Update documentation and comments for UCD generation

This commit is contained in:
Philip Hazel 2022-01-10 16:26:41 +00:00
parent 838cdac4dc
commit 87571b5af3
4 changed files with 95 additions and 101 deletions

View File

@ -10,20 +10,15 @@
# PCRE project by Peter Kankowski in 2008 as part of a previous upgrading of
# Unicode property support. A number of extensions have since been added. The
# main difference in the 2021 upgrade (apart from comments and layout) is that
# the data tables (e.g. list of script names) are now held in a separate Python
# module that is shared with the other Generate scripts.
# the data tables (e.g. list of script names) are now listed in or generated by
# a separate Python module that is shared with the other Generate scripts.
#
# This script must be run in the "maint" directory. It requires eight Unicode
# data tables: DerivedBidiClass.txt, DerivedGeneralCategory.txt,
# GraphemeBreakProperty.txt, PropList.txt, Scripts.txt, ScriptExtensions.txt,
# CaseFolding.txt, and emoji-data.txt. These must be in the Unicode.tables
# subdirectory.
#
# DerivedBidiClass.txt and DerivedGeneralCategory.txt are in the "extracted"
# subdirectory of the Unicode database (UCD) on the Unicode web site;
# GraphemeBreakProperty.txt is in the "auxiliary" subdirectory. PropList.txt,
# Scripts.txt, ScriptExtensions.txt, and CaseFolding.txt are directly in the
# UCD directory.
# This script must be run in the "maint" directory. It requires the following
# Unicode data tables: BidiMirrorring.txt, CaseFolding.txt,
# DerivedBidiClass.txt, DerivedCoreProperties.txt, DerivedGeneralCategory.txt,
# GraphemeBreakProperty.txt, PropList.txt, PropertyAliases.txt,
# PropertyValueAliases.txt, ScriptExtensions.txt, Scripts.txt, and
# emoji-data.txt. These must be in the Unicode.tables subdirectory.
#
# The emoji-data.txt file is found in the "emoji" subdirectory even though it
# is technically part of a different (but coordinated) standard as shown
@ -32,6 +27,11 @@
#
# http://unicode.org/Public/emoji/13.0/ReadMe.txt
#
# DerivedBidiClass.txt and DerivedGeneralCategory.txt are in the "extracted"
# subdirectory of the Unicode database (UCD) on the Unicode web site;
# GraphemeBreakProperty.txt is in the "auxiliary" subdirectory. The other files
# are in the top-level UCD directory.
#
# -----------------------------------------------------------------------------
# Minor modifications made to the original script:
# Added #! line at start
@ -106,6 +106,7 @@
# Changes to the refactored script:
#
# 26-December-2021: Refactoring completed
# 10-January-2022: Addition of general Boolean property support
#
# ----------------------------------------------------------------------------
#
@ -117,7 +118,8 @@
# Conceptually, there is a table of records (of type ucd_record), one for each
# Unicode character. Each record contains the script number, script extension
# value, character type, grapheme break type, offset to caseless matching set,
# offset to the character's other case, and the bidi class/control.
# offset to the character's other case, the bidi class, and offset to bitmap of
# Boolean properties.
#
# A real table covering all Unicode characters would be far too big. It can be
# efficiently compressed by observing that many characters have the same
@ -125,7 +127,7 @@
# the same set of records as other blocks. This leads to a 2-stage lookup
# process.
#
# This script constructs six tables. The ucd_caseless_sets table contains
# This script constructs seven tables. The ucd_caseless_sets table contains
# lists of characters that all match each other caselessly. Each list is
# in order, and is terminated by NOTACHAR (0xffffffff), which is larger than
# any valid character. The first list is empty; this is used for characters
@ -136,13 +138,12 @@
