3 """usage: ./gen-indic-table.py IndicSyllabicCategory.txt IndicPositionalCategory.txt Blocks.txt
6 * https://unicode.org/Public/UCD/latest/ucd/IndicSyllabicCategory.txt
7 * https://unicode.org/Public/UCD/latest/ucd/IndicPositionalCategory.txt
8 * https://unicode.org/Public/UCD/latest/ucd/Blocks.txt
13 if len (sys.argv) != 4:
16 ALLOWED_SINGLES = [0x00A0, 0x25CC]
32 'General Punctuation',
33 'Superscripts and Subscripts',
34 'Devanagari Extended',
39 files = [open (x, encoding='utf-8') for x in sys.argv[1:]]
41 headers = [[f.readline () for i in range (2)] for f in files]
43 unicode_data = [{} for _ in files]
44 for i, f in enumerate (files):
51 fields = [x.strip () for x in line.split (';')]
55 uu = fields[0].split ('..')
56 start = int (uu[0], 16)
64 for u in range (start, end + 1):
65 unicode_data[i][u] = t
67 # Merge data into one dict:
68 defaults = ('Other', 'Not_Applicable', 'No_Block')
70 for i,d in enumerate (unicode_data):
71 for u,v in d.items ():
72 if i == 2 and not u in combined:
75 combined[u] = list (defaults)
77 combined = {k:v for k,v in combined.items() if k in ALLOWED_SINGLES or v[2] in ALLOWED_BLOCKS}
80 # Convert categories & positions types
137 'Avagraha' : 'Symbol',
139 'Brahmi_Joining_Number' : 'PLACEHOLDER', # Don't care.
140 'Cantillation_Mark' : 'A',
142 'Consonant_Dead' : 'C',
143 'Consonant_Final' : 'CM',
144 'Consonant_Head_Letter' : 'C',
145 'Consonant_Initial_Postfixed' : 'C', # TODO
146 'Consonant_Killer' : 'M', # U+17CD only.
147 'Consonant_Medial' : 'CM',
148 'Consonant_Placeholder' : 'PLACEHOLDER',
149 'Consonant_Preceding_Repha' : 'Repha',
150 'Consonant_Prefixed' : 'X', # Don't care.
151 'Consonant_Subjoined' : 'CM',
152 'Consonant_Succeeding_Repha' : 'CM',
153 'Consonant_With_Stacker' : 'CS',
154 'Gemination_Mark' : 'SM', # https://github.com/harfbuzz/harfbuzz/issues/552
155 'Invisible_Stacker' : 'H',
157 'Modifying_Letter' : 'X',
158 'Non_Joiner' : 'ZWNJ',
160 'Number' : 'PLACEHOLDER',
161 'Number_Joiner' : 'PLACEHOLDER', # Don't care.
162 'Pure_Killer' : 'M', # Is like a vowel matra.
163 'Register_Shifter' : 'RS',
164 'Syllable_Modifier' : 'SM',
170 'Vowel_Dependent' : 'M',
171 'Vowel_Independent' : 'V',
174 'Not_Applicable' : 'END',
178 'Bottom' : 'BELOW_C',
181 # These should resolve to the position of the last part of the split sequence.
182 'Bottom_And_Right' : 'POST_C',
183 'Left_And_Right' : 'POST_C',
184 'Top_And_Bottom' : 'BELOW_C',
185 'Top_And_Bottom_And_Left' : 'BELOW_C',
186 'Top_And_Bottom_And_Right' : 'POST_C',
187 'Top_And_Left' : 'ABOVE_C',
188 'Top_And_Left_And_Right' : 'POST_C',
189 'Top_And_Right' : 'POST_C',
191 'Overstruck' : 'AFTER_MAIN',
192 'Visual_order_left' : 'PRE_M',
195 category_overrides = {
197 # These are the variation-selectors. They only appear in the Myanmar grammar
198 # but are not Myanmar-specific
216 # These appear in the OT Myanmar spec, but are not Myanmar-specific
217 0x2015: 'PLACEHOLDER',
218 0x2022: 'PLACEHOLDER',
219 0x25FB: 'PLACEHOLDER',
220 0x25FC: 'PLACEHOLDER',
221 0x25FD: 'PLACEHOLDER',
222 0x25FE: 'PLACEHOLDER',
227 0x0930: 'Ra', # Devanagari
228 0x09B0: 'Ra', # Bengali
229 0x09F0: 'Ra', # Bengali
230 0x0A30: 'Ra', # Gurmukhi No Reph
231 0x0AB0: 'Ra', # Gujarati
232 0x0B30: 'Ra', # Oriya
233 0x0BB0: 'Ra', # Tamil No Reph
234 0x0C30: 'Ra', # Telugu Reph formed only with ZWJ
235 0x0CB0: 'Ra', # Kannada
236 0x0D30: 'Ra', # Malayalam No Reph, Logical Repha
238 # The following act more like the Bindus.
