trunk/maint/MultiStage2.py

#! /usr/bin/python

# Multistage table builder
# (c) Peter Kankowski, 2008

import re
import string
import sys

MAX_UNICODE = 0x110000
NOTACHAR = 0xffffffff

# Parse a line of CaseFolding.txt, Scripts.txt, and DerivedGeneralCategory.txt file
def make_get_names(enum):
        return lambda chardata: enum.index(chardata[1])

def get_case_folding_value(chardata):
        if chardata[1] != 'C' and chardata[1] != 'S':
                return 0
        return int(chardata[2], 16) - int(chardata[0], 16)
        
def get_other_case(chardata):
        if chardata[12] != '':
                return int(chardata[12], 16) - int(chardata[0], 16)
        if chardata[13] != '':
                return int(chardata[13], 16) - int(chardata[0], 16)
        return 0

# Read the whole table in memory
def read_table(file_name, get_value, default_value):
        file = open(file_name, 'r')
        table = [default_value] * MAX_UNICODE
        for line in file:
                line = re.sub(r'#.*', '', line)
                chardata = map(string.strip, line.split(';'))
                if len(chardata) <= 1:
                        continue
                value = get_value(chardata)
                
                m = re.match(r'([0-9a-fA-F]+)(\.\.([0-9a-fA-F]+))?$', chardata[0])
                char = int(m.group(1), 16)
#PH             last = char if m.group(3) is None else int(m.group(3), 16)
                if m.group(3) is None:
                        last = char
                else:
                        last = int(m.group(3), 16)            
                for i in range(char, last + 1):
                        table[i] = value
        file.close()
        return table

# Get the smallest possible C language type for the values
def get_type_size(table):
        type_size = [("uschar", 1), ("pcre_uint16", 2), ("pcre_uint32", 4),
                                 ("signed char", 1), ("pcre_int16", 2), ("pcre_int32", 4)]
        limits = [(0, 255), (0, 65535), (0, 4294967295),
                          (-128, 127), (-32768, 32767), (-2147483648, 2147483647)]
        minval = min(table)
        maxval = max(table)
        for num, (minlimit, maxlimit) in enumerate(limits):
                if minlimit <= minval and maxval <= maxlimit:
                        return type_size[num]
        else:
                raise OverflowError, "Too large to fit into C types"

def get_tables_size(*tables):
        total_size = 0
        for table in tables:
                type, size = get_type_size(table)
                total_size += size * len(table)
        return total_size

# Compress the table into the two stages
def compress_table(table, block_size):
        blocks = {} # Dictionary for finding identical blocks
        stage1 = [] # Stage 1 table contains block numbers (indices into stage 2 table)
        stage2 = [] # Stage 2 table contains the blocks with property values
        table = tuple(table)
        for i in range(0, len(table), block_size):
                block = table[i:i+block_size]
                start = blocks.get(block)
                if start is None:
                        # Allocate a new block
                        start = len(stage2) / block_size
                        stage2 += block
                        blocks[block] = start
                stage1.append(start)
        
        return stage1, stage2

# Print a table
def print_table(table, table_name, block_size = None):
        type, size = get_type_size(table)
        ELEMS_PER_LINE = 16
        
        s = "const %s %s[] = { /* %d bytes" % (type, table_name, size * len(table))
        if block_size:
                s += ", block = %d" % block_size
        print s + " */"
        table = tuple(table)
        if block_size is None:
                fmt = "%3d," * ELEMS_PER_LINE + " /* U+%04X */"
                mult = MAX_UNICODE / len(table)
                for i in range(0, len(table), ELEMS_PER_LINE):
                        print fmt % (table[i:i+ELEMS_PER_LINE] + (i * mult,))
        else:
#PH             fmt = "%3d," * (ELEMS_PER_LINE if block_size > ELEMS_PER_LINE else block_size) + "\n"
                if block_size > ELEMS_PER_LINE:
                        fmt = "%3d," * ELEMS_PER_LINE + "\n"
                else:
                        fmt = "%3d," * block_size + "\n"          
                if block_size > ELEMS_PER_LINE:
                        fmt = fmt * (block_size / ELEMS_PER_LINE)
                for i in range(0, len(table), block_size):
                        print ("/* block %d */\n" + fmt) % ((i / block_size,) + table[i:i+block_size])
        print "};\n"

