From 1b068cedf3c113f27121182b29e096ccb3e1b9c1 Mon Sep 17 00:00:00 2001 From: Thomas Voss Date: Fri, 3 May 2024 18:37:50 +0200 Subject: Add more 2-stage lookup tables --- gen/prop/gcb | 2 +- gen/prop/hst | 2 +- gen/prop/inpc | 2 +- gen/prop/insc | 164 +++++++++++++++++++++++++++++------------------- gen/prop/jg | 153 +++++++++++++++++++++++++++++---------------- gen/prop/jt | 163 +++++++++++++++++++++++++++++------------------- gen/prop/lb | 166 ++++++++++++++++++++++++++++++------------------- gen/prop/nfXX_qc | 185 ++++++++++++++++++++++++++++++++----------------------- gen/prop/nt | 168 +++++++++++++++++++++++++++++++------------------- 9 files changed, 619 insertions(+), 386 deletions(-) (limited to 'gen/prop') diff --git a/gen/prop/gcb b/gen/prop/gcb index 4565d40..06685b0 100755 --- a/gen/prop/gcb +++ b/gen/prop/gcb @@ -104,7 +104,7 @@ def main() -> None: cs = set(Cs) sz_s1 = len(Cs) * isize(len(cs) - 1) - sz_s2 = len(cs) * bs * 2 + sz_s2 = len(cs) * bs sz = sz_s1 + sz_s2 if sz < smallest: diff --git a/gen/prop/hst b/gen/prop/hst index a2765fd..8af4a4b 100755 --- a/gen/prop/hst +++ b/gen/prop/hst @@ -83,7 +83,7 @@ def main() -> None: cs = set(Cs) sz_s1 = len(Cs) * isize(len(cs) - 1) - sz_s2 = len(cs) * bs * 2 + sz_s2 = len(cs) * bs sz = sz_s1 + sz_s2 if sz < smallest: diff --git a/gen/prop/inpc b/gen/prop/inpc index 6a8561f..c4d35ef 100755 --- a/gen/prop/inpc +++ b/gen/prop/inpc @@ -83,7 +83,7 @@ def main() -> None: cs = set(Cs) sz_s1 = len(Cs) * isize(len(cs) - 1) - sz_s2 = len(cs) * bs * 2 + sz_s2 = len(cs) * bs sz = sz_s1 + sz_s2 if sz < smallest: diff --git a/gen/prop/insc b/gen/prop/insc index 7ed2b96..9ee3a34 100755 --- a/gen/prop/insc +++ b/gen/prop/insc @@ -1,63 +1,101 @@ -#!/bin/sh - -set -e -cd "${0%/*}/../.." -exec >lib/unicode/prop/uprop_get_insc.c - -gawk ' -BEGIN { - FS = "( *#.*| +; +)" - - print "/* This file is autogenerated by gen/prop/insc; DO NOT EDIT. */" - print "" - print "#include \"_bsearch.h\"" - print "#include \"macros.h\"" - print "#include \"rune.h\"" - print "#include \"unicode/prop.h\"" - print "" -} - -/^[^#]/ { - n = split($1, a, /\.\./) - lo = strtonum("0X" a[1]) - hi = strtonum("0X" a[n]) - - for (i = lo; i <= hi; i++) { - gsub(/^; /, "", $2) - props[i] = "INSC_" toupper($2) - } -} - -END { - print "static constexpr enum uprop_insc lookup_lat1[LATIN1_MAX] = {" - for (i = 0; i < 0x100; i++) { - if (props[i]) - printf "\t[0x%02X] = %s,\n", i, props[i] - } - print "};" - print "" - print "static const struct {" - print "\trune lo, hi;" - print "\tenum uprop_insc val;" - print "} lookup[] = {" - - for (i = 0x100; i <= 0x10FFFF; i++) { - if (!props[i]) - continue - lo = i - while (props[lo] == props[i + 1]) - i++ - printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i] - } - - print "};" - print "" - print "_MLIB_DEFINE_BSEARCH(enum uprop_insc, lookup, INSC_OTHER)" - print "" - print "enum uprop_insc" - print "uprop_get_insc(rune ch)" - print "{" - print "\treturn ch < lengthof(lookup_lat1) ? lookup_lat1[ch] : mlib_lookup(ch);" - print "}" -} -' data/IndicSyllabicCategory | sed 's/\s*$//' +#!