1#!/usr/bin/env python
2
3import collections
4import copy
5import glob
6import itertools
7from os import path
8import sys
9from xml.etree import ElementTree
10
11from fontTools import ttLib
12
13EMOJI_VS = 0xFE0F
14
15LANG_TO_SCRIPT = {
16    'as': 'Beng',
17    'bn': 'Beng',
18    'cy': 'Latn',
19    'da': 'Latn',
20    'de': 'Latn',
21    'en': 'Latn',
22    'es': 'Latn',
23    'et': 'Latn',
24    'eu': 'Latn',
25    'fr': 'Latn',
26    'ga': 'Latn',
27    'gu': 'Gujr',
28    'hi': 'Deva',
29    'hr': 'Latn',
30    'hu': 'Latn',
31    'hy': 'Armn',
32    'ja': 'Jpan',
33    'kn': 'Knda',
34    'ko': 'Kore',
35    'ml': 'Mlym',
36    'mn': 'Cyrl',
37    'mr': 'Deva',
38    'nb': 'Latn',
39    'nn': 'Latn',
40    'or': 'Orya',
41    'pa': 'Guru',
42    'pt': 'Latn',
43    'sl': 'Latn',
44    'ta': 'Taml',
45    'te': 'Telu',
46    'tk': 'Latn',
47}
48
49def lang_to_script(lang_code):
50    lang = lang_code.lower()
51    while lang not in LANG_TO_SCRIPT:
52        hyphen_idx = lang.rfind('-')
53        assert hyphen_idx != -1, (
54            'We do not know what script the "%s" language is written in.'
55            % lang_code)
56        assumed_script = lang[hyphen_idx+1:]
57        if len(assumed_script) == 4 and assumed_script.isalpha():
58            # This is actually the script
59            return assumed_script.title()
60        lang = lang[:hyphen_idx]
61    return LANG_TO_SCRIPT[lang]
62
63
64def printable(inp):
65    if type(inp) is set:  # set of character sequences
66        return '{' + ', '.join([printable(seq) for seq in inp]) + '}'
67    if type(inp) is tuple:  # character sequence
68        return '<' + (', '.join([printable(ch) for ch in inp])) + '>'
69    else:  # single character
70        return 'U+%04X' % inp
71
72
73def open_font(font):
74    font_file, index = font
75    font_path = path.join(_fonts_dir, font_file)
76    if index is not None:
77        return ttLib.TTFont(font_path, fontNumber=index)
78    else:
79        return ttLib.TTFont(font_path)
80
81
82def get_best_cmap(font):
83    ttfont = open_font(font)
84    all_unicode_cmap = None
85    bmp_cmap = None
86    for cmap in ttfont['cmap'].tables:
87        specifier = (cmap.format, cmap.platformID, cmap.platEncID)
88        if specifier == (4, 3, 1):
89            assert bmp_cmap is None, 'More than one BMP cmap in %s' % (font, )
90            bmp_cmap = cmap
91        elif specifier == (12, 3, 10):
92            assert all_unicode_cmap is None, (
93                'More than one UCS-4 cmap in %s' % (font, ))
94            all_unicode_cmap = cmap
95
96    return all_unicode_cmap.cmap if all_unicode_cmap else bmp_cmap.cmap
97
98
99def get_variation_sequences_cmap(font):
100    ttfont = open_font(font)
101    vs_cmap = None
102    for cmap in ttfont['cmap'].tables:
103        specifier = (cmap.format, cmap.platformID, cmap.platEncID)
104        if specifier == (14, 0, 5):
105            assert vs_cmap is None, 'More than one VS cmap in %s' % (font, )
106            vs_cmap = cmap
107    return vs_cmap
108
109
110def get_emoji_map(font):
111    # Add normal characters
112    emoji_map = copy.copy(get_best_cmap(font))
113    reverse_cmap = {glyph: code for code, glyph in emoji_map.items()}
114
115    # Add variation sequences
116    vs_dict = get_variation_sequences_cmap(font).uvsDict
117    for vs in vs_dict:
118        for base, glyph in vs_dict[vs]:
119            if glyph is None:
120                emoji_map[(base, vs)] = emoji_map[base]
121            else:
122                emoji_map[(base, vs)] = glyph
123
124    # Add GSUB rules
125    ttfont = open_font(font)
126    for lookup in ttfont['GSUB'].table.LookupList.Lookup:
127        assert lookup.LookupType == 4, 'We only understand type 4 lookups'
128        for subtable in lookup.SubTable:
129            ligatures = subtable.ligatures
130            for first_glyph in ligatures:
131                for ligature in ligatures[first_glyph]:
132                    sequence = [first_glyph] + ligature.Component
133                    sequence = [reverse_cmap[glyph] for glyph in sequence]
134                    sequence = tuple(sequence)
135                    # Make sure no starting subsequence of 'sequence' has been
136                    # seen before.