# Extract the unique combinations of properties into records
def combine_tables(*tables):
        records = {}
        index = []
        for t in zip(*tables):
                i = records.get(t)
                if i is None:
                        i = records[t] = len(records)
                index.append(i)
        return index, records

def print_records(records):
        print 'const ucd_record ucd_records[] = { /* %d bytes */' % (len(records) * 4)
        records = zip(records.keys(), records.values())
        records.sort(None, lambda x: x[1])
        for i, record in enumerate(records):
                print ('  {' + '%6d, ' * len(record[0]) + '}, /* %3d */') % (record[0] + (i,))
        print '};\n'

script_names = ['Arabic', 'Armenian', 'Bengali', 'Bopomofo', 'Braille', 'Buginese', 'Buhid', 'Canadian_Aboriginal', \
 'Cherokee', 'Common', 'Coptic', 'Cypriot', 'Cyrillic', 'Deseret', 'Devanagari', 'Ethiopic', 'Georgian', \
 'Glagolitic', 'Gothic', 'Greek', 'Gujarati', 'Gurmukhi', 'Han', 'Hangul', 'Hanunoo', 'Hebrew', 'Hiragana', \
 'Inherited', 'Kannada', 'Katakana', 'Kharoshthi', 'Khmer', 'Lao', 'Latin', 'Limbu', 'Linear_B', 'Malayalam', \
 'Mongolian', 'Myanmar', 'New_Tai_Lue', 'Ogham', 'Old_Italic', 'Old_Persian', 'Oriya', 'Osmanya', 'Runic', \
 'Shavian', 'Sinhala', 'Syloti_Nagri', 'Syriac', 'Tagalog', 'Tagbanwa', 'Tai_Le', 'Tamil', 'Telugu', 'Thaana', \
 'Thai', 'Tibetan', 'Tifinagh', 'Ugaritic', 'Yi', \
 'Balinese', 'Cuneiform', 'Nko', 'Phags_Pa', 'Phoenician']
 
category_names = ['Cc', 'Cf', 'Cn', 'Co', 'Cs', 'Ll', 'Lm', 'Lo', 'Lt', 'Lu',
  'Mc', 'Me', 'Mn', 'Nd', 'Nl', 'No', 'Pc', 'Pd', 'Pe', 'Pf', 'Pi', 'Po', 'Ps',
  'Sc', 'Sk', 'Sm', 'So', 'Zl', 'Zp', 'Zs' ]


script = read_table('Unicode.tables/Scripts.txt', make_get_names(script_names), script_names.index('Common'))
category = read_table('Unicode.tables/DerivedGeneralCategory.txt', make_get_names(category_names), category_names.index('Cn'))
other_case = read_table('Unicode.tables/UnicodeData.txt', get_other_case, 0)
# case_fold = read_table('CaseFolding.txt', get_case_folding_value, 0)

table, records = combine_tables(script, category, other_case)

# Find the optimum block size for the two-stage table
min_size = sys.maxint
for block_size in [2 ** i for i in range(5,10)]:
        size = len(records) * 4
        stage1, stage2 = compress_table(table, block_size)
        size += get_tables_size(stage1, stage2)
        #print "/* block size %5d  => %5d bytes */" % (block_size, size)
        if size < min_size:
                min_size = size
                min_stage1, min_stage2 = stage1, stage2
                min_block_size = block_size

print "#ifdef HAVE_CONFIG_H"
print "#include \"config.h\""
print "#endif"
print "#include \"pcre_internal.h\""
print
print "/* Unicode character database. */"
print "/* This file was autogenerated by MultiStage2.py script. */"
print "/* Total size: %d bytes, block size: %d. */" % (min_size, min_block_size)
print_records(records)
print_table(min_stage1, 'ucd_stage1')
print_table(min_stage2, 'ucd_stage2', min_block_size)
print "#if UCD_BLOCK_SIZE != %d" % min_block_size
print "#error Please correct UCD_BLOCK_SIZE in pcre_internal.h"
print "#endif"

"""

# Three-stage tables:

# Find the optimum block size for 3-stage table
min_size = sys.maxint
for stage3_block in [2 ** i for i in range(2,6)]:
        stage_i, stage3 = compress_table(table, stage3_block)
        for stage2_block in [2 ** i for i in range(5,10)]:
                size = len(records) * 4
                stage1, stage2 = compress_table(stage_i, stage2_block)
                size += get_tables_size(stage1, stage2, stage3)
                # print "/* %5d / %3d  => %5d bytes */" % (stage2_block, stage3_block, size)
                if size < min_size:
                        min_size = size
                        min_stage1, min_stage2, min_stage3 = stage1, stage2, stage3
                        min_stage2_block, min_stage3_block = stage2_block, stage3_block

print "/* Total size: %d bytes" % min_size */
print_records(records)
print_table(min_stage1, 'ucd_stage1')
print_table(min_stage2, 'ucd_stage2', min_stage2_block)
print_table(min_stage3, 'ucd_stage3', min_stage3_block)