# in script runs all come from the same set. The first element in the vector
# contains the number of subsequent elements, which are in ascending order.
#
# The lists of scripts in script_names and script_abbrevs are partitioned into
# two groups. Scripts that appear in at least one character's script extension
# list come first, follwed by "Unknown" and then all the rest. This sorting is
# done certain automatically in the GenerateCommon.py script. A script's number
# is its index in these lists.
# Scripts are partitioned into two groups. Scripts that appear in at least one
# character's script extension list come first, followed by "Unknown" and then
# all the rest. This sorting is done automatically in the GenerateCommon.py
# script. A script's number is its index in the script_names list.
#
# The ucd_script_sets vector contains bitmaps that represent lists of scripts
# The ucd_script_sets table contains bitmaps that represent lists of scripts
# for Script Extensions properties. Each bitmap consists of a fixed number of
# unsigned 32-bit numbers, enough to allocate a bit for every script that is
# used in any character's extension list, that is, enough for every script
@ -151,10 +152,15 @@
# bitmap has no bits set; characters that have no script extensions have zero
# as their script extensions value so that they use this map.
#
# The ucd_records table contains one instance of every unique record that is
# required. The ucd_stage1 table is indexed by a character's block number,
# which is the character's code point divided by 128, since 128 is the size
# of each block. The result of a lookup in ucd_stage1 a "virtual" block number.
# The ucd_boolprop_sets table contains bitmaps that represent lists of Boolean
# properties. Each bitmap consists of a fixed number of unsigned 32-bit
# numbers, enough to allocate a bit for each supported Boolean property.
#
# The ucd_records table contains one instance of every unique character record
# that is required. The ucd_stage1 table is indexed by a character's block
# number, which is the character's code point divided by 128, since 128 is the
# size of each block. The result of a lookup in ucd_stage1 a "virtual" block
# number.
#
# The ucd_stage2 table is a table of "virtual" blocks; each block is indexed by
# the offset of a character within its own block, and the result is the index
@ -165,15 +171,16 @@
#
# Example: lowercase "a" (U+0061) is in block 0
# lookup 0 in stage1 table yields 0
# lookup 97 (0x61) in the first table in stage2 yields 23
# record 23 is { 20, 5, 12, 0, -32, 0, 9, 0 }
# 20 = ucp_Latin => Latin script
# lookup 97 (0x61) in the first table in stage2 yields 35
# record 35 is { 0, 5, 12, 0, -32, 0, 9, 22, 0 }
# 0 = ucp_Latin => Latin script
# 5 = ucp_Ll => Lower case letter
# 12 = ucp_gbOther => Grapheme break property "Other"
# 0 => Not part of a caseless set
# -32 (-0x20) => Other case is U+0041
# 0 => No special Script Extension property
# 9 = ucp_bidiL => Bidi class left-to-right
# 22 => Offset to Boolean properties
# 0 => Dummy value, unused at present
#
# Almost all lowercase latin characters resolve to the same record. One or two
@ -181,36 +188,38 @@
# example, k, K and the Kelvin symbol are such a set).
#
# Example: hiragana letter A (U+3042) is in block 96 (0x60)
# lookup 96 in stage1 table yields 91
# lookup 66 (0x42) in table 91 in stage2 yields 614
# record 614 is { 17, 7, 12, 0, 0, 0, 9, 0 }
# 17 = ucp_Hiragana => Hiragana script
# lookup 96 in stage1 table yields 93
# lookup 66 (0x42) in table 93 in stage2 yields 819
# record 614 is { 20, 7, 12, 0, 0, 0, 9, 41, 0 }
# 20 = ucp_Hiragana => Hiragana script
# 7 = ucp_Lo => Other letter
# 12 = ucp_gbOther => Grapheme break property "Other"
# 0 => Not part of a caseless set
# 0 => No other case
# 0 => No special Script Extension property
# 9 = ucp_bidiL => Bidi class left-to-right
# 41 => Offset to Boolean properties
# 0 => Dummy value, unused at present
#
# Example: vedic tone karshana (U+1CD0) is in block 57 (0x39)
# lookup 57 in stage1 table yields 55
# lookup 80 (0x50) in table 55 in stage2 yields 486
# record 485 is { 78, 12, 