/usr/bin/python3 + +import math + +from lib import * + + +longest = 0 + +def parse(file: str) -> list[bool]: + global longest + + xs = ['INSC_OTHER'] * 0x110000 + with open(file, 'r') as f: + for line in f.readlines(): + if len(line.strip()) == 0 or line[0] == '#': + continue + + parts = line.split(';') + ranges = [int(x, 16) for x in parts[0].strip().split('..')] + prop = 'INSC_' + parts[1].split('#')[0].strip().upper() + longest = max(longest, len(prop)) + + for i in range(ranges[0], ranges[len(ranges) - 1] + 1): + xs[i] = prop + return xs + +def genfile(cs: list[tuple[bool, ...]], blksize: int) -> None: + Cs = cs + cs = list(dict.fromkeys(Cs)) + + print('''\ +/* This file is autogenerated by gen/prop/insc; DO NOT EDIT. */ + +#include "unicode/prop.h" +''') + + print(f'static constexpr {typename(len(cs) - 1)} stage1[] = {{') + for i, c in enumerate(Cs): + print(f'%c%{len(str(len(cs) - 1))}d,' % ('\t' if i % 16 == 0 else ' ', cs.index(c)), end='') + if i % 16 == 15: + print() + print('};') + + print() + + ppc = columns(blksize, longest + 1) + print(f'static constexpr enum uprop_insc stage2[][{blksize}] = {{') + for c in cs: + for i in range(blksize // ppc): + print('\t{' if i == 0 else '\t ', end='') + for j in range(ppc): + print(c[i*ppc + j], end='') + if i < blksize // ppc - 1 or j < ppc - 1: + print(',', end='') + if j < ppc - 1: + print(' ' * (longest + 1 - len(c[i*ppc + j])), end='') + if i < blksize // ppc - 1: + print() + print('},') + print('};') + + print() + + print(f'''\ +enum uprop_insc +uprop_get_insc(rune ch) +{{ + return stage2[stage1[ch / {blksize}]][ch % {blksize}]; +}}''') + +def main() -> None: + cwd_init() + xs = parse('data/IndicSyllabicCategory') + + blksize = -1 + smallest = math.inf + + for bs in powers_of_2(): + if bs > len(xs): + break + Cs = [tuple(x) for x in chunks(xs, bs)] + cs = set(Cs) + + sz_s1 = len(Cs) * isize(len(cs) - 1) + sz_s2 = len(cs) * bs + sz = sz_s1 + sz_s2 + + if sz < smallest: + smallest = sz + blksize = bs + + Cs = [tuple(x) for x in chunks(xs, blksize)] + with open('lib/unicode/prop/uprop_get_insc.c', 'w') as f: + sys.stdout = f + genfile(Cs, blksize) + + report_size(len(xs), smallest) + +if __name__ == '__main__': + main() diff --git a/gen/prop/jg b/gen/prop/jg index e88442d..c0cc31a 100755 --- a/gen/prop/jg +++ b/gen/prop/jg @@ -1,52 +1,101 @@ -#!/bin/sh - -set -e -cd "${0%/*}/../.." -exec >lib/unicode/prop/uprop_get_jg.c - -gawk ' -BEGIN { - FS = "[ ;]+" - - print "/* This file is autogenerated by gen/prop/jg; DO NOT EDIT. */" - print "" - print "#include \"_bsearch.h\"" - print "#include \"rune.h\"" - print "#include \"unicode/prop.h\"" - print "" -} - -/^[A-F0-9]/ { - n = split($1, a, /\.