137                    for sub_len in range(2, len(sequence)+1):
138                        subsequence = sequence[:sub_len]
139                        assert subsequence not in emoji_map
140                    emoji_map[sequence] = ligature.LigGlyph
141
142    return emoji_map
143
144
145def assert_font_supports_any_of_chars(font, chars):
146    best_cmap = get_best_cmap(font)
147    for char in chars:
148        if char in best_cmap:
149            return
150    sys.exit('None of characters in %s were found in %s' % (chars, font))
151
152
153def assert_font_supports_all_of_chars(font, chars):
154    best_cmap = get_best_cmap(font)
155    for char in chars:
156        assert char in best_cmap, (
157            'U+%04X was not found in %s' % (char, font))
158
159
160def assert_font_supports_none_of_chars(font, chars):
161    best_cmap = get_best_cmap(font)
162    for char in chars:
163        assert char not in best_cmap, (
164            'U+%04X was found in %s' % (char, font))
165
166
167def assert_font_supports_all_sequences(font, sequences):
168    vs_dict = get_variation_sequences_cmap(font).uvsDict
169    for base, vs in sorted(sequences):
170        assert vs in vs_dict and (base, None) in vs_dict[vs], (
171            '<U+%04X, U+%04X> was not found in %s' % (base, vs, font))
172
173
174def check_hyphens(hyphens_dir):
175    # Find all the scripts that need automatic hyphenation
176    scripts = set()
177    for hyb_file in glob.iglob(path.join(hyphens_dir, '*.hyb')):
178        hyb_file = path.basename(hyb_file)
179        assert hyb_file.startswith('hyph-'), (
180            'Unknown hyphenation file %s' % hyb_file)
181        lang_code = hyb_file[hyb_file.index('-')+1:hyb_file.index('.')]
182        scripts.add(lang_to_script(lang_code))
183
184    HYPHENS = {0x002D, 0x2010}
185    for script in scripts:
186        fonts = _script_to_font_map[script]
187        assert fonts, 'No fonts found for the "%s" script' % script
188        for font in fonts:
189            assert_font_supports_any_of_chars(font, HYPHENS)
190
191
192class FontRecord(object):
193    def __init__(self, name, scripts, variant, weight, style, font):
194        self.name = name
195        self.scripts = scripts
196        self.variant = variant
197        self.weight = weight
198        self.style = style
199        self.font = font
200
201
202def parse_fonts_xml(fonts_xml_path):
203    global _script_to_font_map, _fallback_chain
204    _script_to_font_map = collections.defaultdict(set)
205    _fallback_chain = []
206    tree = ElementTree.parse(fonts_xml_path)
207    for family in tree.findall('family'):
208        name = family.get('name')
209        variant = family.get('variant')
210        langs = family.get('lang')
211        if name:
212            assert variant is None, (
213                'No variant expected for LGC font %s.' % name)
214            assert langs is None, (
215                'No language expected for LGC fonts %s.' % name)
216        else:
217            assert variant in {None, 'elegant', 'compact'}, (
218                'Unexpected value for variant: %s' % variant)
219
220        if langs:
221            langs = langs.split()
222            scripts = {lang_to_script(lang) for lang in langs}
223        else:
224            scripts = set()
225
226        for child in family:
227            assert child.tag == 'font', (
228                'Unknown tag <%s>' % child.tag)
229            font_file = child.text
230            weight = int(child.get('weight'))
231            assert weight % 100 == 0, (
232                'Font weight "%d" is not a multiple of 100.' % weight)
233
234            style = child.get('style')
235            assert style in {'normal', 'italic'}, (
236                'Unknown style "%s"' % style)
237
238            index = child.get('index')
239            if index:
240                index = int(index)
241
242            _fallback_chain.append(FontRecord(
243                name,
244                frozenset(scripts),
245                variant,
246                weight,
247                style,
248                (font_file, index)))
249
250            if name: # non-empty names are used for default LGC fonts
251                map_scripts = {'Latn', 'Grek', 'Cyrl'}
252            else:
253                map_scripts = scripts
254            for script in map_scripts:
255                _script_to_font_map[script].add((font_file, index))
256
257
258def check_emoji_coverage(all_emoji, equivalent_emoji):