3, 0, 0, 138, 13, 0 }
# 78 = ucp_Inherited => Script inherited from predecessor
# lookup 80 (0x50) in table 55 in stage2 yields 621
# record 621 is { 84, 12, 3, 0, 0, 138, 13, 48, 0 }
# 84 = ucp_Inherited => Script inherited from predecessor
# 12 = ucp_Mn => Non-spacing mark
# 3 = ucp_gbExtend => Grapheme break property "Extend"
# 0 => Not part of a caseless set
# 0 => No other case
# 138 => Script Extension list offset = 138
# 13 = ucp_bidiNSM => Bidi class non-spacing mark
# 48 => Offset to Boolean properties
# 0 => Dummy value, unused at present
#
# At offset 138 in the ucd_script_sets vector we find a bitmap with bits 1, 8,
# 18, and 47 set. This means that this character is expected to be used with
# any of those scripts, which are Bengali, Devanagari, Kannada, and Grantha.
#
# Philip Hazel, last updated 31 December 2021.
# Philip Hazel, last updated 10 January 2022.
##############################################################################
@ -480,7 +489,7 @@ def write_bitsets(list, item_size):
unicode_version = ""
# Some of the tables imported from GenerateCommon.py have alternate comment
# strings for use by GenerateUcpHeader. The comments are now wanted here, so
# strings for use by GenerateUcpHeader. The comments are not wanted here, so
# remove them.
bidi_classes = bidi_classes[::2]
@ -522,22 +531,18 @@ for line in file:
break_props[i] = break_properties.index('Extended_Pictographic')
file.close()
# The Script Extensions property default value is the Script value. Parse the
# file, setting 'Unknown' as the default (this will never be a Script Extension
# value), then scan it and fill in the default from Scripts. Code added by PH
# in October 2018. Positive values are used for just a single script for a
# code point. Negative values are negated offsets in a list of bitsets of
# multiple scripts. Initialize this list with a single entry, as the zeroth
# element is never used.
# Handle script extensions. The get_script_extesion() function maintains a
# list of unique bitmaps representing lists of scripts, returning the offset
# in that list. Initialize the list with an empty set, which is used for
# characters that have no script extensions.
script_lists = [[]]
last_script_extension = ""
scriptx = read_table('Unicode.tables/ScriptExtensions.txt', get_script_extension, 0)
# Find the Boolean properties of each character. This next bit of magic creates
# a list of empty lists. Just using [[]] * MAX_UNICODE gives a list of
# references to the *same* list, which is not what we want.
# a list of empty lists. Using [[]] * MAX_UNICODE gives a list of references to
# the *same* list, which is not what we want.
bprops = [[] for _ in range(MAX_UNICODE)]
@ -601,7 +606,6 @@ for line in file:
file.close()
# Scan each character's boolean property list and created a list of unique
# lists, at the same time, setting the index in that list for each property in
# the bool_props vector.
@ -620,7 +624,6 @@ for c in range(MAX_UNICODE):
bool_props[c] = i
# With the addition of the Script Extensions field, we needed some padding to
# get the Unicode records up to 12 bytes (multiple of 4). Originally this was a
# 16-bit field and padding_dummy[0] was set to 256 to ensure this, but 8 bits
@ -864,14 +867,18 @@ for d in digitsets:
f.write("\n};\n\n")
f.write("""\
/* This vector is a list of script bitsets for the Script Extension property. */
/* This vector is a list of script bitsets for the Script Extension property.
The number of 32-bit words in each bitset is #defined in pcre2_ucp.h as
ucd_script_sets_item_size. */
const uint32_t PRIV(ucd_script_sets)[] = {
""")
write_bitsets(script_lists, script_list_item_size)
f.write("""\
/* This vector is a list of bitsets for Boolean properties. */
/* This vector is a list of bitsets for Boolean properties. The number of
32_bit words in each bitset is #defined as ucd_boolprop_sets_item_size in
pcre2_ucp.h. */
const uint32_t PRIV(ucd_boolprop_sets)[] = {
""")

