2941 # use that to call only those handlers whose index is after it on the
2942 # stack. But this is overkill for what is needed now.
2943
2944 my $self = shift;
2945 trace $_[0] if main::DEBUG && $to_trace;
2946
2947 # Each inserted line is an array, with the first element being 1 to
2948 # indicate that this line has been adjusted
f998e60c 2949 no overloading;
051df77b 2950 push @{$added_lines{pack 'J', $self}}, map { [ 1, $_ ] } @_;
99870f4d
KW
2951 return;
2952 }
2953
2954 sub get_missings {
2955 # Returns the stored up @missings lines' values, and clears the list.
2956 # The values are in an array, consisting of the default in the first
2957 # element, and the property in the 2nd. However, since these lines
2958 # can be stacked up, the return is an array of all these arrays.
2959
2960 my $self = shift;
2961 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2962
ffe43484 2963 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
2964
2965 # If not accepting a list return, just return the first one.
2966 return shift @{$missings{$addr}} unless wantarray;
2967
2968 my @return = @{$missings{$addr}};
2969 undef @{$missings{$addr}};
2970 return @return;
2971 }
2972
2973 sub _insert_property_into_line {
2974 # Add a property field to $_, if this file requires it.
2975
f998e60c 2976 my $self = shift;
ffe43484 2977 my $addr = do { no overloading; pack 'J', $self; };
f998e60c 2978 my $property = $property{$addr};
99870f4d
KW
2979 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2980
2981 $_ =~ s/(;|$)/; $property$1/;
2982 return;
2983 }
2984
2985 sub carp_bad_line {
2986 # Output consistent error messages, using either a generic one, or the
2987 # one given by the optional parameter. To avoid gazillions of the
2988 # same message in case the syntax of a file is way off, this routine
2989 # only outputs the first instance of each message, incrementing a
2990 # count so the totals can be output at the end of the file.
2991
2992 my $self = shift;
2993 my $message = shift;
2994 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
2995
ffe43484 2996 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
2997
2998 $message = 'Unexpected line' unless $message;
2999
3000 # No trailing punctuation so as to fit with our addenda.
3001 $message =~ s/[.:;,]$//;
3002
3003 # If haven't seen this exact message before, output it now. Otherwise
3004 # increment the count of how many times it has occurred
3005 unless ($errors{$addr}->{$message}) {
3006 Carp::my_carp("$message in '$_' in "
f998e60c 3007 . $file{$addr}
99870f4d
KW
3008 . " at line $.. Skipping this line;");
3009 $errors{$addr}->{$message} = 1;
3010 }
3011 else {
3012 $errors{$addr}->{$message}++;
3013 }
3014
3015 # Clear the line to prevent any further (meaningful) processing of it.
3016 $_ = "";
3017
3018 return;
3019 }
3020} # End closure
3021
3022package Multi_Default;
3023
3024# Certain properties in early versions of Unicode had more than one possible
3025# default for code points missing from the files. In these cases, one
3026# default applies to everything left over after all the others are applied,
3027# and for each of the others, there is a description of which class of code
3028# points applies to it. This object helps implement this by storing the
3029# defaults, and for all but that final default, an eval string that generates
3030# the class that it applies to.
3031
3032
3033{ # Closure
3034
3035 main::setup_package();
3036
3037 my %class_defaults;
3038 # The defaults structure for the classes
3039 main::set_access('class_defaults', \%class_defaults);
3040
3041 my %other_default;
3042 # The default that applies to everything left over.
3043 main::set_access('other_default', \%other_default, 'r');
3044
3045
3046 sub new {
3047 # The constructor is called with default => eval pairs, terminated by
3048 # the left-over default. e.g.
3049 # Multi_Default->new(
3050 # 'T' => '$gc->table("Mn") + $gc->table("Cf") - 0x200C
3051 # - 0x200D',
3052 # 'R' => 'some other expression that evaluates to code points',
3053 # .
3054 # .
3055 # .