\./) - lo = strtonum("0X" a[1]) - hi = strtonum("0X" a[n]) - - for (i = lo; i <= hi; i++) - props[i] = "JG_" toupper($2) -} - -END { - print "static const struct {" - print "\trune lo, hi;" - print "\tenum uprop_jg val;" - print "} lookup[] = {" - - for (i = 0; i <= 0x10FFFF; i++) { - if (!props[i]) - continue - for (lo = i; props[lo] == props[i + 1]; i++) - ; - printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i] - } - - print "};" - print "" - print "_MLIB_DEFINE_BSEARCH(enum uprop_jg, lookup, JG_NO_JOINING_GROUP)" - print "" - print "enum uprop_jg" - print "uprop_get_jg(rune ch)" - print "{" - print "\treturn ch < lookup[0].lo ? JG_NO_JOINING_GROUP : mlib_lookup(ch);" - print "}" -} -' data/DerivedJoiningGroup | sed 's/\s*$//' +#!/usr/bin/python3 + +import math + +from lib import * + + +longest = 0 + +def parse(file: str) -> list[bool]: + global longest + + xs = ['JG_NO_JOINING_GROUP'] * 0x110000 + with open(file, 'r') as f: + for line in f.readlines(): + if len(line.strip()) == 0 or line[0] == '#': + continue + + parts = line.split(';') + ranges = [int(x, 16) for x in parts[0].strip().split('..')] + prop = 'JG_' + parts[1].split('#')[0].strip().upper() + longest = max(longest, len(prop)) + + for i in range(ranges[0], ranges[len(ranges) - 1] + 1): + xs[i] = prop + return xs + +def genfile(cs: list[tuple[bool, ...]], blksize: int) -> None: + Cs = cs + cs = list(dict.fromkeys(Cs)) + + print('''\ +/* This file is autogenerated by gen/prop/jg; DO NOT EDIT. */ + +#include "unicode/prop.h" +''') + + print(f'static constexpr {typename(len(cs) - 1)} stage1[] = {{') + for i, c in enumerate(Cs): + print(f'%c%{len(str(len(cs) - 1))}d,' % ('\t' if i % 16 == 0 else ' ', cs.index(c)), end='') + if i % 16 == 15: + print() + print('};') + + print() + + ppc = columns(blksize, longest + 1) + print(f'static constexpr enum uprop_jg stage2[][{blksize}] = {{') + for c in cs: + for i in range(blksize // ppc): + print('\t{' if i == 0 else '\t ', end='') + for j in range(ppc): + print(c[i*ppc + j], end='') + if i < blksize // ppc - 1 or j < ppc - 1: + print(',', end='') + if j < ppc - 1: + print(' ' * (longest + 1 - len(c[i*ppc + j])), end='') + if i < blksize // ppc - 1: + print() + print('},') + print('};') + + print() + + print(f'''\ +enum uprop_jg +uprop_get_jg(rune ch) +{{ + return stage2[stage1[ch / {blksize}]][ch % {blksize}]; +}}''') + +def main() -> None: + cwd_init() + xs = parse('data/DerivedJoiningGroup') + + blksize = -1 + smallest = math.inf + + for bs in powers_of_2(): + if bs > len(xs): + break + Cs = [tuple(x) for x in chunks(xs, bs)] + cs = set(Cs) + + sz_s1 = len(Cs) * isize(len(cs) - 1) + sz_s2 = len(cs) * bs + sz = sz_s1 + sz_s2 + + if sz < smallest: + smallest = sz + blksize = bs + + Cs = [tuple(x) for x in chunks(xs, blksize)] + with open('lib/unicode/prop/uprop_get_jg.c', 'w') as f: + sys.stdout = f + genfile(Cs, blksize) + + report_size(len(xs), smallest) + +if __name__ == '__main__': + main() diff --git a/gen/prop/jt b/gen/prop/jt index d8ca9ff..23a4c9e 100755 --- a/gen/prop/jt +++ b/gen/prop/jt @@ -1,62 +1,101 @@ -#!/bin/sh - -set -e -cd "${0%/*}/../.." -exec >lib/unicode/prop/uprop_get_jt.c - -gawk ' -BEGIN { - FS = "[ ;]+" - - print "/* This file is autogenerated by gen/prop/jt; DO NOT EDIT. */" - print "" - print "#include \"_bsearch.h\"" - print "#include \"macros.h\"" - print "#include \"rune.h\"" - print "#include \"unicode/prop.h\"" - print "" -} - -/^[A-F0-9]/ { - n = split($1, a, /\.