"""
1	#! /usr/bin/python
2
3	# Multistage table builder
4	# (c) Peter Kankowski, 2008
5
6	import re
7	import string
8	import sys
9
10	MAX_UNICODE = 0x110000
11	NOTACHAR = 0xffffffff
12
13	# Parse a line of CaseFolding.txt, Scripts.txt, and DerivedGeneralCategory.txt file
14	def make_get_names(enum):
15	return lambda chardata: enum.index(chardata[1])
16
17	def get_case_folding_value(chardata):
18	if chardata[1] != 'C' and chardata[1] != 'S':
19	return 0
20	return int(chardata[2], 16) - int(chardata[0], 16)
21
22	def get_other_case(chardata):
23	if chardata[12] != '':
24	return int(chardata[12], 16) - int(chardata[0], 16)
25	if chardata[13] != '':
26	return int(chardata[13], 16) - int(chardata[0], 16)
27	return 0
28
29	# Read the whole table in memory
30	def read_table(file_name, get_value, default_value):
31	file = open(file_name, 'r')
32	table = [default_value] * MAX_UNICODE
33	for line in file:
34	line = re.sub(r'#.*', '', line)
35	chardata = map(string.strip, line.split(';'))
36	if len(chardata) <= 1:
37	continue
38	value = get_value(chardata)
39
40	m = re.match(r'([0-9a-fA-F]+)(\.\.([0-9a-fA-F]+))?$', chardata[0])
41	char = int(m.group(1), 16)
42	#PH last = char if m.group(3) is None else int(m.group(3), 16)
43	if m.group(3) is None:
44	last = char
45	else:
46	last = int(m.group(3), 16)
47	for i in range(char, last + 1):
48	table[i] = value
49	file.close()
50	return table
51
52	# Get the smallest possible C language type for the values
53	def get_type_size(table):
54	type_size = [("uschar", 1), ("pcre_uint16", 2), ("pcre_uint32", 4),
55	("signed char", 1), ("pcre_int16", 2), ("pcre_int32", 4)]
56	limits = [(0, 255), (0, 65535), (0, 4294967295),
57	(-128, 127), (-32768, 32767), (-2147483648, 2147483647)]
58	minval = min(table)
59	maxval = max(table)
60	for num, (minlimit, maxlimit) in enumerate(limits):
61	if minlimit <= minval and maxval <= maxlimit:
62	return type_size[num]
63	else:
64	raise OverflowError, "Too large to fit into C types"
65
66	def get_tables_size(*tables):
67	total_size = 0
68	for table in tables:
69	type, size = get_type_size(table)
70	total_size += size * len(table)
71	return total_size
72
73	# Compress the table into the two stages
74	def compress_table(table, block_size):
75	blocks = {} # Dictionary for finding identical blocks
76	stage1 = [] # Stage 1 table contains block numbers (indices into stage 2 table)
77	stage2 = [] # Stage 2 table contains the blocks with property values
78	table = tuple(table)
79	for i in range(0, len(table), block_size):
80	block = table[i:i+block_size]
81	start = blocks.get(block)
82	if start is None:
83	# Allocate a new block
84	start = len(stage2) / block_size
85	stage2 += block
86	blocks[block] = start
87	stage1.append(start)
88
89	return stage1, stage2
90
91	# Print a table
92	def print_table(table, table_name, block_size = None):
93	type, size = get_type_size(table)
94	ELEMS_PER_LINE = 16
95
96	s = "const %s %s[] = { /* %d bytes" % (type, table_name, size * len(table))
97	if block_size:
98	s += ", block = %d" % block_size
99	print s + " */"
100	table = tuple(table)
101	if block_size is None:
102	fmt = "%3d," * ELEMS_PER_LINE + " /* U+%04X */"
103	mult = MAX_UNICODE / len(table)
104	for i in range(0, len(table), ELEMS_PER_LINE):
105	print fmt % (table[i:i+ELEMS_PER_LINE] + (i * mult,))
106	else:
107	#PH fmt = "%3d," * (ELEMS_PER_LINE if block_size > ELEMS_PER_LINE else block_size) + "\n"
108	if block_size > ELEMS_PER_LINE:
109	fmt = "%3d," * ELEMS_PER_LINE + "\n"
110	else:
111	fmt = "%3d," * block_size + "\n"
112	if block_size > ELEMS_PER_LINE:
113	fmt = fmt * (block_size / ELEMS_PER_LINE)
114	for i in range(0, len(table), block_size):
115	print ("/* block %d */\n" + fmt) % ((i / block_size,) + table[i:i+block_size])
116	print "};\n"
117
118	# Extract the unique combinations of properties into records
119	def combine_tables(*tables):
120	records = {}
121	index = []
122	for t in zip(*tables):
123	i = records.get(t)