259    emoji_fonts = [
260        record.font for record in _fallback_chain
261        if 'Zsye' in record.scripts]
262    assert len(emoji_fonts) == 1, 'There are %d emoji fonts.' % len(emoji_fonts)
263    emoji_font = emoji_fonts[0]
264    coverage = get_emoji_map(emoji_font)
265
266    for sequence in all_emoji:
267        assert sequence in coverage, (
268            '%s is not supported in the emoji font.' % printable(sequence))
269
270    for sequence in coverage:
271        if sequence in {0x0000, 0x000D, 0x0020}:
272            # The font needs to support a few extra characters, which is OK
273            continue
274        assert sequence in all_emoji, (
275            'Emoji font should not support %s.' % printable(sequence))
276
277    for first, second in sorted(equivalent_emoji.items()):
278        assert coverage[first] == coverage[second], (
279            '%s and %s should map to the same glyph.' % (
280                printable(first),
281                printable(second)))
282
283    for glyph in set(coverage.values()):
284        maps_to_glyph = [seq for seq in coverage if coverage[seq] == glyph]
285        if len(maps_to_glyph) > 1:
286            # There are more than one sequences mapping to the same glyph. We
287            # need to make sure they were expected to be equivalent.
288            equivalent_seqs = set()
289            for seq in maps_to_glyph:
290                equivalent_seq = seq
291                while equivalent_seq in equivalent_emoji:
292                    equivalent_seq = equivalent_emoji[equivalent_seq]
293                equivalent_seqs.add(equivalent_seq)
294            assert len(equivalent_seqs) == 1, (
295                'The sequences %s should not result in the same glyph %s' % (
296                    printable(equivalent_seqs),
297                    glyph))
298
299
300def check_emoji_defaults(default_emoji):
301    missing_text_chars = _emoji_properties['Emoji'] - default_emoji
302    emoji_font_seen = False
303    for record in _fallback_chain:
304        if 'Zsye' in record.scripts:
305            emoji_font_seen = True
306            # No need to check the emoji font
307            continue
308        # For later fonts, we only check them if they have a script
309        # defined, since the defined script may get them to a higher
310        # score even if they appear after the emoji font.
311        if emoji_font_seen and not record.scripts:
312            continue
313
314        # Check default emoji-style characters
315        assert_font_supports_none_of_chars(record.font, sorted(default_emoji))
316
317        # Mark default text-style characters appearing in fonts above the emoji
318        # font as seen
319        if not emoji_font_seen:
320            missing_text_chars -= set(get_best_cmap(record.font))
321
322    # Noto does not have monochrome glyphs for Unicode 7.0 wingdings and
323    # webdings yet.
324    missing_text_chars -= _chars_by_age['7.0']
325    # TODO: Remove these after b/26113320 is fixed
326    missing_text_chars -= {
327        0x263A, # WHITE SMILING FACE
328        0x270C, # VICTORY HAND
329        0x2744, # SNOWFLAKE
330        0x2764, # HEAVY BLACK HEART
331    }
332    assert missing_text_chars == set(), (
333        'Text style version of some emoji characters are missing: ' + repr(missing_text_chars))
334
335
336# Setting reverse to true returns a dictionary that maps the values to sets of
337# characters, useful for some binary properties. Otherwise, we get a
338# dictionary that maps characters to the property values, assuming there's only
339# one property in the file.
340def parse_unicode_datafile(file_path, reverse=False):
341    if reverse:
342        output_dict = collections.defaultdict(set)
343    else:
344        output_dict = {}
345    with open(file_path) as datafile:
346        for line in datafile:
347            if '#' in line:
348                line = line[:line.index('#')]
349            line = line.strip()
350            if not line:
351                continue
352
353            chars, prop = line.split(';')
354            chars = chars.strip()
355            prop = prop.strip()
356
357            if ' ' in chars:  # character sequence
358                sequence = [int(ch, 16) for ch in chars.split(' ')]
359                additions = [tuple(sequence)]
360            elif '..' in chars:  # character range
361                char_start, char_end = chars.split('..')