\./) - lo = strtonum("0X" a[1]) - hi = strtonum("0X" a[n]) - - for (i = lo; i <= hi; i++) - props[i] = "JT_" $2 -} - -END { - print "static constexpr enum uprop_jt lookup_lat1[] = {" - for (i = 0; i < 0x100; i++) { - if (i % 8 == 0) - printf "\t" - printf "%s%s", (props[i] ? props[i] : "JT_U") ",", \ - i % 8 == 7 ? "\n" : " " - } - print "};" - print "" - print "static const struct {" - print "\trune lo, hi;" - print "\tenum uprop_jt val;" - print "} lookup[] = {" - - for (i = 0x100; i <= 0x10FFFF; i++) { - if (!props[i]) - continue - for (lo = i; props[lo] == props[i + 1]; i++) - ; - printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i] - } - - print "};" - print "" - print "_MLIB_DEFINE_BSEARCH(enum uprop_jt, lookup, JT_U)" - print "" - print "enum uprop_jt" - print "uprop_get_jt(rune ch)" - print "{" - print "\treturn ch < lengthof(lookup_lat1) ? lookup_lat1[ch] : mlib_lookup(ch);" - print "}" -} -' data/DerivedJoiningType | sed 's/\s*$//' +#!/usr/bin/python3 + +import math + +from lib import * + + +longest = 0 + +def parse(file: str) -> list[bool]: + global longest + + xs = ['JT_U'] * 0x110000 + with open(file, 'r') as f: + for line in f.readlines(): + if len(line.strip()) == 0 or line[0] == '#': + continue + + parts = line.split(';') + ranges = [int(x, 16) for x in parts[0].strip().split('..')] + prop = 'JT_' + parts[1].split('#')[0].strip() + longest = max(longest, len(prop)) + + for i in range(ranges[0], ranges[len(ranges) - 1] + 1): + xs[i] = prop + return xs + +def genfile(cs: list[tuple[bool, ...]], blksize: int) -> None: + Cs = cs + cs = list(dict.fromkeys(Cs)) + + print('''\ +/* This file is autogenerated by gen/prop/jt; DO NOT EDIT. */ + +#include "unicode/prop.h" +''') + + print(f'static constexpr {typename(len(cs) - 1)} stage1[] = {{') + for i, c in enumerate(Cs): + print(f'%c%{len(str(len(cs) - 1))}d,' % ('\t' if i % 16 == 0 else ' ', cs.index(c)), end='') + if i % 16 == 15: + print() + print('};') + + print() + + ppc = columns(blksize, longest + 1) + print(f'static constexpr enum uprop_jt stage2[][{blksize}] = {{') + for c in cs: + for i in range(blksize // ppc): + print('\t{' if i == 0 else '\t ', end='') + for j in range(ppc): + print(c[i*ppc + j], end='') + if i < blksize // ppc - 1 or j < ppc - 1: + print(',', end='') + if j < ppc - 1: + print(' ' * (longest + 1 - len(c[i*ppc + j])), end='') + if i < blksize // ppc - 1: + print() + print('},') + print('};') + + print() + + print(f'''\ +enum uprop_jt +uprop_get_jt(rune ch) +{{ + return stage2[stage1[ch / {blksize}]][ch % {blksize}]; +}}''') + +def main() -> None: + cwd_init() + xs = parse('data/DerivedJoiningType') + + blksize = -1 + smallest = math.inf + + for bs in powers_of_2(): + if bs > len(xs): + break + Cs = [tuple(x) for x in chunks(xs, bs)] + cs = set(Cs) + + sz_s1 = len(Cs) * isize(len(cs) - 1) + sz_s2 = len(cs) * bs + sz = sz_s1 + sz_s2 + + if sz < smallest: + smallest = sz + blksize = bs + + Cs = [tuple(x) for x in chunks(xs, blksize)] + with open('lib/unicode/prop/uprop_get_jt.c', 'w') as f: + sys.stdout = f + genfile(Cs, blksize) + + report_size(len(xs), smallest) + +if __name__ == '__main__': + main() diff --git a/gen/prop/lb b/gen/prop/lb index 2abab6c..e003d13 100755 --- a/gen/prop/lb +++ b/gen/prop/lb @@ -1,65 +1,101 @@ -#!