aboutsummaryrefslogtreecommitdiff
path: root/src/gen-indic-table.py
diff options
context:
space:
mode:
Diffstat (limited to 'src/gen-indic-table.py')
-rwxr-xr-xsrc/gen-indic-table.py505
1 files changed, 465 insertions, 40 deletions
diff --git a/src/gen-indic-table.py b/src/gen-indic-table.py
index 367e55e27..4ef970265 100755
--- a/src/gen-indic-table.py
+++ b/src/gen-indic-table.py
@@ -26,7 +26,6 @@ ALLOWED_BLOCKS = [
'Telugu',
'Kannada',
'Malayalam',
- 'Sinhala',
'Myanmar',
'Khmer',
'Vedic Extensions',
@@ -41,8 +40,7 @@ files = [open (x, encoding='utf-8') for x in sys.argv[1:]]
headers = [[f.readline () for i in range (2)] for f in files]
-data = [{} for _ in files]
-values = [{} for _ in files]
+unicode_data = [{} for _ in files]
for i, f in enumerate (files):
for line in f:
@@ -64,15 +62,12 @@ for i, f in enumerate (files):
t = fields[1]
for u in range (start, end + 1):
- data[i][u] = t
- values[i][t] = values[i].get (t, 0) + end - start + 1
+ unicode_data[i][u] = t
# Merge data into one dict:
defaults = ('Other', 'Not_Applicable', 'No_Block')
-for i,v in enumerate (defaults):
- values[i][v] = values[i].get (v, 0) + 1
combined = {}
-for i,d in enumerate (data):
+for i,d in enumerate (unicode_data):
for u,v in d.items ():
if i == 2 and not u in combined:
continue
@@ -80,14 +75,415 @@ for i,d in enumerate (data):
combined[u] = list (defaults)
combined[u][i] = v
combined = {k:v for k,v in combined.items() if k in ALLOWED_SINGLES or v[2] in ALLOWED_BLOCKS}
-data = combined
-del combined
+
+
+# Convert categories & positions types
+
+categories = {
+ 'indic' : [
+ 'X',
+ 'C',
+ 'V',
+ 'N',
+ 'H',
+ 'ZWNJ',
+ 'ZWJ',
+ 'M',
+ 'SM',
+ 'A',
+ 'VD',
+ 'PLACEHOLDER',
+ 'DOTTEDCIRCLE',
+ 'RS',
+ 'MPst',
+ 'Repha',
+ 'Ra',
+ 'CM',
+ 'Symbol',
+ 'CS',
+ ],
+ 'khmer' : [
+ 'VAbv',
+ 'VBlw',
+ 'VPre',
+ 'VPst',
+
+ 'Robatic',
+ 'Xgroup',
+ 'Ygroup',
+ ],
+ 'myanmar' : [
+ 'VAbv',
+ 'VBlw',
+ 'VPre',
+ 'VPst',
+
+ 'IV',
+ 'As',
+ 'DB',
+ 'GB',
+ 'MH',
+ 'MR',
+ 'MW',
+ 'MY',
+ 'PT',
+ 'VS',
+ 'ML',
+ ],
+}
+
+category_map = {
+ 'Other' : 'X',
+ 'Avagraha' : 'Symbol',
+ 'Bindu' : 'SM',
+ 'Brahmi_Joining_Number' : 'PLACEHOLDER', # Don't care.
+ 'Cantillation_Mark' : 'A',
+ 'Consonant' : 'C',
+ 'Consonant_Dead' : 'C',
+ 'Consonant_Final' : 'CM',
+ 'Consonant_Head_Letter' : 'C',
+ 'Consonant_Initial_Postfixed' : 'C', # TODO
+ 'Consonant_Killer' : 'M', # U+17CD only.
+ 'Consonant_Medial' : 'CM',
+ 'Consonant_Placeholder' : 'PLACEHOLDER',
+ 'Consonant_Preceding_Repha' : 'Repha',
+ 'Consonant_Prefixed' : 'X', # Don't care.