362                char_start = int(char_start, 16)
363                char_end = int(char_end, 16)
364                additions = xrange(char_start, char_end+1)
365            else:  # singe character
366                additions = [int(chars, 16)]
367            if reverse:
368                output_dict[prop].update(additions)
369            else:
370                for addition in additions:
371                    assert addition not in output_dict
372                    output_dict[addition] = prop
373    return output_dict
374
375
376def parse_standardized_variants(file_path):
377    emoji_set = set()
378    text_set = set()
379    with open(file_path) as datafile:
380        for line in datafile:
381            if '#' in line:
382                line = line[:line.index('#')]
383            line = line.strip()
384            if not line:
385                continue
386            sequence, description, _ = line.split(';')
387            sequence = sequence.strip().split(' ')
388            base = int(sequence[0], 16)
389            vs = int(sequence[1], 16)
390            description = description.strip()
391            if description == 'text style':
392                text_set.add((base, vs))
393            elif description == 'emoji style':
394                emoji_set.add((base, vs))
395    return text_set, emoji_set
396
397
398def parse_ucd(ucd_path):
399    global _emoji_properties, _chars_by_age
400    global _text_variation_sequences, _emoji_variation_sequences
401    global _emoji_sequences, _emoji_zwj_sequences
402    _emoji_properties = parse_unicode_datafile(
403        path.join(ucd_path, 'emoji-data.txt'), reverse=True)
404    _chars_by_age = parse_unicode_datafile(
405        path.join(ucd_path, 'DerivedAge.txt'), reverse=True)
406    sequences = parse_standardized_variants(
407        path.join(ucd_path, 'StandardizedVariants.txt'))
408    _text_variation_sequences, _emoji_variation_sequences = sequences
409    _emoji_sequences = parse_unicode_datafile(
410        path.join(ucd_path, 'emoji-sequences.txt'))
411    _emoji_zwj_sequences = parse_unicode_datafile(
412        path.join(ucd_path, 'emoji-zwj-sequences.txt'))
413
414    # filter modern pentathlon, as it seems likely to be removed from final spec
415    # also filter rifle
416    def is_excluded(n):
417        return n in [0x1f93b, 0x1f946]
418
419    def contains_excluded(t):
420        if type(t) == int:
421            return is_excluded(t)
422        return any(is_excluded(cp) for cp in t)
423
424    # filter modern pentathlon, as it seems likely to be removed from final spec
425    _emoji_properties['Emoji'] = set(
426        t for t in _emoji_properties['Emoji'] if not contains_excluded(t))
427    _emoji_sequences = dict(
428        (t, v) for (t, v) in _emoji_sequences.items() if not contains_excluded(t))
429
430def flag_sequence(territory_code):
431    return tuple(0x1F1E6 + ord(ch) - ord('A') for ch in territory_code)
432
433
434UNSUPPORTED_FLAGS = frozenset({
435    flag_sequence('BL'), flag_sequence('BQ'), flag_sequence('DG'),
436    flag_sequence('EA'), flag_sequence('EH'), flag_sequence('FK'),
437    flag_sequence('GF'), flag_sequence('GP'), flag_sequence('GS'),
438    flag_sequence('MF'), flag_sequence('MQ'), flag_sequence('NC'),
439    flag_sequence('PM'), flag_sequence('RE'), flag_sequence('TF'),
440    flag_sequence('WF'), flag_sequence('XK'), flag_sequence('YT'),
441})
442
443EQUIVALENT_FLAGS = {
444    flag_sequence('BV'): flag_sequence('NO'),
445    flag_sequence('CP'): flag_sequence('FR'),
446    flag_sequence('HM'): flag_sequence('AU'),
447    flag_sequence('SJ'): flag_sequence('NO'),
448    flag_sequence('UM'): flag_sequence('US'),
449}
450
451COMBINING_KEYCAP = 0x20E3
452
453LEGACY_ANDROID_EMOJI = {
454    0xFE4E5: flag_sequence('JP'),
455    0xFE4E6: flag_sequence('US'),
456    0xFE4E7: flag_sequence('FR'),