View File

@ -42,7 +42,7 @@
# Note subsequent changes here:
#
# 27-December-2021: Added support for 4-letter script abbreviations.
# xx-January-2022: Further updates for Boolean property support
# 10-January-2022: Further updates for Boolean property support
# -----------------------------------------------------------------------------
@ -112,20 +112,18 @@ utt_table += list(zip(std_category_names, category_names, ['PT_PC'] * len(catego
utt_table += list(zip(std_general_category_names, general_category_names, ['PT_GC'] * len(general_category_names)))
utt_table += list(zip(std_bidi_class_names, bidi_class_names, ['PT_BIDICL'] * len(bidi_class_names)))
for name in bool_properties:
utt_table.append((stdname(name), name, 'PT_BOOL'))
if name in abbreviations:
for abbrev in abbreviations[name]:
utt_table.append((stdname(abbrev), name, 'PT_BOOL'))
# Now add specials and synonyms. Note both the standardized and capitalized
# forms are needed.
utt_table.append(('any', 'Any', 'PT_ANY'))
utt_table.append(('l&', 'L&', 'PT_LAMP'))
utt_table.append(('lc', 'LC', 'PT_LAMP'))
utt_table.append(('l&', 'L&', 'PT_LAMP'))
utt_table.append(('lc', 'LC', 'PT_LAMP'))
utt_table.append(('xan', 'Xan', 'PT_ALNUM'))
utt_table.append(('xps', 'Xps', 'PT_PXSPACE'))
utt_table.append(('xsp', 'Xsp', 'PT_SPACE'))