+ 'Consonant_Subjoined' : 'CM',
+ 'Consonant_Succeeding_Repha' : 'CM',
+ 'Consonant_With_Stacker' : 'CS',
+ 'Gemination_Mark' : 'SM', # https://github.com/harfbuzz/harfbuzz/issues/552
+ 'Invisible_Stacker' : 'H',
+ 'Joiner' : 'ZWJ',
+ 'Modifying_Letter' : 'X',
+ 'Non_Joiner' : 'ZWNJ',
+ 'Nukta' : 'N',
+ 'Number' : 'PLACEHOLDER',
+ 'Number_Joiner' : 'PLACEHOLDER', # Don't care.
+ 'Pure_Killer' : 'M', # Is like a vowel matra.
+ 'Register_Shifter' : 'RS',
+ 'Syllable_Modifier' : 'SM',
+ 'Tone_Letter' : 'X',
+ 'Tone_Mark' : 'N',
+ 'Virama' : 'H',
+ 'Visarga' : 'SM',
+ 'Vowel' : 'V',
+ 'Vowel_Dependent' : 'M',
+ 'Vowel_Independent' : 'V',
+}
+position_map = {
+ 'Not_Applicable' : 'END',
+
+ 'Left' : 'PRE_C',
+ 'Top' : 'ABOVE_C',
+ 'Bottom' : 'BELOW_C',
+ 'Right' : 'POST_C',
+
+ # These should resolve to the position of the last part of the split sequence.
+ 'Bottom_And_Right' : 'POST_C',
+ 'Left_And_Right' : 'POST_C',
+ 'Top_And_Bottom' : 'BELOW_C',
+ 'Top_And_Bottom_And_Left' : 'BELOW_C',
+ 'Top_And_Bottom_And_Right' : 'POST_C',
+ 'Top_And_Left' : 'ABOVE_C',
+ 'Top_And_Left_And_Right' : 'POST_C',
+ 'Top_And_Right' : 'POST_C',
+
+ 'Overstruck' : 'AFTER_MAIN',
+ 'Visual_order_left' : 'PRE_M',
+}
+
+category_overrides = {
+
+ # These are the variation-selectors. They only appear in the Myanmar grammar
+ # but are not Myanmar-specific
+ 0xFE00: 'VS',
+ 0xFE01: 'VS',
+ 0xFE02: 'VS',
+ 0xFE03: 'VS',
+ 0xFE04: 'VS',
+ 0xFE05: 'VS',
+ 0xFE06: 'VS',
+ 0xFE07: 'VS',
+ 0xFE08: 'VS',
+ 0xFE09: 'VS',
+ 0xFE0A: 'VS',
+ 0xFE0B: 'VS',
+ 0xFE0C: 'VS',
+ 0xFE0D: 'VS',
+ 0xFE0E: 'VS',
+ 0xFE0F: 'VS',
+
+ # These appear in the OT Myanmar spec, but are not Myanmar-specific
+ 0x2015: 'PLACEHOLDER',
+ 0x2022: 'PLACEHOLDER',
+ 0x25FB: 'PLACEHOLDER',
+ 0x25FC: 'PLACEHOLDER',
+ 0x25FD: 'PLACEHOLDER',
+ 0x25FE: 'PLACEHOLDER',
+
+
+ # Indic
+
+ 0x0930: 'Ra', # Devanagari
+ 0x09B0: 'Ra', # Bengali
+ 0x09F0: 'Ra', # Bengali
+ 0x0A30: 'Ra', # Gurmukhi No Reph
+ 0x0AB0: 'Ra', # Gujarati
+ 0x0B30: 'Ra', # Oriya
+ 0x0BB0: 'Ra', # Tamil No Reph
+ 0x0C30: 'Ra', # Telugu Reph formed only with ZWJ
+ 0x0CB0: 'Ra', # Kannada
+ 0x0D30: 'Ra', # Malayalam No Reph, Logical Repha
+
+ # The following act more like the Bindus.
+ 0x0953: 'SM',
+ 0x0954: 'SM',
+
+ # U+0A40 GURMUKHI VOWEL SIGN II may be preceded by U+0A02 GURMUKHI SIGN BINDI.