3056 # 'U'));
3057
3058 my $class = shift;
3059
3060 my $self = bless \do{my $anonymous_scalar}, $class;
ffe43484 3061 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3062
3063 while (@_ > 1) {
3064 my $default = shift;
3065 my $eval = shift;
3066 $class_defaults{$addr}->{$default} = $eval;
3067 }
3068
3069 $other_default{$addr} = shift;
3070
3071 return $self;
3072 }
3073
3074 sub get_next_defaults {
3075 # Iterates and returns the next class of defaults.
3076 my $self = shift;
3077 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3078
ffe43484 3079 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3080
3081 return each %{$class_defaults{$addr}};
3082 }
3083}
3084
3085package Alias;
3086
3087# An alias is one of the names that a table goes by. This class defines them
3088# including some attributes. Everything is currently setup in the
3089# constructor.
3090
3091
3092{ # Closure
3093
3094 main::setup_package();
3095
3096 my %name;
3097 main::set_access('name', \%name, 'r');
3098
3099 my %loose_match;
c12f2655 3100 # Should this name match loosely or not.
99870f4d
KW
3101 main::set_access('loose_match', \%loose_match, 'r');
3102
33e96e72
KW
3103 my %make_re_pod_entry;
3104 # Some aliases should not get their own entries in the re section of the
3105 # pod, because they are covered by a wild-card, and some we want to
3106 # discourage use of. Binary
f82fe4ba 3107 main::set_access('make_re_pod_entry', \%make_re_pod_entry, 'r', 's');
99870f4d 3108
fd1e3e84
KW
3109 my %ucd;
3110 # Is this documented to be accessible via Unicode::UCD
3111 main::set_access('ucd', \%ucd, 'r', 's');
3112
99870f4d
KW
3113 my %status;
3114 # Aliases have a status, like deprecated, or even suppressed (which means
3115 # they don't appear in documentation). Enum
3116 main::set_access('status', \%status, 'r');
3117
0eac1e20 3118 my %ok_as_filename;
99870f4d
KW
3119 # Similarly, some aliases should not be considered as usable ones for
3120 # external use, such as file names, or we don't want documentation to
3121 # recommend them. Boolean
0eac1e20 3122 main::set_access('ok_as_filename', \%ok_as_filename, 'r');
99870f4d
KW
3123
3124 sub new {
3125 my $class = shift;
3126
3127 my $self = bless \do { my $anonymous_scalar }, $class;
ffe43484 3128 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3129
3130 $name{$addr} = shift;
3131 $loose_match{$addr} = shift;
33e96e72 3132 $make_re_pod_entry{$addr} = shift;
0eac1e20 3133 $ok_as_filename{$addr} = shift;
99870f4d 3134 $status{$addr} = shift;
fd1e3e84 3135 $ucd{$addr} = shift;
99870f4d
KW
3136
3137 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3138
3139 # Null names are never ok externally
0eac1e20 3140 $ok_as_filename{$addr} = 0 if $name{$addr} eq "";
99870f4d
KW
3141
3142 return $self;
3143 }
3144}
3145
3146package Range;
3147
3148# A range is the basic unit for storing code points, and is described in the
3149# comments at the beginning of the program. Each range has a starting code
3150# point; an ending code point (not less than the starting one); a value
3151# that applies to every code point in between the two end-points, inclusive;
3152# and an enum type that applies to the value. The type is for the user's
3153# convenience, and has no meaning here, except that a non-zero type is
3154# considered to not obey the normal Unicode rules for having standard forms.
3155#
3156# The same structure is used for both map and match tables, even though in the
3157# latter, the value (and hence type) is irrelevant and could be used as a
3158# comment. In map tables, the value is what all the code points in the range
3159# map to. Type 0 values have the standardized version of the value stored as
3160# well, so as to not have to recalculate it a lot.
3161
3162sub trace { return main::trace(@_); }
3163
3164{ # Closure
3165
3166 main::setup_package();
3167
3168 my %start;
3169 main::set_access('start', \%start, 'r', 's');
3170
3171 my %end;
3172 main::set_access('end', \%end, 'r', 's');
3173
3174 my %value;
3175 main::set_access('value', \%value, 'r');
3176
3177 my %type;
3178 main::set_access('type', \%type, 'r');
3179
3180 my %standard_form;
3181 # The value in internal standard form. Defined only if the type is 0.
