blob: e8b76dfb5fbdd26585ccf1414c7145f18c2dab6b [file] [log] [blame]
#!/usr/bin/env python3
# flake8: noqa: F821
import logging
logging.basicConfig(format='%(levelname)s: %(message)s', level=logging.INFO)
"""usage: ./gen-use-table.py IndicSyllabicCategory.txt IndicPositionalCategory.txt ArabicShaping.txt DerivedCoreProperties.txt UnicodeData.txt Blocks.txt Scripts.txt IndicSyllabicCategory-Additional.txt IndicPositionalCategory-Additional.txt
Input files:
* https://unicode.org/Public/UCD/latest/ucd/IndicSyllabicCategory.txt
* https://unicode.org/Public/UCD/latest/ucd/IndicPositionalCategory.txt
* https://unicode.org/Public/UCD/latest/ucd/ArabicShaping.txt
* https://unicode.org/Public/UCD/latest/ucd/DerivedCoreProperties.txt
* https://unicode.org/Public/UCD/latest/ucd/UnicodeData.txt
* https://unicode.org/Public/UCD/latest/ucd/Blocks.txt
* https://unicode.org/Public/UCD/latest/ucd/Scripts.txt
* ms-use/IndicSyllabicCategory-Additional.txt
* ms-use/IndicPositionalCategory-Additional.txt
"""
import sys
if len (sys.argv) != 10:
sys.exit (__doc__)
DISABLED_SCRIPTS = {
'Arabic',
'Lao',
'Samaritan',
'Syriac',
'Thai',
}
files = [open (x, encoding='utf-8') for x in sys.argv[1:]]
headers = [[f.readline () for i in range (2)] for j,f in enumerate(files) if j != 4]
for j in range(7, 9):
for line in files[j]:
line = line.rstrip()
if not line:
break
headers[j - 1].append(line)
headers.append (["UnicodeData.txt does not have a header."])
unicode_data = [{} for _ in files]
values = [{} for _ in files]
for i, f in enumerate (files):
for line in f:
j = line.find ('#')
if j >= 0:
line = line[:j]
fields = [x.strip () for x in line.split (';')]
if len (fields) == 1:
continue
uu = fields[0].split ('..')
start = int (uu[0], 16)
if len (uu) == 1:
end = start
else:
end = int (uu[1], 16)
t = fields[1 if i not in [2, 4] else 2]
if i == 2:
t = 'jt_' + t
elif i == 3 and t != 'Default_Ignorable_Code_Point':
continue
elif i == 7 and t == 'Consonant_Final_Modifier':
# TODO: https://github.com/MicrosoftDocs/typography-issues/issues/336
t = 'Syllable_Modifier'
elif i == 8 and t == 'NA':
t = 'Not_Applicable'
i0 = i if i < 7 else i - 7
for u in range (start, end + 1):
unicode_data[i0][u] = t
values[i0][t] = values[i0].get (t, 0) + end - start + 1
defaults = ('Other', 'Not_Applicable', 'jt_X', '', 'Cn', 'No_Block', 'Unknown')
# Merge data into one dict:
for i,v in enumerate (defaults):
values[i][v] = values[i].get (v, 0) + 1
combined = {}
for i,d in enumerate (unicode_data):
for u,v in d.items ():
if not u in combined:
if i >= 4:
continue
combined[u] = list (defaults)
combined[u][i] = v
combined = {k: v for k, v in combined.items() if v[6] not in DISABLED_SCRIPTS}
property_names = [
# General_Category
'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',
# Indic_Syllabic_Category
'Other',
'Bindu',
'Visarga',
'Avagraha',
'Nukta',
'Virama',
'Pure_Killer',
'Invisible_Stacker',
'Vowel_Independent',
'Vowel_Dependent',
'Vowel',
'Consonant_Placeholder',
'Consonant',
'Consonant_Dead',
'Consonant_With_Stacker',
'Consonant_Prefixed',
'Consonant_Preceding_Repha',