+ 0x0A40: 'MPst',
+
+ # The following act like consonants.
+ 0x0A72: 'C',
+ 0x0A73: 'C',
+ 0x1CF5: 'C',
+ 0x1CF6: 'C',
+
+ # TODO: The following should only be allowed after a Visarga.
+ # For now, just treat them like regular tone marks.
+ 0x1CE2: 'A',
+ 0x1CE3: 'A',
+ 0x1CE4: 'A',
+ 0x1CE5: 'A',
+ 0x1CE6: 'A',
+ 0x1CE7: 'A',
+ 0x1CE8: 'A',
+
+ # TODO: The following should only be allowed after some of
+ # the nasalization marks, maybe only for U+1CE9..U+1CF1.
+ # For now, just treat them like tone marks.
+ 0x1CED: 'A',
+
+ # The following take marks in standalone clusters, similar to Avagraha.
+ 0xA8F2: 'Symbol',
+ 0xA8F3: 'Symbol',
+ 0xA8F4: 'Symbol',
+ 0xA8F5: 'Symbol',
+ 0xA8F6: 'Symbol',
+ 0xA8F7: 'Symbol',
+ 0x1CE9: 'Symbol',
+ 0x1CEA: 'Symbol',
+ 0x1CEB: 'Symbol',
+ 0x1CEC: 'Symbol',
+ 0x1CEE: 'Symbol',
+ 0x1CEF: 'Symbol',
+ 0x1CF0: 'Symbol',
+ 0x1CF1: 'Symbol',
+
+ 0x0A51: 'M', # https://github.com/harfbuzz/harfbuzz/issues/524
+
+ # According to ScriptExtensions.txt, these Grantha marks may also be used in Tamil,
+ # so the Indic shaper needs to know their categories.
+ 0x11301: 'SM',
+ 0x11302: 'SM',
+ 0x11303: 'SM',
+ 0x1133B: 'N',
+ 0x1133C: 'N',
+
+ 0x0AFB: 'N', # https://github.com/harfbuzz/harfbuzz/issues/552
+ 0x0B55: 'N', # https://github.com/harfbuzz/harfbuzz/issues/2849
+
+ 0x09FC: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/pull/1613
+ 0x0C80: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/pull/623
+ 0x0D04: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/pull/3511
+
+ 0x25CC: 'DOTTEDCIRCLE',
+
+
+ # Khmer
+
+ 0x179A: 'Ra',
+
+ 0x17CC: 'Robatic',
+ 0x17C9: 'Robatic',
+ 0x17CA: 'Robatic',
+
+ 0x17C6: 'Xgroup',
+ 0x17CB: 'Xgroup',
+ 0x17CD: 'Xgroup',
+ 0x17CE: 'Xgroup',
+ 0x17CF: 'Xgroup',
+ 0x17D0: 'Xgroup',
+ 0x17D1: 'Xgroup',
+
+ 0x17C7: 'Ygroup',
+ 0x17C8: 'Ygroup',
+ 0x17DD: 'Ygroup',
+ 0x17D3: 'Ygroup', # Just guessing. Uniscribe doesn't categorize it.
+
+ 0x17D9: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/issues/2384
+
+
+ # Myanmar
+
+ # https://docs.microsoft.com/en-us/typography/script-development/myanmar#analyze
+
+ 0x104E: 'C', # The spec says C, IndicSyllableCategory says Consonant_Placeholder
+
+ 0x1004: 'Ra',
+ 0x101B: 'Ra',
+ 0x105A: 'Ra',
+
+ 0x1032: 'A',
+ 0x1036: 'A',
+
+ 0x103A: 'As',
+
+ #0x1040: 'D0', # XXX The spec says D0, but Uniscribe doesn't seem to do.