3182 main::set_access('standard_form', \%standard_form);
3183
3184 # Note that if these fields change, the dump() method should as well
3185
3186 sub new {
3187 return Carp::carp_too_few_args(\@_, 3) if main::DEBUG && @_ < 3;
3188 my $class = shift;
3189
3190 my $self = bless \do { my $anonymous_scalar }, $class;
ffe43484 3191 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3192
3193 $start{$addr} = shift;
3194 $end{$addr} = shift;
3195
3196 my %args = @_;
3197
3198 my $value = delete $args{'Value'}; # Can be 0
3199 $value = "" unless defined $value;
3200 $value{$addr} = $value;
3201
3202 $type{$addr} = delete $args{'Type'} || 0;
3203
3204 Carp::carp_extra_args(\%args) if main::DEBUG && %args;
3205
99870f4d
KW
3206 return $self;
3207 }
3208
3209 use overload
3210 fallback => 0,
3211 qw("") => "_operator_stringify",
3212 "." => \&main::_operator_dot,
1285127e 3213 ".=" => \&main::_operator_dot_equal,
99870f4d
KW
3214 ;
3215
3216 sub _operator_stringify {
3217 my $self = shift;
ffe43484 3218 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3219
3220 # Output it like '0041..0065 (value)'
3221 my $return = sprintf("%04X", $start{$addr})
3222 . '..'
3223 . sprintf("%04X", $end{$addr});
3224 my $value = $value{$addr};
3225 my $type = $type{$addr};
3226 $return .= ' (';
3227 $return .= "$value";
3228 $return .= ", Type=$type" if $type != 0;
3229 $return .= ')';
3230
3231 return $return;
3232 }
3233
3234 sub standard_form {
c292d35a
NC
3235 # Calculate the standard form only if needed, and cache the result.
3236 # The standard form is the value itself if the type is special.
3237 # This represents a considerable CPU and memory saving - at the time
3238 # of writing there are 368676 non-special objects, but the standard
3239 # form is only requested for 22047 of them - ie about 6%.
99870f4d
KW
3240
3241 my $self = shift;
3242 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3243
ffe43484 3244 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3245
3246 return $standard_form{$addr} if defined $standard_form{$addr};
c292d35a
NC
3247
3248 my $value = $value{$addr};
3249 return $value if $type{$addr};
3250 return $standard_form{$addr} = main::standardize($value);
99870f4d
KW
3251 }
3252
3253 sub dump {
3254 # Human, not machine readable. For machine readable, comment out this
3255 # entire routine and let the standard one take effect.
3256 my $self = shift;
3257 my $indent = shift;
3258 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3259
ffe43484 3260 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3261
3262 my $return = $indent
3263 . sprintf("%04X", $start{$addr})
3264 . '..'
3265 . sprintf("%04X", $end{$addr})
3266 . " '$value{$addr}';";
3267 if (! defined $standard_form{$addr}) {
3268 $return .= "(type=$type{$addr})";
3269 }
3270 elsif ($standard_form{$addr} ne $value{$addr}) {
3271 $return .= "(standard '$standard_form{$addr}')";
3272 }
3273 return $return;
3274 }
3275} # End closure
3276
3277package _Range_List_Base;
3278
3279# Base class for range lists. A range list is simply an ordered list of
3280# ranges, so that the ranges with the lowest starting numbers are first in it.
3281#
3282# When a new range is added that is adjacent to an existing range that has the
3283# same value and type, it merges with it to form a larger range.
3284#
3285# Ranges generally do not overlap, except that there can be multiple entries
3286# of single code point ranges. This is because of NameAliases.txt.
3287#
3288# In this program, there is a standard value such that if two different
3289# values, have the same standard value, they are considered equivalent. This
3290# value was chosen so that it gives correct results on Unicode data
3291
3292# There are a number of methods to manipulate range lists, and some operators
3293# are overloaded to handle them.