/bin/sh - -set -e -cd "${0%/*}/../.." -exec >lib/unicode/prop/uprop_get_lb.c - -gawk ' -BEGIN { - FS = "( *#.*| *; +)" - - print "/* This file is autogenerated by gen/prop/lb; DO NOT EDIT. */" - print "" - print "#include \"_bsearch.h\"" - print "#include \"macros.h\"" - print "#include \"rune.h\"" - print "#include \"unicode/prop.h\"" - print "" -} - -/^[^#]/ { - n = split($1, a, /\.\./) - lo = strtonum("0X" a[1]) - hi = strtonum("0X" a[n]) - - for (i = lo; i <= hi; i++) { - gsub(/^; /, "", $2) - props[i] = "LB_" $2 - } -} - -END { - print "static constexpr enum uprop_lb lookup_lat1[] = {" - for (i = 0; i < 0x100; i++) { - if (i % 8 == 0) - printf "\t" - printf "%-6s%s", (props[i] ? props[i] : "LB_XX") ",", \ - i % 8 == 7 ? "\n" : " " - } - print "};" - print "" - print "static const struct {" - print "\trune lo, hi;" - print "\tenum uprop_lb val;" - print "} lookup[] = {" - - for (i = 0x100; i <= 0x10FFFF; i++) { - if (!props[i]) - continue - lo = i - while (props[lo] == props[i + 1]) - i++ - printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i] - } - - print "};" - print "" - print "_MLIB_DEFINE_BSEARCH(enum uprop_lb, lookup, LB_XX)" - print "" - print "enum uprop_lb" - print "uprop_get_lb(rune ch)" - print "{" - print "\treturn ch < lengthof(lookup_lat1) ? lookup_lat1[ch] : mlib_lookup(ch);" - print "}" -} -' data/DerivedLineBreak | sed 's/\s*$//' +#!/usr/bin/python3 + +import math + +from lib import * + + +longest = 0 + +def parse(file: str) -> list[bool]: + global longest + + xs = ['LB_XX'] * 0x110000 + with open(file, 'r') as f: + for line in f.readlines(): + if len(line.strip()) == 0 or line[0] == '#': + continue + + parts = line.split(';') + ranges = [int(x, 16) for x in parts[0].strip().split('..')] + prop = 'LB_' + parts[1].split('#')[0].strip() + longest = max(longest, len(prop)) + + for i in range(ranges[0], ranges[len(ranges) - 1] + 1): + xs[i] = prop + return xs + +def genfile(cs: list[tuple[bool, ...]], blksize: int) -> None: + Cs = cs + cs = list(dict.fromkeys(Cs)) + + print('''\ +/* This file is autogenerated by gen/prop/lb; DO NOT EDIT. */ + +#include "unicode/prop.h" +''') + + print(f'static constexpr {typename(len(cs) - 1)} stage1[] = {{') + for i, c in enumerate(Cs): + print(f'%c%{len(str(len(cs) - 1))}d,' % ('\t' if i % 16 == 0 else ' ', cs.index(c)), end='') + if i % 16 == 15: + print() + print('};') + + print() + + ppc = columns(blksize, longest + 1) + print(f'static constexpr enum uprop_lb stage2[][{blksize}] = {{') + for c in cs: + for i in range(blksize // ppc): + print('\t{' if i == 0 else '\t ', end='') + for j in range(ppc): + print(c[i*ppc + j], end='') + if i < blksize // ppc - 1 or j < ppc - 1: + print(',', end='') + if j < ppc - 1: + print(' ' * (longest + 1 - len(c[i*ppc + j])), end='') + if i < blksize // ppc - 1: + print() + print('},') + print('};') + + print() + + print(f'''\ +enum uprop_lb +uprop_get_lb(rune ch) +{{ + return stage2[stage1[ch / {blksize}]][ch % {blksize}]; +}}''') + +def main() -> None: + cwd_init() + xs = parse('data/DerivedLineBreak') + + blksize = -1 + smallest = math.inf + + for bs in powers_of_2(): + if bs > len(xs): + break + Cs = [tuple(x) for x in chunks(xs, bs)] + cs = set(Cs) + + sz_s1 = len(Cs) * isize(len(cs) - 1) + sz_s2 = len(cs) * bs + sz = sz_s1 + sz_s2 + + if sz < smallest: + smallest = sz + blksize = bs + + Cs = [tuple(x) for x in chunks(xs, blksize)] + with open('lib/unicode/prop/uprop_get_lb.c', 'w') as f: + sys.stdout = f + genfile(Cs, blksize) + + report_size(len(xs), smallest) + +if __name__ == '__main__': + main() diff --git a/gen/prop/nfXX_qc b/gen/prop/nfXX_qc index fed603e..24fcf45 100755 --- a/gen/prop/nfXX_qc +++ b/gen/prop/nfXX_qc @@ -1,77 +1,110 @@ -#!/bin/sh - -set -e -cd "${0%/*}/../.." - -for x in c d kc kd -do - gawk -v s=$x ' - BEGIN { - FS = "[ ;]+" - - _default = "NF" toupper(s) "_QC_Y" - - print "/* This file is autogenerated by gen/prop/nfXX_qc; DO NOT EDIT. */" - print "" - print "/* The macros.h include may be unused */" - print "" - print "#include \"_bsearch.h\"" - print "#include \"macros.h\"" - print "#include \"rune.h\"" - print "#include \"unicode/prop.h\"" - print "" - } - - /^[A-F0-9]/ && $2 == "NF" toupper(s) "_QC" { - n = split($1, a, /\.\./) - lo = strtonum("0X" a[1]) - hi = strtonum("0X" a[n]) - - for (i = lo; i <= hi; i++) - props[i] = "NF" toupper(s) "_QC_" $3 - if (lo < 0x100) - want_lat1_tbl = 1 - } - - END { - if (want_lat1_tbl) { - print "static constexpr enum uprop_nf" s "_qc lookup_lat1[] = {" - for (i = 0; i < 0x100; i++) { - if (i % 8 == 0) - printf "\t" - printf "%s,%s", props[i] ? props[i] : _default, \ - i % 8 == 7 ? "\n" : " " - } - print "};" - print "" - } - - print "static const struct {" - print "\trune lo, hi;" - print "\tenum uprop_nf" s "_qc val;" - print "} lookup[] = {" - - for (i = want_lat1_tbl ? 0x100 : 0; i <= 0x10FFFF; i++) { - if (!props[i]) +#!/usr/bin/python3 + +import math + +from lib import * + + +longest = 0 +TYPES = ['c', 'd', 'kc', 'kd'] + +def parse(file: str, _type: str) -> list[bool]: + global longest + + _type = _type.upper() + + xs = [f'NF{_type}_QC_Y'] * 0x110000 + with open(file, 'r') as f: + for line in f.readlines(): + if ( + len(line.strip()) == 0 + or line[0] == '#' + or f'NF{_type}_QC' not in line + ): continue - for (lo = i; props[lo] == props[i + 1]; i++) - ; - printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i] - } - - print "};" - print "" - print "_MLIB_DEFINE_BSEARCH(enum uprop_nf" s "_qc, lookup, " _default ")" - print "" - print "enum uprop_nf" s "_qc" - print "uprop_get_nf" s "_qc(rune ch)" - print "{" - if (want_lat1_tbl) - print "\treturn ch < lengthof(lookup_lat1) ? lookup_lat1[ch] : mlib_lookup(ch);" - else - print "\treturn ch < lookup[0].lo ? " _default " : mlib_lookup(ch);" - print "}" - } - ' data/DerivedNormalizationProps \ - | sed 's/\s*$//' >lib/unicode/prop/uprop_get_nf${x}_qc.c -done + + parts = line.split(';') + ranges = [int(x, 16) for x in parts[0].strip().split('..')] + prop = f'NF{_type}_QC_' + parts[2].split('#')[0].strip() + longest = max(longest, len(prop)) + + for i in range(ranges[0], ranges[len(ranges) - 1] + 1): + xs[i] = prop + return xs + +def genfile(cs: list[tuple[bool, ...]], blksize: int, _type: str) -> None: + Cs = cs + cs = list(dict.fromkeys(Cs)) + + print('''\ +/* This file is autogenerated by gen/prop/nfXX_qc; DO NOT EDIT. */ + +#include "unicode/prop.h" +''') + + print(f'static constexpr {typename(len(cs) - 1)} stage1[] = {{') + for i, c in enumerate(Cs): + print(f'%c%{len(str(len(cs) - 1))}d,' % ('\t' if i % 16 == 0 else ' ', cs.index(c)), end='') + if i % 16 == 15: + print() + print('};') + + print() + + ppc = columns(blksize, longest + 1) + print(f'static constexpr enum uprop_nf{_type}_qc stage2[][{blksize}] = {{') + for c in cs: + for i in range(blksize // ppc): + print('\t{' if i == 0 else '\t ', end='') + for j in range(ppc): + print(c[i*ppc + j], end='') + if i < blksize // ppc - 1 or j < ppc - 1: + print(',', end='') + if j < ppc - 1: + print(' ' * (longest + 1 - len(c[i*ppc + j])), end='') + if i < blksize // ppc - 1: + print() + print('},') + print('};') + + print() + + print(f'''\ +enum uprop_nf{_type}_qc +uprop_get_nf{_type}_qc(rune ch) +{{ + return stage2[stage1[ch / {blksize}]][ch % {blksize}]; +}}''') + +def main(_type: str) -> None: + cwd_init() + xs = parse('data/DerivedNormalizationProps', _type) + + blksize = -1 + smallest = math.inf + + for bs in powers_of_2(): + if bs > len(xs): + break + Cs = [tuple(x) for x in chunks(xs, bs)] + cs = set(Cs) + + sz_s1 = len(Cs) * isize(len(cs) - 1) + sz_s2 = len(cs) * bs + sz = sz_s1 + sz_s2 + + if sz < smallest: + smallest = sz + blksize = bs + + Cs = [tuple(x) for x in chunks(xs, blksize)] + with open(f'lib/unicode/prop/uprop_get_nf{_type}_qc.c', 'w') as f: + sys.stdout = f + genfile(Cs, blksize, _type) + + report_size(len(xs), smallest) + +if __name__ == '__main__': + for _type in TYPES: + longest = 0 + main(_type) diff --git a/gen/prop/nt b/gen/prop/nt index 835bc3f..477789c 100755 --- a/gen/prop/nt +++ b/gen/prop/nt @@ -1,69 +1,107 @@ -#!/bin/sh - -set -e -cd "${0%/*}/../.." -exec >lib/unicode/prop/uprop_get_nt.c - -gawk ' -BEGIN { - FS = "( *#.*| +; +)" - - map["Decimal"] = "DE" - map["Digit"] = "DI" - map["Numeric"] = "NU" - - print "/* This file is autogenerated by gen/prop/nt; DO NOT EDIT. */" - print "" - print "#include \"_bsearch.h\"" - print "#include \"macros.h\"" - print "#include \"rune.h\"" - print "#include \"unicode/prop.h\"" - print "" -} +#!/usr/bin/python3 -/^[^#]/ { - n = split($1, a, /\.