+
+ 0x103E: 'MH',
+ 0x1060: 'ML',
+ 0x103C: 'MR',
+ 0x103D: 'MW',
+ 0x1082: 'MW',
+ 0x103B: 'MY',
+ 0x105E: 'MY',
+ 0x105F: 'MY',
+
+ 0x1063: 'PT',
+ 0x1064: 'PT',
+ 0x1069: 'PT',
+ 0x106A: 'PT',
+ 0x106B: 'PT',
+ 0x106C: 'PT',
+ 0x106D: 'PT',
+ 0xAA7B: 'PT',
+
+ 0x1038: 'SM',
+ 0x1087: 'SM',
+ 0x1088: 'SM',
+ 0x1089: 'SM',
+ 0x108A: 'SM',
+ 0x108B: 'SM',
+ 0x108C: 'SM',
+ 0x108D: 'SM',
+ 0x108F: 'SM',
+ 0x109A: 'SM',
+ 0x109B: 'SM',
+ 0x109C: 'SM',
+
+ 0x104A: 'PLACEHOLDER',
+}
+position_overrides = {
+
+ 0x0A51: 'BELOW_C', # https://github.com/harfbuzz/harfbuzz/issues/524
+
+ 0x0B01: 'BEFORE_SUB', # Oriya Bindu is BeforeSub in the spec.
+}
+
+def matra_pos_left(u, block):
+ return "PRE_M"
+def matra_pos_right(u, block):
+ if block == 'Devanagari': return 'AFTER_SUB'
+ if block == 'Bengali': return 'AFTER_POST'
+ if block == 'Gurmukhi': return 'AFTER_POST'
+ if block == 'Gujarati': return 'AFTER_POST'
+ if block == 'Oriya': return 'AFTER_POST'
+ if block == 'Tamil': return 'AFTER_POST'
+ if block == 'Telugu': return 'BEFORE_SUB' if u <= 0x0C42 else 'AFTER_SUB'
+ if block == 'Kannada': return 'BEFORE_SUB' if u < 0x0CC3 or u > 0x0CD6 else 'AFTER_SUB'
+ if block == 'Malayalam': return 'AFTER_POST'
+ return 'AFTER_SUB'
+def matra_pos_top(u, block):
+ # BENG and MLYM don't have top matras.
+ if block == 'Devanagari': return 'AFTER_SUB'
+ if block == 'Gurmukhi': return 'AFTER_POST' # Deviate from spec
+ if block == 'Gujarati': return 'AFTER_SUB'
+ if block == 'Oriya': return 'AFTER_MAIN'
+ if block == 'Tamil': return 'AFTER_SUB'
+ if block == 'Telugu': return 'BEFORE_SUB'
+ if block == 'Kannada': return 'BEFORE_SUB'
+ return 'AFTER_SUB'
+def matra_pos_bottom(u, block):
+ if block == 'Devanagari': return 'AFTER_SUB'
+ if block == 'Bengali': return 'AFTER_SUB'
+ if block == 'Gurmukhi': return 'AFTER_POST'
+ if block == 'Gujarati': return 'AFTER_POST'
+ if block == 'Oriya': return 'AFTER_SUB'
+ if block == 'Tamil': return 'AFTER_POST'
+ if block == 'Telugu': return 'BEFORE_SUB'
+ if block == 'Kannada': return 'BEFORE_SUB'
+ if block == 'Malayalam': return 'AFTER_POST'
+ return "AFTER_SUB"
+def indic_matra_position(u, pos, block): # Reposition matra
+ if pos == 'PRE_C': return matra_pos_left(u, block)
+ if pos == 'POST_C': return matra_pos_right(u, block)
+ if pos == 'ABOVE_C': return matra_pos_top(u, block)
+ if pos == 'BELOW_C': return matra_pos_bottom(u, block)
+ assert (False)
+
+def position_to_category(pos):
+ if pos == 'PRE_C': return 'VPre'
+ if pos == 'ABOVE_C': return 'VAbv'
+ if pos == 'BELOW_C': return 'VBlw'
+ if pos == 'POST_C': return 'VPst'
+ assert(False)
+
+
+defaults = (category_map[defaults[0]], position_map[defaults[1]], defaults[2])
+
+indic_data = {}
+for k, (cat, pos, block) in combined.items():
+ cat = category_map[cat]
+ pos = position_map[pos]
+ indic_data[k] = (cat, pos, block)
+
+for k,new_cat in category_overrides.items():