View File

@ -22,13 +22,14 @@ GenerateCommon.py
GenerateTest26.py
A Python script that generates input and expected output test data for test
26, which tests Unicode property support.
26, which tests certain aspects of Unicode property support.
GenerateUcd.py
A Python script that generates the file pcre2_ucd.c from GenerateCommon.py
and Unicode data files, which are themselves downloaded from the Unicode web
site. The generated file contains the tables for a 2-stage lookup of Unicode
properties, along with some auxiliary tables.
properties, along with some auxiliary tables. The script starts with a long
comment that gives details of the tables it constructs.
GenerateUcpHeader.py
A Python script that generates the file pcre2_ucp.h from GenerateCommon.py
@ -38,7 +39,7 @@ GenerateUcpHeader.py
GenerateUcpTables.py
A Python script that generates the file pcre2_ucptables.c from
GenerateCommon.py and Unicode data files. The generated file contains tables
for looking up Unicode properties.
for looking up Unicode property names.
ManyConfigTests
A shell script that runs "configure, make, test" a number of times with
@ -63,11 +64,11 @@ Unicode.tables
ScriptExtensions.txt are where to look for script information.
ucptest.c
A short C program for testing the Unicode property macros that do lookups in
the pcre2_ucd.c data, mainly useful after rebuilding the Unicode property
table. Compile and run this in the "maint" directory (see comments at its
head). This program can also be used to find characters with specific
properties.
A program for testing the Unicode property macros that do lookups in the
pcre2_ucd.c data, mainly useful after rebuilding the Unicode property tables.
Compile and run this in the "maint" directory (see comments at its head).
This program can also be used to find characters with specific properties and
to list which properties are supported.
ucptestdata
A directory containing four files, testinput{1,2} and testoutput{1,2}, for
@ -85,35 +86,18 @@ utf8.c
Updating to a new Unicode release
=================================
**** WORK IN PROGRESS ****
Work is going on in the area of Unicode property handling. What follows here is
now out-of-date. It will be updated once the current project is complete.
Updating to any new Unicode release is best left till then also.
06-January-2022
When there is a new release of Unicode, the files in Unicode.tables must be
refreshed from the web site. If the new version of Unicode adds new character
scripts, the lists in GenerateCommon.py must be updated. I have been adding
each new group at the end of the relevant list, with a comment.
refreshed from the web site. Once that is done, the four Python scripts that
generate files from the Unicode data can be run from within the "maint"
directory.
NOTE: Both the pcre2syntax.3 and pcre2pattern.3 man pages contain lists of
supported Unicode scripts that also have to be updated. These lists are in
alphabetical order.
Note: Previously, it was necessary to update lists of scripts and their
abbreviations by hand before running the Python scripts. This is no longer
necessary because the scripts have been upgraded to extract this information
themselves. Also, there used to be explicit lists of script in two of the man
pages. This is no longer the case.
There are two lists in GenerateCommon.py: the full names and the abbreviations
that are found in the ScriptExtensions.txt file. A list of script names and
their abbreviations can be found in the PropertyValueAliases.txt file on the
Unicode web site. There is also a Wikipedia page that lists them, and notes the
Unicode version in which they were introduced:
https://en.wikipedia.org/wiki/Unicode_scripts#Table_of_Unicode_scripts
Once the script name lists have been updated, the three generator scripts can
be run from within the maint directory. If you get the error "ValueError:
list.index(x): x not in list", the cause is usually a missing (or misspelt)
name in one of the lists. You can give an output file name as an argument, but
You can give an output file name as an argument to the following scripts, but
by default:
GenerateUcd.py creates pcre2_ucd.c )
@ -121,17 +105,18 @@ GenerateUcpHeader.py creates pcre2_ucp.h ) in the current directory
GenerateUcpTables.py creates pcre2_ucptables.c )
These files can be compared against the existing versions in the src directory
to check on any changes before replacing the old files, but you can generate
directly into the final location by running
to check on any changes before replacing the old files, but you can also
generate directly into the final location by running:
./GenerateUcd.py ../src/pcre2_ucd.c
./GenerateUcpHeader.py ../src/pcre2_ucp.h
./GenerateUcpTables.py ../src/pcre2_ucptables.c
The ucptest program can be compiled and used to check that the new tables work
properly, using the data files in ucptestdata to check a number of test
characters. See the comments at the start of ucptest.c. If there are new
scripts, adding a few tests to the files in ucptestdata is a good idea.
Once the .c and .h files are in the ../src directory, the ucptest program can
be compiled and used to check that the new tables work properly. The data files
in ucptestdata are set up to check a number of test characters. See the
comments at the start of ucptest.c. If there are new scripts, adding a few
tests to the files in ucptestdata is a good idea.
Finally, you should run the GenerateTest26.py script to regenerate new versions
of the input and expected output from a series of Unicode property tests that
@ -476,4 +461,4 @@ years.
Philip Hazel
Email local part: Philip.Hazel
Email domain: gmail.com
Last updated: 31 December 2021
Last updated: 10 January 2022

View File

@ -168,7 +168,9 @@ const uint32_t PRIV(ucd_digit_sets)[] = {
0x1e959, 0x1fbf9,
};
/* This vector is a list of script bitsets for the Script Extension property. */
/* This vector is a list of script bitsets for the Script Extension property.
The number of 32-bit words in each bitset is #defined in pcre2_ucp.h as
ucd_script_sets_item_size. */
const uint32_t PRIV(ucd_script_sets)[] = {
0x00000000u, 0x00000000u, 0x00000000u,
@ -236,7 +238,9 @@ const uint32_t PRIV(ucd_script_sets)[] = {
0x2000ffc0u, 0x3984a010u, 0x00000001u,
};
/* This vector is a list of bitsets for Boolean properties. */
/* This vector is a list of bitsets for Boolean properties. The number of
32_bit words in each bitset is #defined as ucd_boolprop_sets_item_size in
pcre2_ucp.h. */
const uint32_t PRIV(ucd_boolprop_sets)[] = {
0x00000000u, 0x00000000u,