242 # U+0A40 GURMUKHI VOWEL SIGN II may be preceded by U+0A02 GURMUKHI SIGN BINDI.
245 # The following act like consonants.
251 # TODO: The following should only be allowed after a Visarga.
252 # For now, just treat them like regular tone marks.
261 # TODO: The following should only be allowed after some of
262 # the nasalization marks, maybe only for U+1CE9..U+1CF1.
263 # For now, just treat them like tone marks.
266 # The following take marks in standalone clusters, similar to Avagraha.
282 0x0A51: 'M', # https://github.com/harfbuzz/harfbuzz/issues/524
284 # According to ScriptExtensions.txt, these Grantha marks may also be used in Tamil,
285 # so the Indic shaper needs to know their categories.
292 0x0AFB: 'N', # https://github.com/harfbuzz/harfbuzz/issues/552
293 0x0B55: 'N', # https://github.com/harfbuzz/harfbuzz/issues/2849
295 0x09FC: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/pull/1613
296 0x0C80: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/pull/623
297 0x0D04: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/pull/3511
299 0x25CC: 'DOTTEDCIRCLE',
321 0x17D3: 'Ygroup', # Just guessing. Uniscribe doesn't categorize it.
323 0x17D9: 'PLACEHOLDER', # https://github.com/harfbuzz/harfbuzz/issues/2384
328 # https://docs.microsoft.com/en-us/typography/script-development/myanmar#analyze
330 0x104E: 'C', # The spec says C, IndicSyllableCategory says Consonant_Placeholder
341 #0x1040: 'D0', # XXX The spec says D0, but Uniscribe doesn't seem to do.
374 0x104A: 'PLACEHOLDER',
376 position_overrides = {
378 0x0A51: 'BELOW_C', # https://github.com/harfbuzz/harfbuzz/issues/524
380 0x0B01: 'BEFORE_SUB', # Oriya Bindu is BeforeSub in the spec.
383 def matra_pos_left(u, block):
385 def matra_pos_right(u, block):
386 if block == 'Devanagari': return 'AFTER_SUB'
387 if block == 'Bengali': return 'AFTER_POST'
388 if block == 'Gurmukhi': return 'AFTER_POST'
389 if block == 'Gujarati': return 'AFTER_POST'
390 if block == 'Oriya': return 'AFTER_POST'
391 if block == 'Tamil': return 'AFTER_POST'
392 if block == 'Telugu': return 'BEFORE_SUB' if u <= 0x0C42 else 'AFTER_SUB'
393 if block == 'Kannada': return 'BEFORE_SUB' if u < 0x0CC3 or u > 0x0CD6 else 'AFTER_SUB'
394 if block == 'Malayalam': return 'AFTER_POST'
396 def matra_pos_top(u, block):
397 # BENG and MLYM don't have top matras.