+ (cat, pos, _) = indic_data.get(k, defaults)
+ indic_data[k] = (new_cat, pos, unicode_data[2][k])
+
+# We only expect position for certain types
+positioned_categories = ('CM', 'SM', 'RS', 'H', 'M', 'MPst')
+for k, (cat, pos, block) in indic_data.items():
+ if cat not in positioned_categories:
+ pos = 'END'
+ indic_data[k] = (cat, pos, block)
+
+# Position overrides are more complicated
+
+# Keep in sync with CONSONANT_FLAGS in the shaper
+consonant_categories = ('C', 'CS', 'Ra','CM', 'V', 'PLACEHOLDER', 'DOTTEDCIRCLE')
+matra_categories = ('M', 'MPst')
+smvd_categories = ('SM', 'VD', 'A', 'Symbol')
+for k, (cat, pos, block) in indic_data.items():
+ if cat in consonant_categories:
+ pos = 'BASE_C'
+ elif cat in matra_categories:
+ if block.startswith('Khmer') or block.startswith('Myanmar'):
+ cat = position_to_category(pos)
+ else:
+ pos = indic_matra_position(k, pos, block)
+ elif cat in smvd_categories:
+ pos = 'SMVD';
+ indic_data[k] = (cat, pos, block)
+
+for k,new_pos in position_overrides.items():
+ (cat, pos, _) = indic_data.get(k, defaults)
+ indic_data[k] = (cat, new_pos, unicode_data[2][k])
+
+
+values = [{_: 1} for _ in defaults]
+for vv in indic_data.values():
+ for i,v in enumerate(vv):
+ values[i][v] = values[i].get (v, 0) + 1
+
+
+
# Move the outliers NO-BREAK SPACE and DOTTED CIRCLE out
singles = {}
for u in ALLOWED_SINGLES:
- singles[u] = data[u]
- del data[u]
+ singles[u] = indic_data[u]
+ del indic_data[u]
print ("/* == Start of generated table == */")
print ("/*")
@@ -106,37 +502,63 @@ print ('#include "hb.hh"')
print ()
print ('#ifndef HB_NO_OT_SHAPE')
print ()
-print ('#include "hb-ot-shape-complex-indic.hh"')
+print ('#include "hb-ot-shaper-indic.hh"')
+print ()
+print ('#pragma GCC diagnostic push')
+print ('#pragma GCC diagnostic ignored "-Wunused-macros"')
+print ()
+
+# Print categories
+for shaper in categories:
+ print ('#include "hb-ot-shaper-%s-machine.hh"' % shaper)
+print ()
+done = {}
+for shaper, shaper_cats in categories.items():
+ print ('/* %s */' % shaper)
+ for cat in shaper_cats:
+ v = shaper[0].upper()
+ if cat not in done:
+ print ("#define OT_%s %s_Cat(%s)" % (cat, v, cat))
+ done[cat] = v
+ else:
+ print ('static_assert (OT_%s == %s_Cat(%s), "");' % (cat, v, cat))
print ()
# Shorten values
short = [{
- "Bindu": 'Bi',
- "Cantillation_Mark": 'Ca',
- "Joiner": 'ZWJ',
- "Non_Joiner": 'ZWNJ',
- "Number": 'Nd',
- "Visarga": 'Vs',
- "Vowel": 'Vo',
- "Vowel_Dependent": 'M',
- "Consonant_Prefixed": 'CPrf',
- "Other": 'x',
+ "Repha": 'Rf',
+ "PLACEHOLDER": 'GB',
+ "DOTTEDCIRCLE": 'DC',
+ "VPst": 'VR',
+ "VPre": 'VL',
+ "Robatic": 'Rt',
+ "Xgroup": 'Xg',
+ "Ygroup": 'Yg',
+ "As": 'As',
},{
- "Not_Applicable": 'x',
+ "END": 'X',
+ "BASE_C": 'C',
+ "ABOVE_C": 'T',
+ "BELOW_C": 'B',
+ "POST_C": 'R',
+ "PRE_C": 'L',
+ "PRE_M": 'LM',
+ "AFTER_MAIN": 'A',
+ "AFTER_SUB": 'AS',
+ "BEFORE_SUB": 'BS',
+ "AFTER_POST": 'AP',
+ "SMVD": 'SM',
}]