'Consonant_Succeeding_Repha',
'Consonant_Subjoined',
'Consonant_Medial',
'Consonant_Final',
'Consonant_Head_Letter',
'Consonant_Initial_Postfixed',
'Modifying_Letter',
'Tone_Letter',
'Tone_Mark',
'Gemination_Mark',
'Cantillation_Mark',
'Register_Shifter',
'Syllable_Modifier',
'Consonant_Killer',
'Non_Joiner',
'Joiner',
'Number_Joiner',
'Number',
'Brahmi_Joining_Number',
'Symbol_Modifier',
'Hieroglyph',
'Hieroglyph_Joiner',
'Hieroglyph_Segment_Begin',
'Hieroglyph_Segment_End',
# Indic_Positional_Category
'Not_Applicable',
'Right',
'Left',
'Visual_Order_Left',
'Left_And_Right',
'Top',
'Bottom',
'Top_And_Bottom',
'Top_And_Bottom_And_Left',
'Top_And_Right',
'Top_And_Left',
'Top_And_Left_And_Right',
'Bottom_And_Left',
'Bottom_And_Right',
'Top_And_Bottom_And_Right',
'Overstruck',
# Joining_Type
'jt_C',
'jt_D',
'jt_L',
'jt_R',
'jt_T',
'jt_U',
'jt_X',
]
class PropertyValue(object):
def __init__(self, name_):
self.name = name_
def __str__(self):
return self.name
def __eq__(self, other):
return self.name == (other if isinstance(other, str) else other.name)
def __ne__(self, other):
return not (self == other)
def __hash__(self):
return hash(str(self))
property_values = {}
for name in property_names:
value = PropertyValue(name)
assert value not in property_values
assert value not in globals()
property_values[name] = value
globals().update(property_values)
def is_BASE(U, UISC, UDI, UGC, AJT):
return (UISC in [Number, Consonant, Consonant_Head_Letter,
Tone_Letter,
Vowel_Independent,
] or
# TODO: https://github.com/MicrosoftDocs/typography-issues/issues/484
AJT in [jt_C, jt_D, jt_L, jt_R] and UISC != Joiner or
(UGC == Lo and UISC in [Avagraha, Bindu, Consonant_Final, Consonant_Medial,
Consonant_Subjoined, Vowel, Vowel_Dependent]))
def is_BASE_NUM(U, UISC, UDI, UGC, AJT):
return UISC == Brahmi_Joining_Number
def is_BASE_OTHER(U, UISC, UDI, UGC, AJT):
if UISC == Consonant_Placeholder: return True
return U in [0x2015, 0x2022, 0x25FB, 0x25FC, 0x25FD, 0x25FE]
def is_CGJ(U, UISC, UDI, UGC, AJT):
# Also includes VARIATION_SELECTOR and ZWJ
return UISC == Joiner or UDI and UGC in [Mc, Me, Mn]
def is_CONS_FINAL(U, UISC, UDI, UGC, AJT):
return ((UISC == Consonant_Final and UGC != Lo) or
UISC == Consonant_Succeeding_Repha)
def is_CONS_FINAL_MOD(U, UISC, UDI, UGC, AJT):
return UISC == Syllable_Modifier
def is_CONS_MED(U, UISC, UDI, UGC, AJT):
# Consonant_Initial_Postfixed is new in Unicode 11; not in the spec.
return (UISC == Consonant_Medial and UGC != Lo or
UISC == Consonant_Initial_Postfixed)
def is_CONS_MOD(U, UISC, UDI, UGC, AJT):
return UISC in [Nukta, Gemination_Mark, Consonant_Killer]
def is_CONS_SUB(U, UISC, UDI, UGC, AJT):
return UISC == Consonant_Subjoined and UGC != Lo
def is_CONS_WITH_STACKER(U, UISC, UDI, UGC, AJT):
return UISC == Consonant_With_Stacker
def is_HALANT(U, UISC, UDI, UGC, AJT):
return UISC == Virama and not is_HALANT_OR_VOWEL_MODIFIER(U, UISC, UDI, UGC, AJT)
def is_HALANT_OR_VOWEL_MODIFIER(U, UISC, UDI, UGC, AJT):
# Split off of HALANT
return U == 0x0DCA
def is_HALANT_NUM(U, UISC, UDI, UGC, AJT):
return UISC == Number_Joiner