3294
99870f4d
KW
3295sub trace { return main::trace(@_); }
3296
3297{ # Closure
3298
3299 our $addr;
3300
5b348b71
KW
3301 # Max is initialized to a negative value that isn't adjacent to 0, for
3302 # simpler tests
3303 my $max_init = -2;
3304
99870f4d
KW
3305 main::setup_package();
3306
3307 my %ranges;
3308 # The list of ranges
3309 main::set_access('ranges', \%ranges, 'readable_array');
3310
3311 my %max;
3312 # The highest code point in the list. This was originally a method, but
3313 # actual measurements said it was used a lot.
3314 main::set_access('max', \%max, 'r');
3315
3316 my %each_range_iterator;
3317 # Iterator position for each_range()
3318 main::set_access('each_range_iterator', \%each_range_iterator);
3319
3320 my %owner_name_of;
3321 # Name of parent this is attached to, if any. Solely for better error
3322 # messages.
3323 main::set_access('owner_name_of', \%owner_name_of, 'p_r');
3324
3325 my %_search_ranges_cache;
3326 # A cache of the previous result from _search_ranges(), for better
3327 # performance
3328 main::set_access('_search_ranges_cache', \%_search_ranges_cache);
3329
3330 sub new {
3331 my $class = shift;
3332 my %args = @_;
3333
3334 # Optional initialization data for the range list.
3335 my $initialize = delete $args{'Initialize'};
3336
3337 my $self;
3338
3339 # Use _union() to initialize. _union() returns an object of this
3340 # class, which means that it will call this constructor recursively.
3341 # But it won't have this $initialize parameter so that it won't
3342 # infinitely loop on this.
3343 return _union($class, $initialize, %args) if defined $initialize;
3344
3345 $self = bless \do { my $anonymous_scalar }, $class;
ffe43484 3346 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3347
3348 # Optional parent object, only for debug info.
3349 $owner_name_of{$addr} = delete $args{'Owner'};
3350 $owner_name_of{$addr} = "" if ! defined $owner_name_of{$addr};
3351
3352 # Stringify, in case it is an object.
3353 $owner_name_of{$addr} = "$owner_name_of{$addr}";
3354
3355 # This is used only for error messages, and so a colon is added
3356 $owner_name_of{$addr} .= ": " if $owner_name_of{$addr} ne "";
3357
3358 Carp::carp_extra_args(\%args) if main::DEBUG && %args;
3359
5b348b71 3360 $max{$addr} = $max_init;
99870f4d
KW
3361
3362 $_search_ranges_cache{$addr} = 0;
3363 $ranges{$addr} = [];
3364
3365 return $self;
3366 }
3367
3368 use overload
3369 fallback => 0,
3370 qw("") => "_operator_stringify",
3371 "." => \&main::_operator_dot,
1285127e 3372 ".=" => \&main::_operator_dot_equal,
99870f4d
KW
3373 ;
3374
3375 sub _operator_stringify {
3376 my $self = shift;
ffe43484 3377 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3378
3379 return "Range_List attached to '$owner_name_of{$addr}'"
3380 if $owner_name_of{$addr};
3381 return "anonymous Range_List " . \$self;
3382 }
3383
3384 sub _union {
3385 # Returns the union of the input code points. It can be called as
3386 # either a constructor or a method. If called as a method, the result
3387 # will be a new() instance of the calling object, containing the union
3388 # of that object with the other parameter's code points; if called as
d59563d0 3389 # a constructor, the first parameter gives the class that the new object
99870f4d
KW
3390 # should be, and the second parameter gives the code points to go into
3391 # it.
3392 # In either case, there are two parameters looked at by this routine;
3393 # any additional parameters are passed to the new() constructor.
3394 #
3395 # The code points can come in the form of some object that contains
3396 # ranges, and has a conventionally named method to access them; or
3397 # they can be an array of individual code points (as integers); or
3398 # just a single code point.