\./) - lo = strtonum("0X" a[1]) - hi = strtonum("0X" a[n]) +import math + +from lib import * - for (i = lo; i <= hi; i++) { - gsub(/^; /, "", $2) - props[i] = "NT_" map[$2] - } -} -END { - print "static constexpr enum uprop_nt lookup_lat1[] = {" - for (i = 0; i < 0x100; i++) { - if (i % 8 == 0) - printf "\t" - printf "%-8s%s", (props[i] ? props[i] : "NT_NONE") ",", \ - i % 8 == 7 ? "\n" : " " - } - print "};" - print "" - print "static const struct {" - print "\trune lo, hi;" - print "\tenum uprop_nt val;" - print "} lookup[] = {" - - for (i = 0x100; i <= 0x10FFFF; i++) { - if (!props[i]) - continue - lo = i - while (props[lo] == props[i + 1]) - i++ - printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i] - } - - print "};" - print "" - print "_MLIB_DEFINE_BSEARCH(enum uprop_nt, lookup, NT_NONE)" - print "" - print "enum uprop_nt" - print "uprop_get_nt(rune ch)" - print "{" - print "\treturn ch < lengthof(lookup_lat1) ? lookup_lat1[ch] : mlib_lookup(ch);" - print "}" +MAP = { + 'Decimal': 'DE', + 'Digit': 'DI', + 'Numeric': 'NU', } -' data/DerivedNumericType | sed 's/\s*$//' + +longest = 0 + +def parse(file: str) -> list[bool]: + global longest + + xs = ['NT_NONE'] * 0x110000 + with open(file, 'r') as f: + for line in f.readlines(): + if len(line.strip()) == 0 or line[0] == '#': + continue + + parts = line.split(';') + ranges = [int(x, 16) for x in parts[0].strip().split('..')] + prop = 'NT_' + MAP[parts[1].split('#')[0].strip()] + longest = max(longest, len(prop), len('NT_NONE')) + + for i in range(ranges[0], ranges[len(ranges) - 1] + 1): + xs[i] = prop + return xs + +def genfile(cs: list[tuple[bool, ...]], blksize: int) -> None: + Cs = cs + cs = list(dict.fromkeys(Cs)) + + print('''\ +/* This file is autogenerated by gen/prop/nt; DO NOT EDIT. */ + +#include "unicode/prop.h" +''') + + print(f'static constexpr {typename(len(cs) - 1)} stage1[] = {{') + for i, c in enumerate(Cs): + print(f'%c%{len(str(len(cs) - 1))}d,' % ('\t' if i % 16 == 0 else ' ', cs.index(c)), end='') + if i % 16 == 15: + print() + print('};') + + print() + + ppc = columns(blksize, longest + 1) + print(f'static constexpr enum uprop_nt stage2[][{blksize}] = {{') + for c in cs: + for i in range(blksize // ppc): + print('\t{' if i == 0 else '\t ', end='') + for j in range(ppc): + print(c[i*ppc + j], end='') + if i < blksize // ppc - 1 or j < ppc - 1: + print(',', end='') + if j < ppc - 1: + print(' ' * (longest + 1 - len(c[i*ppc + j])), end='') + if i < blksize // ppc - 1: + print() + print('},') + print('};') + + print() + + print(f'''\ +enum uprop_nt +uprop_get_nt(rune ch) +{{ + return stage2[stage1[ch / {blksize}]][ch % {blksize}]; +}}''') + +def main() -> None: + cwd_init() + xs = parse('data/DerivedNumericType') + + blksize = -1 + smallest = math.inf + + for bs in powers_of_2(): + if bs > len(xs): + break + Cs = [tuple(x) for x in chunks(xs, bs)] + cs = set(Cs) + + sz_s1 = len(Cs) * isize(len(cs) - 1) + sz_s2 = len(cs) * bs + sz = sz_s1 + sz_s2 + + if sz < smallest: + smallest = sz + blksize = bs + + Cs = [tuple(x) for x in chunks(xs, blksize)] + with open('lib/unicode/prop/uprop_get_nt.c', 'w') as f: + sys.stdout = f + genfile(Cs, blksize) + + report_size(len(xs), smallest) + +if __name__ == '__main__': + main() -- cgit v1.2.3