def is_HIEROGLYPH(U, UISC, UDI, UGC, AJT):
return UISC == Hieroglyph
def is_HIEROGLYPH_JOINER(U, UISC, UDI, UGC, AJT):
return UISC == Hieroglyph_Joiner
def is_HIEROGLYPH_SEGMENT_BEGIN(U, UISC, UDI, UGC, AJT):
return UISC == Hieroglyph_Segment_Begin
def is_HIEROGLYPH_SEGMENT_END(U, UISC, UDI, UGC, AJT):
return UISC == Hieroglyph_Segment_End
def is_INVISIBLE_STACKER(U, UISC, UDI, UGC, AJT):
# Split off of HALANT
return (UISC == Invisible_Stacker
and not is_SAKOT(U, UISC, UDI, UGC, AJT)
)
def is_ZWNJ(U, UISC, UDI, UGC, AJT):
return UISC == Non_Joiner
def is_OTHER(U, UISC, UDI, UGC, AJT):
# Also includes BASE_IND and SYM
return ((UGC == Po or UISC in [Consonant_Dead, Joiner, Modifying_Letter, Other])
and not is_BASE(U, UISC, UDI, UGC, AJT)
and not is_BASE_OTHER(U, UISC, UDI, UGC, AJT)
and not is_CGJ(U, UISC, UDI, UGC, AJT)
and not is_SYM_MOD(U, UISC, UDI, UGC, AJT)
and not is_Word_Joiner(U, UISC, UDI, UGC, AJT)
)
def is_REPHA(U, UISC, UDI, UGC, AJT):
return UISC in [Consonant_Preceding_Repha, Consonant_Prefixed]
def is_SAKOT(U, UISC, UDI, UGC, AJT):
# Split off of HALANT
return U == 0x1A60
def is_SYM_MOD(U, UISC, UDI, UGC, AJT):
return UISC == Symbol_Modifier
def is_VOWEL(U, UISC, UDI, UGC, AJT):
return (UISC == Pure_Killer or
UGC != Lo and UISC in [Vowel, Vowel_Dependent])
def is_VOWEL_MOD(U, UISC, UDI, UGC, AJT):
return (UISC in [Tone_Mark, Cantillation_Mark, Register_Shifter, Visarga] or
UGC != Lo and UISC == Bindu)
def is_Word_Joiner(U, UISC, UDI, UGC, AJT):
# Also includes Rsv
return (UDI and U not in [0x115F, 0x1160, 0x3164, 0xFFA0, 0x1BCA0, 0x1BCA1, 0x1BCA2, 0x1BCA3]
and UISC == Other
and not is_CGJ(U, UISC, UDI, UGC, AJT)
) or UGC == Cn
use_mapping = {
'B': is_BASE,
'N': is_BASE_NUM,
'GB': is_BASE_OTHER,
'CGJ': is_CGJ,
'F': is_CONS_FINAL,
'FM': is_CONS_FINAL_MOD,
'M': is_CONS_MED,
'CM': is_CONS_MOD,
'SUB': is_CONS_SUB,
'CS': is_CONS_WITH_STACKER,
'H': is_HALANT,
'HVM': is_HALANT_OR_VOWEL_MODIFIER,
'HN': is_HALANT_NUM,
'IS': is_INVISIBLE_STACKER,
'G': is_HIEROGLYPH,
'J': is_HIEROGLYPH_JOINER,
'SB': is_HIEROGLYPH_SEGMENT_BEGIN,
'SE': is_HIEROGLYPH_SEGMENT_END,
'ZWNJ': is_ZWNJ,
'O': is_OTHER,
'R': is_REPHA,
'Sk': is_SAKOT,
'SM': is_SYM_MOD,
'V': is_VOWEL,
'VM': is_VOWEL_MOD,
'WJ': is_Word_Joiner,
}
use_positions = {
'F': {
'Abv': [Top],
'Blw': [Bottom],
'Pst': [Right],
},
'M': {
'Abv': [Top],
'Blw': [Bottom, Bottom_And_Left, Bottom_And_Right],
'Pst': [Right],
'Pre': [Left, Top_And_Bottom_And_Left],
},
'CM': {
'Abv': [Top],
'Blw': [Bottom, Overstruck],
},
'V': {
'Abv': [Top, Top_And_Bottom, Top_And_Bottom_And_Right, Top_And_Right],
'Blw': [Bottom, Overstruck, Bottom_And_Right],
'Pst': [Right],
'Pre': [Left, Top_And_Left, Top_And_Left_And_Right, Left_And_Right],
},
'VM': {
'Abv': [Top],
'Blw': [Bottom, Overstruck],
'Pst': [Right],
'Pre': [Left],
},
'SM': {
'Abv': [Top],
'Blw': [Bottom],
},
'H': None,
'HVM': None,
'IS': None,
'B': None,
'FM': {
'Abv': [Top],
'Blw': [Bottom],
'Pst': [Not_Applicable],
},
'R': None,
'SUB': None,
}
def map_to_use(data):
out = {}
items = use_mapping.items()
for U, (UISC, UIPC, AJT, UDI, UGC, UBlock, _) in data.items():