3399 #
3400 # If they are ranges, this routine doesn't make any effort to preserve
3198cc57
KW
3401 # the range values and types of one input over the other. Therefore
3402 # this base class should not allow _union to be called from other than
99870f4d
KW
3403 # initialization code, so as to prevent two tables from being added
3404 # together where the range values matter. The general form of this
3405 # routine therefore belongs in a derived class, but it was moved here
3406 # to avoid duplication of code. The failure to overload this in this
3407 # class keeps it safe.
3198cc57
KW
3408 #
3409 # It does make the effort during initialization to accept tables with
3410 # multiple values for the same code point, and to preserve the order
3411 # of these. If there is only one input range or range set, it doesn't
3412 # sort (as it should already be sorted to the desired order), and will
3413 # accept multiple values per code point. Otherwise it will merge
3414 # multiple values into a single one.
99870f4d
KW
3415
3416 my $self;
3417 my @args; # Arguments to pass to the constructor
3418
3419 my $class = shift;
3420
3421 # If a method call, will start the union with the object itself, and
3422 # the class of the new object will be the same as self.
3423 if (ref $class) {
3424 $self = $class;
3425 $class = ref $self;
3426 push @args, $self;
3427 }
3428
3429 # Add the other required parameter.
3430 push @args, shift;
3431 # Rest of parameters are passed on to the constructor
3432
3433 # Accumulate all records from both lists.
3434 my @records;
3198cc57 3435 my $input_count = 0;
99870f4d
KW
3436 for my $arg (@args) {
3437 #local $to_trace = 0 if main::DEBUG;
3438 trace "argument = $arg" if main::DEBUG && $to_trace;
3439 if (! defined $arg) {
3440 my $message = "";
3441 if (defined $self) {
f998e60c 3442 no overloading;
051df77b 3443 $message .= $owner_name_of{pack 'J', $self};
99870f4d 3444 }
ada6088e 3445 Carp::my_carp_bug($message . "Undefined argument to _union. No union done.");
99870f4d
KW
3446 return;
3447 }
3198cc57 3448
99870f4d
KW
3449 $arg = [ $arg ] if ! ref $arg;
3450 my $type = ref $arg;
3451 if ($type eq 'ARRAY') {
3452 foreach my $element (@$arg) {
3453 push @records, Range->new($element, $element);
3198cc57 3454 $input_count++;
99870f4d
KW
3455 }
3456 }
3457 elsif ($arg->isa('Range')) {
3458 push @records, $arg;
3198cc57 3459 $input_count++;
99870f4d
KW
3460 }
3461 elsif ($arg->can('ranges')) {
3462 push @records, $arg->ranges;
3198cc57 3463 $input_count++;
99870f4d
KW
3464 }
3465 else {
3466 my $message = "";
3467 if (defined $self) {
f998e60c 3468 no overloading;
051df77b 3469 $message .= $owner_name_of{pack 'J', $self};
99870f4d
KW
3470 }
3471 Carp::my_carp_bug($message . "Cannot take the union of a $type. No union done.");
3472 return;
3473 }
3474 }
3475
3476 # Sort with the range containing the lowest ordinal first, but if
3477 # two ranges start at the same code point, sort with the bigger range
3478 # of the two first, because it takes fewer cycles.
3198cc57
KW
3479 if ($input_count > 1) {
3480 @records = sort { ($a->start <=> $b->start)
99870f4d
KW
3481 or
3482 # if b is shorter than a, b->end will be
3483 # less than a->end, and we want to select
3484 # a, so want to return -1
3485 ($b->end <=> $a->end)
3486 } @records;
3198cc57 3487 }
99870f4d
KW
3488
3489 my $new = $class->new(@_);
3490
3491 # Fold in records so long as they add new information.
3492 for my $set (@records) {
3493 my $start = $set->start;
3494 my $end = $set->end;
d59563d0 3495 my $value = $set->value;
3198cc57 3496 my $type = $set->type;
99870f4d 3497 if ($start > $new->max) {
3198cc57 3498 $new->_add_delete('+', $start, $end, $value, Type => $type);
99870f4d
KW
3499 }
3500 elsif ($end > $new->max) {
3198cc57
KW
3501 $new->_add_delete('+', $new->max +1, $end, $value,
3502 Type => $type);
3503 }
3504 elsif ($input_count == 1) {
3505 # Here, overlaps existing range, but is from a single input,
3506 # so preserve the multiple values from that input.