124	if i is None:
125	i = records[t] = len(records)
126	index.append(i)
127	return index, records
128
129	def print_records(records):
130	print 'const ucd_record ucd_records[] = { /* %d bytes /' % (len(records) 4)
131	records = zip(records.keys(), records.values())
132	records.sort(None, lambda x: x[1])
133	for i, record in enumerate(records):
134	print (' {' + '%6d, ' * len(record[0]) + '}, /* %3d */') % (record[0] + (i,))
135	print '};\n'
136
137	script_names = ['Arabic', 'Armenian', 'Bengali', 'Bopomofo', 'Braille', 'Buginese', 'Buhid', 'Canadian_Aboriginal', \
138	'Cherokee', 'Common', 'Coptic', 'Cypriot', 'Cyrillic', 'Deseret', 'Devanagari', 'Ethiopic', 'Georgian', \
139	'Glagolitic', 'Gothic', 'Greek', 'Gujarati', 'Gurmukhi', 'Han', 'Hangul', 'Hanunoo', 'Hebrew', 'Hiragana', \
140	'Inherited', 'Kannada', 'Katakana', 'Kharoshthi', 'Khmer', 'Lao', 'Latin', 'Limbu', 'Linear_B', 'Malayalam', \
141	'Mongolian', 'Myanmar', 'New_Tai_Lue', 'Ogham', 'Old_Italic', 'Old_Persian', 'Oriya', 'Osmanya', 'Runic', \
142	'Shavian', 'Sinhala', 'Syloti_Nagri', 'Syriac', 'Tagalog', 'Tagbanwa', 'Tai_Le', 'Tamil', 'Telugu', 'Thaana', \
143	'Thai', 'Tibetan', 'Tifinagh', 'Ugaritic', 'Yi', \
144	'Balinese', 'Cuneiform', 'Nko', 'Phags_Pa', 'Phoenician']
145
146	category_names = ['Cc', 'Cf', 'Cn', 'Co', 'Cs', 'Ll', 'Lm', 'Lo', 'Lt', 'Lu',
147	'Mc', 'Me', 'Mn', 'Nd', 'Nl', 'No', 'Pc', 'Pd', 'Pe', 'Pf', 'Pi', 'Po', 'Ps',
148	'Sc', 'Sk', 'Sm', 'So', 'Zl', 'Zp', 'Zs' ]
149
150
151	script = read_table('Unicode.tables/Scripts.txt', make_get_names(script_names), script_names.index('Common'))
152	category = read_table('Unicode.tables/DerivedGeneralCategory.txt', make_get_names(category_names), category_names.index('Cn'))
153	other_case = read_table('Unicode.tables/UnicodeData.txt', get_other_case, 0)
154	# case_fold = read_table('CaseFolding.txt', get_case_folding_value, 0)
155
156	table, records = combine_tables(script, category, other_case)
157
158	# Find the optimum block size for the two-stage table
159	min_size = sys.maxint
160	for block_size in [2 ** i for i in range(5,10)]:
161	size = len(records) * 4
162	stage1, stage2 = compress_table(table, block_size)
163	size += get_tables_size(stage1, stage2)
164	#print "/* block size %5d => %5d bytes */" % (block_size, size)
165	if size < min_size:
166	min_size = size
167	min_stage1, min_stage2 = stage1, stage2
168	min_block_size = block_size
169
170	print "#ifdef HAVE_CONFIG_H"
171	print "#include \"config.h\""
172	print "#endif"
173	print "#include \"pcre_internal.h\""
174	print
175	print "/* Unicode character database. */"
176	print "/* This file was autogenerated by MultiStage2.py script. */"
177	print "/* Total size: %d bytes, block size: %d. */" % (min_size, min_block_size)
178	print_records(records)
179	print_table(min_stage1, 'ucd_stage1')
180	print_table(min_stage2, 'ucd_stage2', min_block_size)
181	print "#if UCD_BLOCK_SIZE != %d" % min_block_size
182	print "#error Please correct UCD_BLOCK_SIZE in pcre_internal.h"
183	print "#endif"
184
185	"""
186
187	# Three-stage tables:
188
189	# Find the optimum block size for 3-stage table
190	min_size = sys.maxint
191	for stage3_block in [2 ** i for i in range(2,6)]:
192	stage_i, stage3 = compress_table(table, stage3_block)
193	for stage2_block in [2 ** i for i in range(5,10)]:
194	size = len(records) * 4
195	stage1, stage2 = compress_table(stage_i, stage2_block)
196	size += get_tables_size(stage1, stage2, stage3)
197	# print "/* %5d / %3d => %5d bytes */" % (stage2_block, stage3_block, size)
198	if size < min_size:
199	min_size = size
200	min_stage1, min_stage2, min_stage3 = stage1, stage2, stage3
201	min_stage2_block, min_stage3_block = stage2_block, stage3_block
202
203	print "/* Total size: %d bytes" % min_size */
204	print_records(records)
205	print_table(min_stage1, 'ucd_stage1')
206	print_table(min_stage2, 'ucd_stage2', min_stage2_block)
207	print_table(min_stage3, 'ucd_stage3', min_stage3_block)
208
209	"""