# Resolve Indic_Syllabic_Category
# TODO: These don't have UISC assigned in Unicode 13.0.0, but have UIPC
if 0x1CE2 <= U <= 0x1CE8: UISC = Cantillation_Mark
# Tibetan:
# TODO: These don't have UISC assigned in Unicode 13.0.0, but have UIPC
if 0x0F18 <= U <= 0x0F19 or 0x0F3E <= U <= 0x0F3F: UISC = Vowel_Dependent
# TODO: U+1CED should only be allowed after some of
# the nasalization marks, maybe only for U+1CE9..U+1CF1.
if U == 0x1CED: UISC = Tone_Mark
values = [k for k,v in items if v(U, UISC, UDI, UGC, AJT)]
assert len(values) == 1, "%s %s %s %s %s %s" % (hex(U), UISC, UDI, UGC, AJT, values)
USE = values[0]
# Resolve Indic_Positional_Category
# TODO: https://github.com/harfbuzz/harfbuzz/pull/1037
# and https://github.com/harfbuzz/harfbuzz/issues/1631
if U in [0x11302, 0x11303, 0x114C1]: UIPC = Top
assert (UIPC in [Not_Applicable, Visual_Order_Left] or U == 0x0F7F or
USE in use_positions), "%s %s %s %s %s %s %s" % (hex(U), UIPC, USE, UISC, UDI, UGC, AJT)
pos_mapping = use_positions.get(USE, None)
if pos_mapping:
values = [k for k,v in pos_mapping.items() if v and UIPC in v]
assert len(values) == 1, "%s %s %s %s %s %s %s %s" % (hex(U), UIPC, USE, UISC, UDI, UGC, AJT, values)
USE = USE + values[0]
out[U] = (USE, UBlock)
return out
use_data = map_to_use(combined)
print ("/* == Start of generated table == */")
print ("/*")
print (" * The following table is generated by running:")
print (" *")
print (" * {} IndicSyllabicCategory.txt IndicPositionalCategory.txt ArabicShaping.txt DerivedCoreProperties.txt UnicodeData.txt Blocks.txt Scripts.txt IndicSyllabicCategory-Additional.txt IndicPositionalCategory-Additional.txt".format (sys.argv[0]))
print (" *")
print (" * on files with these headers:")
print (" *")
for h in headers:
for l in h:
print (" * %s" % (l.strip()))
print (" */")
print ()
print ("#ifndef HB_OT_SHAPER_USE_TABLE_HH")
print ("#define HB_OT_SHAPER_USE_TABLE_HH")
print ()
print ('#include "hb.hh"')
print ()
print ('#include "hb-ot-shaper-use-machine.hh"')
print ()
total = 0
used = 0
last_block = None
def print_block (block, start, end, use_data):
global total, used, last_block
if block and block != last_block:
print ()
print ()
print (" /* %s */" % block)
if start % 16:
print (' ' * (20 + (start % 16 * 6)), end='')
num = 0
assert start % 8 == 0
assert (end+1) % 8 == 0
for u in range (start, end+1):
if u % 16 == 0:
print ()
print (" /* %04X */" % u, end='')
if u in use_data:
num += 1
d = use_data.get (u)
if d is not None:
d = d[0]
elif u in unicode_data[4]:
d = 'O'
else:
d = 'WJ'
print ("%6s," % d, end='')
total += end - start + 1
used += num
if block:
last_block = block
uu = sorted (use_data.keys ())
last = -100000
num = 0
offset = 0
starts = []
ends = []
print ('#pragma GCC diagnostic push')
print ('#pragma GCC diagnostic ignored "-Wunused-macros"')
for k,v in sorted(use_mapping.items()):
if k in use_positions and use_positions[k]: continue
print ("#define %s USE(%s) /* %s */" % (k, k, v.__name__[3:]))
for k,v in sorted(use_positions.items()):
if not v: continue
for suf in v.keys():
tag = k + suf
print ("#define %s USE(%s)" % (tag, tag))
print ('#pragma GCC diagnostic pop')
print ("")
import packTab
data = {u:v[0] for u,v in use_data.items()}
DEFAULT = 5
COMPACT = 9
for compression in (DEFAULT, COMPACT):
logging.info(' Compression=%d:' % compression)
print()
if compression == DEFAULT:
print('#ifndef HB_OPTIMIZE_SIZE')
elif compression == COMPACT:
print('#else')
else:
assert False
print()
code = packTab.Code('hb_use')
sol = packTab.pack_table(data, compression=compression, default='O')
logging.info(' FullCost=%d' % (sol.fullCost))
sol.genCode(code, f'get_category')
code.print_c(linkage='static inline')
print ()
print('#endif')
print ()
for k in sorted(use_mapping.keys()):
if k in use_positions and use_positions[k]: continue
print ("#undef %s" % k)
for k,v in sorted(use_positions.items()):
if not v: continue
for suf in v.keys():
tag = k + suf
print ("#undef %s" % tag)
print ()
print ()
print ("#endif /* HB_OT_SHAPER_USE_TABLE_HH */")
print ("/* == End of generated table == */")