Add more 2-stage lookup tables

3 files changed, 320 insertions, 189 deletions

diff --git a/gen/prop/gcb b/gen/prop/gcb
index 17d3136..4565d40 100755
--- a/gen/prop/gcb
+++ b/gen/prop/gcb
@@ -1,83 +1,122 @@
-#!/bin/sh
-
-set -e
-cd "${0%/*}/../.."
-exec >lib/unicode/prop/uprop_get_gcb.c
-
-gawk '
-BEGIN {
-	FS = "( *#.*| *; +)"
-
-	map["Control"]            = "CN"
-	map["CR"]                 = "CR"
-	map["E_Base"]             = "EB"
-	map["E_Base_GAZ"]         = "EBG"
-	map["E_Modifier"]         = "EM"
-	map["Extend"]             = "EX"
-	map["Glue_After_Zwj"]     = "GAZ"
-	map["LF"]                 = "LF"
-	map["L"]                  = "L"
-	map["LV"]                 = "LV"
-	map["LVT"]                = "LVT"
-	map["Other"]              = "XX"
-	map["Prepend"]            = "PP"
-	map["Regional_Indicator"] = "RI"
-	map["SpacingMark"]        = "SM"
-	map["T"]                  = "T"
-	map["V"]                  = "V"
-	map["ZWJ"]                = "ZWJ"
-
-	print "/* This file is autogenerated by gen/prop/gcb; 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] = "GCB_" map[$2]
-	}
-}
 