3507 $new->_add_delete('+', $start, $end, $value, Type => $type,
3508 Replace => $MULTIPLE_AFTER);
99870f4d
KW
3509 }
3510 }
3511
3512 return $new;
3513 }
3514
3515 sub range_count { # Return the number of ranges in the range list
3516 my $self = shift;
3517 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3518
f998e60c 3519 no overloading;
051df77b 3520 return scalar @{$ranges{pack 'J', $self}};
99870f4d
KW
3521 }
3522
3523 sub min {
3524 # Returns the minimum code point currently in the range list, or if
3525 # the range list is empty, 2 beyond the max possible. This is a
3526 # method because used so rarely, that not worth saving between calls,
3527 # and having to worry about changing it as ranges are added and
3528 # deleted.
3529
3530 my $self = shift;
3531 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3532
ffe43484 3533 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3534
3535 # If the range list is empty, return a large value that isn't adjacent
3536 # to any that could be in the range list, for simpler tests
2d88a86a 3537 return $MAX_WORKING_CODEPOINT + 2 unless scalar @{$ranges{$addr}};
99870f4d
KW
3538 return $ranges{$addr}->[0]->start;
3539 }
3540
3541 sub contains {
3542 # Boolean: Is argument in the range list? If so returns $i such that:
3543 # range[$i]->end < $codepoint <= range[$i+1]->end
3544 # which is one beyond what you want; this is so that the 0th range
3545 # doesn't return false
3546 my $self = shift;
3547 my $codepoint = shift;
3548 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3549
99870f4d
KW
3550 my $i = $self->_search_ranges($codepoint);
3551 return 0 unless defined $i;
3552
3553 # The search returns $i, such that
3554 # range[$i-1]->end < $codepoint <= range[$i]->end
3555 # So is in the table if and only iff it is at least the start position
3556 # of range $i.
f998e60c 3557 no overloading;
051df77b 3558 return 0 if $ranges{pack 'J', $self}->[$i]->start > $codepoint;
99870f4d
KW
3559 return $i + 1;
3560 }
3561
2f7a8815
KW
3562 sub containing_range {
3563 # Returns the range object that contains the code point, undef if none
3564
3565 my $self = shift;
3566 my $codepoint = shift;
3567 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3568
3569 my $i = $self->contains($codepoint);
3570 return unless $i;
3571
3572 # contains() returns 1 beyond where we should look
3573 no overloading;
3574 return $ranges{pack 'J', $self}->[$i-1];
3575 }
3576
99870f4d
KW
3577 sub value_of {
3578 # Returns the value associated with the code point, undef if none
3579
3580 my $self = shift;
3581 my $codepoint = shift;
3582 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3583
d69c231b
KW
3584 my $range = $self->containing_range($codepoint);
3585 return unless defined $range;
99870f4d 3586
d69c231b 3587 return $range->value;
99870f4d
KW
3588 }
3589
0a9dbafc
KW
3590 sub type_of {
3591 # Returns the type of the range containing the code point, undef if
3592 # the code point is not in the table
3593
3594 my $self = shift;
3595 my $codepoint = shift;
3596 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3597
3598 my $range = $self->containing_range($codepoint);
3599 return unless defined $range;
3600
3601 return $range->type;
3602 }
3603
99870f4d
KW
3604 sub _search_ranges {
3605 # Find the range in the list which contains a code point, or where it
3606 # should go if were to add it. That is, it returns $i, such that:
3607 # range[$i-1]->end < $codepoint <= range[$i]->end
3608 # Returns undef if no such $i is possible (e.g. at end of table), or
3609 # if there is an error.