all_shorts = [{},{}]
# Add some of the values, to make them more readable, and to avoid duplicates
-
for i in range (2):
for v,s in short[i].items ():
all_shorts[i][s] = v
-what = ["INDIC_SYLLABIC_CATEGORY", "INDIC_MATRA_CATEGORY"]
-what_short = ["ISC", "IMC"]
-print ('#pragma GCC diagnostic push')
-print ('#pragma GCC diagnostic ignored "-Wunused-macros"')
+what = ["OT", "POS"]
+what_short = ["_OT", "_POS"]
cat_defs = []
for i in range (2):
vv = sorted (values[i].keys ())
@@ -150,7 +572,7 @@ for i in range (2):
raise Exception ("Duplicate short value alias", v, all_shorts[i][s])
all_shorts[i][s] = v
short[i][v] = s
- cat_defs.append ((what_short[i] + '_' + s, what[i] + '_' + v.upper (), str (values[i][v]), v))
+ cat_defs.append ((what_short[i] + '_' + s, what[i] + '_' + (v.upper () if i else v), str (values[i][v]), v))
maxlen_s = max ([len (c[0]) for c in cat_defs])
maxlen_l = max ([len (c[1]) for c in cat_defs])
@@ -163,7 +585,9 @@ for s in what_short:
print ()
print ('#pragma GCC diagnostic pop')
print ()
-print ("#define _(S,M) INDIC_COMBINE_CATEGORIES (ISC_##S, IMC_##M)")
+print ("#define INDIC_COMBINE_CATEGORIES(S,M) ((S) | ((M) << 8))")
+print ()
+print ("#define _(S,M) INDIC_COMBINE_CATEGORIES (%s_##S, %s_##M)" % tuple(what_short))
print ()
print ()
@@ -193,7 +617,7 @@ def print_block (block, start, end, data):
if block:
last_block = block
-uu = sorted (data.keys ())
+uu = sorted (indic_data)
last = -100000
num = 0
@@ -204,17 +628,17 @@ print ("static const uint16_t indic_table[] = {")
for u in uu:
if u <= last:
continue
- block = data[u][2]
+ block = indic_data[u][2]
start = u//8*8
end = start+1
- while end in uu and block == data[end][2]:
+ while end in uu and block == indic_data[end][2]:
end += 1
end = (end-1)//8*8 + 7
if start != last + 1:
- if start - last <= 1+16*3:
- print_block (None, last+1, start-1, data)
+ if start - last <= 1+16*2:
+ print_block (None, last+1, start-1, indic_data)
else:
if last >= 0:
ends.append (last + 1)
@@ -224,7 +648,7 @@ for u in uu:
print ("#define indic_offset_0x%04xu %d" % (start, offset))
starts.append (start)
- print_block (block, start, end, data)
+ print_block (block, start, end, indic_data)
last = end
ends.append (last + 1)
offset += ends[-1] - starts[-1]
@@ -254,10 +678,11 @@ for p in sorted(pages):
print (" default:")
print (" break;")
print (" }")
-print (" return _(x,x);")
+print (" return _(X,X);")
print ("}")
print ()
print ("#undef _")
+print ("#undef INDIC_COMBINE_CATEGORIES")
for i in range (2):
print ()
vv = sorted (values[i].keys ())
@@ -269,6 +694,6 @@ print ('#endif')
print ()
print ("/* == End of generated table == */")
-# Maintain at least 30% occupancy in the table */
-if occupancy < 30:
+# Maintain at least 50% occupancy in the table */
+if occupancy < 50:
raise Exception ("Table too sparse, please investigate: ", occupancy)
generated by cgit v1.2.3 (git 2.25.1) at 2024-06-13 16:20:57 +0000