457    0xFE4E8: flag_sequence('DE'),
458    0xFE4E9: flag_sequence('IT'),
459    0xFE4EA: flag_sequence('GB'),
460    0xFE4EB: flag_sequence('ES'),
461    0xFE4EC: flag_sequence('RU'),
462    0xFE4ED: flag_sequence('CN'),
463    0xFE4EE: flag_sequence('KR'),
464    0xFE82C: (ord('#'), COMBINING_KEYCAP),
465    0xFE82E: (ord('1'), COMBINING_KEYCAP),
466    0xFE82F: (ord('2'), COMBINING_KEYCAP),
467    0xFE830: (ord('3'), COMBINING_KEYCAP),
468    0xFE831: (ord('4'), COMBINING_KEYCAP),
469    0xFE832: (ord('5'), COMBINING_KEYCAP),
470    0xFE833: (ord('6'), COMBINING_KEYCAP),
471    0xFE834: (ord('7'), COMBINING_KEYCAP),
472    0xFE835: (ord('8'), COMBINING_KEYCAP),
473    0xFE836: (ord('9'), COMBINING_KEYCAP),
474    0xFE837: (ord('0'), COMBINING_KEYCAP),
475}
476
477ZWJ_IDENTICALS = {
478    # KISS
479    (0x1F469, 0x200D, 0x2764, 0x200D, 0x1F48B, 0x200D, 0x1F468): 0x1F48F,
480    # COUPLE WITH HEART
481    (0x1F469, 0x200D, 0x2764, 0x200D, 0x1F468): 0x1F491,
482    # FAMILY
483    (0x1F468, 0x200D, 0x1F469, 0x200D, 0x1F466): 0x1F46A,
484}
485
486def compute_expected_emoji():
487    equivalent_emoji = {}
488    sequence_pieces = set()
489    all_sequences = set()
490    all_sequences.update(_emoji_variation_sequences)
491
492    for sequence in _emoji_sequences.keys():
493        sequence = tuple(ch for ch in sequence if ch != EMOJI_VS)
494        all_sequences.add(sequence)
495        sequence_pieces.update(sequence)
496
497    for sequence in _emoji_zwj_sequences.keys():
498        sequence = tuple(ch for ch in sequence if ch != EMOJI_VS)
499        all_sequences.add(sequence)
500        sequence_pieces.update(sequence)
501        # Add reverse of all emoji ZWJ sequences, which are added to the fonts
502        # as a workaround to get the sequences work in RTL text.
503        reversed_seq = tuple(reversed(sequence))
504        all_sequences.add(reversed_seq)
505        equivalent_emoji[reversed_seq] = sequence
506
507    # Add all two-letter flag sequences, as even the unsupported ones should
508    # resolve to a flag tofu.
509    all_letters = [chr(code) for code in range(ord('A'), ord('Z')+1)]
510    all_two_letter_codes = itertools.product(all_letters, repeat=2)
511    all_flags = {flag_sequence(code) for code in all_two_letter_codes}
512    all_sequences.update(all_flags)
513    tofu_flags = UNSUPPORTED_FLAGS | (all_flags - set(_emoji_sequences.keys()))
514
515    all_emoji = (
516        _emoji_properties['Emoji'] |
517        all_sequences |
518        sequence_pieces |
519        set(LEGACY_ANDROID_EMOJI.keys()))
520    default_emoji = (
521        _emoji_properties['Emoji_Presentation'] |
522        all_sequences |
523        set(LEGACY_ANDROID_EMOJI.keys()))
524
525    first_tofu_flag = sorted(tofu_flags)[0]
526    for flag in tofu_flags:
527        if flag != first_tofu_flag:
528            equivalent_emoji[flag] = first_tofu_flag
529    equivalent_emoji.update(EQUIVALENT_FLAGS)
530    equivalent_emoji.update(LEGACY_ANDROID_EMOJI)
531    equivalent_emoji.update(ZWJ_IDENTICALS)
532    for seq in _emoji_variation_sequences:
533        equivalent_emoji[seq] = seq[0]
534
535    return all_emoji, default_emoji, equivalent_emoji
536
537
538def main():
539    target_out = sys.argv[1]
540    global _fonts_dir
541    _fonts_dir = path.join(target_out, 'fonts')
542
543    fonts_xml_path = path.join(target_out, 'etc', 'fonts.xml')
544    parse_fonts_xml(fonts_xml_path)
545
546    hyphens_dir = path.join(target_out, 'usr', 'hyphen-data')
547    check_hyphens(hyphens_dir)
548
549    check_emoji = sys.argv[2]
550    if check_emoji == 'true':
551        ucd_path = sys.argv[3]
552        parse_ucd(ucd_path)
553        all_emoji, default_emoji, equivalent_emoji = compute_expected_emoji()
554        check_emoji_coverage(all_emoji, equivalent_emoji)
555        check_emoji_defaults(default_emoji)
556
557
558if __name__ == '__main__':
559    main()
560