398 if block == 'Devanagari': return 'AFTER_SUB'
399 if block == 'Gurmukhi': return 'AFTER_POST' # Deviate from spec
400 if block == 'Gujarati': return 'AFTER_SUB'
401 if block == 'Oriya': return 'AFTER_MAIN'
402 if block == 'Tamil': return 'AFTER_SUB'
403 if block == 'Telugu': return 'BEFORE_SUB'
404 if block == 'Kannada': return 'BEFORE_SUB'
406 def matra_pos_bottom(u, block):
407 if block == 'Devanagari': return 'AFTER_SUB'
408 if block == 'Bengali': return 'AFTER_SUB'
409 if block == 'Gurmukhi': return 'AFTER_POST'
410 if block == 'Gujarati': return 'AFTER_POST'
411 if block == 'Oriya': return 'AFTER_SUB'
412 if block == 'Tamil': return 'AFTER_POST'
413 if block == 'Telugu': return 'BEFORE_SUB'
414 if block == 'Kannada': return 'BEFORE_SUB'
415 if block == 'Malayalam': return 'AFTER_POST'
417 def indic_matra_position(u, pos, block): # Reposition matra
418 if pos == 'PRE_C': return matra_pos_left(u, block)
419 if pos == 'POST_C': return matra_pos_right(u, block)
420 if pos == 'ABOVE_C': return matra_pos_top(u, block)
421 if pos == 'BELOW_C': return matra_pos_bottom(u, block)
424 def position_to_category(pos):
425 if pos == 'PRE_C': return 'VPre'
426 if pos == 'ABOVE_C': return 'VAbv'
427 if pos == 'BELOW_C': return 'VBlw'
428 if pos == 'POST_C': return 'VPst'
432 defaults = (category_map[defaults[0]], position_map[defaults[1]], defaults[2])
435 for k, (cat, pos, block) in combined.items():
436 cat = category_map[cat]
437 pos = position_map[pos]
438 indic_data[k] = (cat, pos, block)
440 for k,new_cat in category_overrides.items():
441 (cat, pos, _) = indic_data.get(k, defaults)
442 indic_data[k] = (new_cat, pos, unicode_data[2][k])
444 # We only expect position for certain types
445 positioned_categories = ('CM', 'SM', 'RS', 'H', 'M', 'MPst')
446 for k, (cat, pos, block) in indic_data.items():
447 if cat not in positioned_categories:
449 indic_data[k] = (cat, pos, block)
451 # Position overrides are more complicated
453 # Keep in sync with CONSONANT_FLAGS in the shaper
454 consonant_categories = ('C', 'CS', 'Ra','CM', 'V', 'PLACEHOLDER', 'DOTTEDCIRCLE')
455 matra_categories = ('M', 'MPst')
456 smvd_categories = ('SM', 'VD', 'A', 'Symbol')
457 for k, (cat, pos, block) in indic_data.items():
458 if cat in consonant_categories:
460 elif cat in matra_categories:
461 if block.startswith('Khmer') or block.startswith('Myanmar'):
462 cat = position_to_category(pos)
464 pos = indic_matra_position(k, pos, block)
465 elif cat in smvd_categories:
467 indic_data[k] = (cat, pos, block)
469 for k,new_pos in position_overrides.items():
470 (cat, pos, _) = indic_data.get(k, defaults)
471 indic_data[k] = (cat, new_pos, unicode_data[2][k])
474 values = [{_: 1} for _ in defaults]
475 for vv in indic_data.values():
476 for i,v in enumerate(vv):
477 values[i][v] = values[i].get (v, 0) + 1
482 # Move the outliers NO-BREAK SPACE and DOTTED CIRCLE out
484 for u in ALLOWED_SINGLES:
485 singles[u] = indic_data[u]
488 print ("/* == Start of generated table == */")
490 print (" * The following table is generated by running:")
492 print (" * ./gen-indic-table.py IndicSyllabicCategory.txt IndicPositionalCategory.txt Blocks.txt")
494 print (" * on files with these headers:")
498 print (" * %s" % (l.strip()))
501 print ('#include "hb.hh"')
503 print ('#ifndef HB_NO_OT_SHAPE')
505 print ('#include "hb-ot-shaper-indic.hh"')
507 print ('#pragma GCC diagnostic push')
508 print ('#pragma GCC diagnostic ignored "-Wunused-macros"')
512 for shaper in categories:
513 print ('#include "hb-ot-shaper-%s-machine.hh"' % shaper)
516 for shaper, shaper_cats in categories.items():
517 print ('/* %s */' % shaper)
518 for cat in shaper_cats:
519 v = shaper[0].upper()
521 print ("#define OT_%s %s_Cat(%s)" % (cat, v, cat))
524 print ('static_assert (OT_%s == %s_Cat(%s), "");' % (cat, v, cat))
531 "DOTTEDCIRCLE": 'DC',
554 # Add some of the values, to make them more readable, and to avoid duplicates
557 for v,s in short[i].items ():
561 what_short = ["_OT", "_POS"]
564 vv = sorted (values[i].keys ())
566 v_no_and = v.replace ('_And_', '_')
570 s = ''.join ([c for c in v_no_and if ord ('A') <= ord (c) <= ord ('Z')])
571 if s in all_shorts[i]:
572 raise Exception ("Duplicate short value alias", v, all_shorts[i][s])
575 cat_defs.append ((what_short[i] + '_' + s, what[i] + '_' + (v.upper () if i else v), str (values[i][v]), v))
577 maxlen_s = max ([len (c[0]) for c in cat_defs])
578 maxlen_l = max ([len (c[1]) for c in cat_defs])
579 maxlen_n = max ([len (c[2]) for c in cat_defs])
582 for c in [c for c in cat_defs if s in c[0]]:
583 print ("#define %s %s /* %s chars; %s */" %
584 (c[0].ljust (maxlen_s), c[1].ljust (maxlen_l), c[2].rjust (maxlen_n), c[3]))
586 print ('#pragma GCC diagnostic pop')
588 print ("#define INDIC_COMBINE_CATEGORIES(S,M) ((S) | ((M) << 8))")
590 print ("#define _(S,M) INDIC_COMBINE_CATEGORIES (%s_##S, %s_##M)" % tuple(what_short))
597 def print_block (block, start, end, data):
598 global total, used, last_block
599 if block and block != last_block:
602 print (" /* %s */" % block)
604 assert start % 8 == 0
605 assert (end+1) % 8 == 0
606 for u in range (start, end+1):
609 print (" /* %04X */" % u, end="")
612 d = data.get (u, defaults)
613 print ("%9s" % ("_(%s,%s)," % (short[0][d[0]], short[1][d[1]])), end="")
615 total += end - start + 1
620 uu = sorted (indic_data)
627 print ("static const uint16_t indic_table[] = {")
631 block = indic_data[u][2]
635 while end in uu and block == indic_data[end][2]:
637 end = (end-1)//8*8 + 7
639 if start != last + 1:
640 if start - last <= 1+16*2:
641 print_block (None, last+1, start-1, indic_data)
644 ends.append (last + 1)
645 offset += ends[-1] - starts[-1]
648 print ("#define indic_offset_0x%04xu %d" % (start, offset))
649 starts.append (start)
651 print_block (block, start, end, indic_data)
653 ends.append (last + 1)
654 offset += ends[-1] - starts[-1]
657 occupancy = used * 100. / total
659 print ("}; /* Table items: %d; occupancy: %d%% */" % (offset, occupancy))
662 print ("hb_indic_get_categories (hb_codepoint_t u)")
664 print (" switch (u >> %d)" % page_bits)
666 pages = set ([u>>page_bits for u in starts+ends+list (singles.keys ())])
667 for p in sorted(pages):
668 print (" case 0x%0Xu:" % p)
669 for u,d in singles.items ():
670 if p != u>>page_bits: continue
671 print (" if (unlikely (u == 0x%04Xu)) return _(%s,%s);" % (u, short[0][d[0]], short[1][d[1]]))
672 for (start,end) in zip (starts, ends):
673 if p not in [start>>page_bits, end>>page_bits]: continue
674 offset = "indic_offset_0x%04xu" % start
675 print (" if (hb_in_range<hb_codepoint_t> (u, 0x%04Xu, 0x%04Xu)) return indic_table[u - 0x%04Xu + %s];" % (start, end-1, start, offset))
681 print (" return _(X,X);")
685 print ("#undef INDIC_COMBINE_CATEGORIES")
688 vv = sorted (values[i].keys ())
690 print ("#undef %s_%s" %
691 (what_short[i], short[i][v]))
695 print ("/* == End of generated table == */")
697 # Maintain at least 50% occupancy in the table */
699 raise Exception ("Table too sparse, please investigate: ", occupancy)