-END {
-	print "static constexpr enum uprop_gcb lookup_lat1[] = {"
-	for (i = 0; i < 0x100; i++) {
-		if (i % 8 == 0)
-			printf "\t"
-		printf "%-7s%s", (props[i] ? props[i] : "GCB_XX") ",", \
-			i % 8 == 7 ? "\n" : " "
-	}
-	print "};"
-	print ""
-	print "static const struct {"
-	print "\trune lo, hi;"
-	print "\tenum uprop_gcb 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_gcb, lookup, GCB_XX)"
-	print ""
-	print "enum uprop_gcb"
-	print "uprop_get_gcb(rune ch)"
-	print "{"
-	print "\treturn ch < lengthof(lookup_lat1) ? lookup_lat1[ch] : mlib_lookup(ch);"
-	print "}"
+MAP = {
+	'Control':            'CN',
+	'CR':                 'CR',
+	'E_Base':             'EB',
+	'E_Base_GAZ':         'EBG',
+	'E_Modifier':         'EM',
+	'Extend':             'EX',
+	'Glue_After_Zwj':     'GAZ',
+	'LF':                 'LF',
+	'L':                  'L',
+	'LV':                 'LV',
+	'LVT':                'LVT',
+	'Other':              'XX',
+	'Prepend':            'PP',
+	'Regional_Indicator': 'RI',
+	'SpacingMark':        'SM',
+	'T':                  'T',
+	'V':                  'V',
+	'ZWJ':                'ZWJ',
 }
-' data/GraphemeBreakProperty | sed 's/\s*$//'
+
+longest = 0
+
+def parse(file: str) -> list[bool]:
+	global longest
+
+	xs = ['GCB_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 = 'GCB_' + MAP[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/gcb; 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_gcb 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_gcb
+uprop_get_gcb(rune ch)
+{{
+	return stage2[stage1[ch / {blksize}]][ch % {blksize}];
+}}''')
+
+def main() -> None:
+	cwd_init()
+	xs = parse('data/GraphemeBreakProperty')
+
+	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 * 2
+		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_gcb.c', 'w') as f:
+		sys.stdout = f
+		genfile(Cs, blksize)
+
+	report_size(len(xs), smallest)
+
+if __name__ == '__main__':
+	main()
diff --git a/gen/prop/hst b/gen/prop/hst
index 3cb241d..a2765fd 100755
--- a/gen/prop/hst
+++ b/gen/prop/hst
@@ -1,55 +1,101 @@
-#!/bin/sh
-
-set -e
-cd "${0%/*}/../.."
-exec >lib/unicode/prop/uprop_get_hst.c
-
-gawk '
-BEGIN {
-	FS = "( *#.*| +; +)"
-
-	print "/* This file is autogenerated by gen/prop/hst; DO NOT EDIT. */"
-	print ""
-	print "#include \"_bsearch.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] = "HST_" toupper($2)
-	}
-}
-
-END {
-	print "static const struct {"
-	print "\trune lo, hi;"
-	print "\tenum uprop_hst val;"
-	print "} lookup[] = {"
-
-	for (i = 0x1100; i <= 0x10FFFF; i++) {
-		if (!props[i])
-			continue
-		lo = i
-		while (props[lo] == props[i + 1])
-			i++
-		printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i]
-	}
-
-	print "};"
-	print ""
-	print "_MLIB_DEFINE_BSEARCH(enum uprop_hst, lookup, HST_NA)"
-	print ""
-	print "enum uprop_hst"
-	print "uprop_get_hst(rune ch)"
-	print "{"
-	print "\treturn ch < lookup[0].lo ? HST_NA : mlib_lookup(ch);"
-	print "}"
-}
-' data/HangulSyllableType | sed 's/\s*$//'
+#!/usr/bin/python3
+
+import math
+
+from lib import *
+
+
+longest = 0
+
+def parse(file: str) -> list[bool]:
+	global longest
+
+	xs = ['HST_NA'] * 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 = 'HST_' + 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/hst; 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_hst 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_hst
+uprop_get_hst(rune ch)
+{{
+	return stage2[stage1[ch / {blksize}]][ch % {blksize}];
+}}''')
+
+def main() -> None:
+	cwd_init()
+	xs = parse('data/HangulSyllableType')
+
+	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 * 2
+		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_hst.c', 'w') as f:
+		sys.stdout = f
+		genfile(Cs, blksize)
+
+	report_size(len(xs), smallest)
+
+if __name__ == '__main__':
+	main()
diff --git a/gen/prop/inpc b/gen/prop/inpc
index b4bd85c..6a8561f 100755
--- a/gen/prop/inpc
+++ b/gen/prop/inpc
@@ -1,55 +1,101 @@
-#!/bin/sh
-
-set -e
-cd "${0%/*}/../.."
-exec >lib/unicode/prop/uprop_get_inpc.c
-
-gawk '
-BEGIN {
-	FS = "( *#.*| +; +)"
-
-	print "/* This file is autogenerated by gen/prop/inpc; DO NOT EDIT. */"
-	print ""
-	print "#include \"_bsearch.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] = "INPC_" toupper($2)
-	}
-}
-
-END {
-	print "static const struct {"
-	print "\trune lo, hi;"
-	print "\tenum uprop_inpc val;"
-	print "} lookup[] = {"
-
-	for (i = 0x900; i <= 0x10FFFF; i++) {
-		if (!props[i])
-			continue
-		lo = i
-		while (props[lo] == props[i + 1])
-			i++
-		printf "\t{RUNE_C(0x%06X), RUNE_C(0x%06X), %s},\n", lo, i, props[i]
-	}
-
-	print "};"
-	print ""
-	print "_MLIB_DEFINE_BSEARCH(enum uprop_inpc, lookup, INPC_NA)"
-	print ""
-	print "enum uprop_inpc"
-	print "uprop_get_inpc(rune ch)"
-	print "{"
-	print "\treturn ch < lookup[0].lo ? INPC_NA : mlib_lookup(ch);"
-	print "}"
-}
-' data/IndicPositionalCategory | sed 's/\s*$//'
+#!/usr/bin/python3
+
+import math
+
+from lib import *
+
+
+longest = 0
+
+def parse(file: str) -> list[bool]:
+	global longest
+
+	xs = ['INPC_NA'] * 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 = 'INPC_' + 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/inpc; 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_inpc 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_inpc
+uprop_get_inpc(rune ch)
+{{
+	return stage2[stage1[ch / {blksize}]][ch % {blksize}];
+}}''')
+
+def main() -> None:
+	cwd_init()
+	xs = parse('data/IndicPositionalCategory')
+
+	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 * 2
+		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_inpc.c', 'w') as f:
+		sys.stdout = f
+		genfile(Cs, blksize)
+
+	report_size(len(xs), smallest)
+
+if __name__ == '__main__':
+	main()