3610
3611 my $self = shift;
3612 my $code_point = shift;
3613 Carp::carp_extra_args(\@_) if main::DEBUG && @_;
3614
ffe43484 3615 my $addr = do { no overloading; pack 'J', $self; };
99870f4d
KW
3616
3617 return if $code_point > $max{$addr};
3618 my $r = $ranges{$addr}; # The current list of ranges
3619 my $range_list_size = scalar @$r;
3620 my $i;
3621
3622 use integer; # want integer division
3623
3624 # Use the cached result as the starting guess for this one, because,
3625 # an experiment on 5.1 showed that 90% of the time the cache was the
3626 # same as the result on the next call (and 7% it was one less).
3627 $i = $_search_ranges_cache{$addr};
3628 $i = 0 if $i >= $range_list_size; # Reset if no longer valid (prob.
3629 # from an intervening deletion
3630 #local $to_trace = 1 if main::DEBUG;
3631 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);
3632 return $i if $code_point <= $r->[$i]->end
3633 && ($i == 0 || $r->[$i-1]->end < $code_point);
3634
3635 # Here the cache doesn't yield the correct $i. Try adding 1.
3636 if ($i < $range_list_size - 1
3637 && $r->[$i]->end < $code_point &&
3638 $code_point <= $r->[$i+1]->end)
3639 {
3640 $i++;
3641 trace "next \$i is correct: $i" if main::DEBUG && $to_trace;
3642 $_search_ranges_cache{$addr} = $i;
3643 return $i;
3644 }
3645
3646 # Here, adding 1 also didn't work. We do a binary search to
3647 # find the correct position, starting with current $i
3648 my $lower = 0;
3649 my $upper = $range_list_size - 1;
3650 while (1) {
3651 trace "top of loop i=$i:", sprintf("%04X", $r->[$lower]->start), "[$lower] .. ", sprintf("%04X", $r->[$i]->start), "[$i] .. ", sprintf("%04X", $r->[$upper]->start), "[$upper]" if main::DEBUG && $to_trace;
3652
3653 if ($code_point <= $r->[$i]->end) {
3654
3655 # Here we have met the upper constraint. We can quit if we
3656 # also meet the lower one.
3657 last if $i == 0 || $r->[$i-1]->end < $code_point;
3658
3659 $upper = $i; # Still too high.
3660
3661 }
3662 else {
3663
3664 # Here, $r[$i]->end < $code_point, so look higher up.
3665 $lower = $i;
3666 }
3667
3668 # Split search domain in half to try again.
3669 my $temp = ($upper + $lower) / 2;
3670
3671 # No point in continuing unless $i changes for next time
3672 # in the loop.
3673 if ($temp == $i) {
3674
3675 # We can't reach the highest element because of the averaging.
3676 # So if one below the upper edge, force it there and try one
3677 # more time.
3678 if ($i == $range_list_size - 2) {
3679
3680 trace "Forcing to upper edge" if main::DEBUG && $to_trace;
3681 $i = $range_list_size - 1;
3682
3683 # Change $lower as well so if fails next time through,
3684 # taking the average will yield the same $i, and we will
3685 # quit with the error message just below.
3686 $lower = $i;
3687 next;
3688 }
3689 Carp::my_carp_bug("$owner_name_of{$addr}Can't find where the range ought to go. No action taken.");
3690 return;
3691 }
3692 $i = $temp;
3693 } # End of while loop
3694
3695 if (main::DEBUG && $to_trace) {
3696 trace 'i-1=[', $i-1, ']', $r->[$i-1] if $i;
3697 trace "i= [ $i ]", $r->[$i];
3698 trace 'i+1=[', $i+1, ']', $r->[$i+1] if $i < $range_list_size - 1;
3699 }
3700
3701 # Here we have found the offset. Cache it as a starting point for the
3702 # next call.
3703 $_search_ranges_cache{$addr} = $i;
3704 return $i;
3705 }
3706
3707 sub _add_delete {
3708 # Add, replace or delete ranges to or from a list. The $type
3709 # parameter gives which:
3710 # '+' => insert or replace a range, returning a list of any changed
3711 # ranges.
3712 # '-' => delete a range, returning a list of any deleted ranges.
3713 #
3714 # The next three parameters give respectively the start, end, and
3715 # value associated with the range. 'value' should be null unless the
3716 # operation is '+';
3717 #
3718 # The range list is kept sorted so that the range with the lowest