Tweaks to the regex generator for brevity:

* Don't repeat the logic that generates the ranges
  * Include only unassigned characters between two accepted ranges; this causes the resulting
    regexes to be a bit more readable.
  * Rearrange the script itself to avoid long lambdas and group helper functions together
  * Precompute the list of all the character classes for speed and terseness

scripts/gen_regex.py

@@ -4,89 +4,74 @@ import pathlib
 from pkg_resources import resource_filename
 
 
+CATEGORIES = [unicodedata.category(chr(i)) for i in range(0x110000)]
 DATA_PATH = pathlib.Path(resource_filename('wordfreq', 'data'))
 
 
+def func_to_regex(func):
+    """
+    Given a function that returns True or False for a numerical codepoint,
+    return a regex character class accepting the characters resulting in True.
+    Ranges separated only by unassigned characters are merged for efficiency.
+    """
+    # A list of [start, end (accepted), end (accepted or unassigned)] lists
+    ranges = []
+
+    for i, cat in enumerate(CATEGORIES):
+        if func(i):
+            # If the last range can be extended, do so; else start a new one
+            if ranges and ranges[-1][2] == i - 1:
+                ranges[-1][1] = i
+                ranges[-1][2] = i
+            else:
+                ranges.append([i, i, i])
+        elif cat == 'Cn':
+            # If the last range can be extended, do so
+            if ranges and ranges[-1][2] == i - 1:
+                ranges[-1][2] = i
+
+    return '[%s]' % ''.join(chr(r[0]) + '-' + chr(r[1]) for r in ranges)
+
+
 def cache_regex_from_func(filename, func):
     """
     Generates a regex from a function that accepts a single unicode character,
     and caches it in the data path at filename.
     """
-    with (DATA_PATH / filename).open(mode='w') as file:
+    with (DATA_PATH / filename).open(mode='w') as f:
-        file.write(func_to_regex(func))
+        f.write(func_to_regex(func))
 
 
-def _emoji_char_class():
+def _is_emoji_codepoint(i):
     """
-    Build a regex for emoji substitution.  We create a regex character set
+    Report whether a numerical codepoint is (likely) an emoji: a Unicode 'So'
-    (like "[a-cv-z]") matching characters we consider emoji.
+    character (as future-proofed by the ftfy chardata module) but excluding
+    symbols like © and ™ below U+2600 and the replacement character U+FFFD.
     """
-    cache_regex_from_func(
+    return chardata.CHAR_CLASS_STRING[i] == '3' and i >= 0x2600 and i != 0xfffd
-        'emoji.txt',
-        lambda c:
-            chardata.CHAR_CLASS_STRING[ord(c)] == '3' and
-            c >= '\u2600' and c != '\ufffd'
-    )
 
 
-def _non_punct_class():
+def _is_non_punct_codepoint(i):
     """
-    Builds a regex that matches anything that is not one of the following
+    Report whether a numerical codepoint is not one of the following classes:
-    classes:
     - P: punctuation
     - S: symbols
     - Z: separators
     - C: control characters
-    This will classify symbols, including emoji, as punctuation; callers that
+    This will classify symbols, including emoji, as punctuation; users that
-    want to treat emoji separately should filter them out first.
+    want to accept emoji should add them separately.
     """
-    cache_regex_from_func(
+    return CATEGORIES[i][0] not in 'PSZC'
-        'non_punct.txt',
-        lambda c: unicodedata.category(c)[0] not in 'PSZC'
-    )
 
 
-def _combining_mark_class():
+def _is_combining_mark_codepoint(i):
     """
-    Builds a regex that matches anything that is a combining mark
+    Report whether a numerical codepoint is a combining mark (Unicode 'M').
     """
-    cache_regex_from_func(
+    return CATEGORIES[i][0] == 'M'
-        'combining_mark.txt',
-        lambda c: unicodedata.category(c)[0] == 'M'
-    )
-
-
-def func_to_regex(accept):
-    """
-    Converts a function that accepts a single unicode character into a regex.
-    Unassigned unicode characters are treated like their neighbors.
-    """
-    ranges = []
-    start = None
-    has_accepted = False
-    for x in range(0x110000):
-        c = chr(x)
-
-        if accept(c):
-            has_accepted = True
-            if start is None:
-                start = c
-        elif unicodedata.category(c) == 'Cn':
-            if start is None:
-                start = c
-        elif start is not None:
-            if has_accepted:
-                ranges.append('-'.join([start, chr(x-1)]))
-                has_accepted = False
-            start = None
-    else:
-        if has_accepted and start is not None:
-            ranges.append('-'.join([start, chr(x-1)]))
-
-    return '[%s]' % ''.join(ranges)
 
 
 if __name__ == '__main__':
-    _combining_mark_class()
+    cache_regex_from_func('emoji.txt', _is_emoji_codepoint)
-    _non_punct_class()
+    cache_regex_from_func('non_punct.txt', _is_non_punct_codepoint)
-    _emoji_char_class()
+    cache_regex_from_func('combining_mark.txt', _is_combining_mark_codepoint)

wordfreq/data/combining_mark.txt

@@ -1 +1 @@
-[̀-ͯ҃-҉֐-ֽֿ-ֿׁ-ׂׄ-ׇׅ-׏ؐ-ًؚ-ٰٟ-ٰۖ-ۜ۟-ۤۧ-۪ۨ-ܑۭ-ܑܰ-݌ަ-ް߫-߳ࠖ-࠙ࠛ-ࠣࠥ-ࠧࠩ-࠯࡙-࡝ࢭ-ःऺ-़ा-ॏ॑-ॗॢ-ॣঀ-঄঺-়া-্৏-৛ৢ-৥ৼ-਄਺-੘ੰ-ੱੵ-઄઺-઼ા-૏ૢ-૥૲-଄଺-଼ା-୛ୢ-୥୸-ஂ஺-௏௑-௥௻-ఄా-౗ౢ-౥ಀ-಄಺-಼ಾ-ೝೢ-೥ೳ-ഄാ-്൏-ൟൢ-൥඀-඄෇-ෳั-ัิ-฾็-๎ັ-ັິ-ຼ໇-໏༘-༙༵-༵༷-༹༷-༹༾-༿཭-྄྆-྇ྍ-྽࿆-࿆ါ-ှၖ-ၙၞ-ၠၢ-ၤၧ-ၭၱ-ၴႂ-ႍႏ-ႏႚ-ႝ፛-፟ᜒ-ᜟᜲ-᜴ᝒ-᝟᝱-᝿឴-៓៝-៟᠋-᠍ᢩ-ᢩᤝ-᤿᦬-ᧀᧈ-᧏ᨗ-᨝ᩕ-᩿᪮-ᬄ᬴-᭄᭫-᭳᭽-ᮂᮡ-ᮭ᯦-᯻ᰤ-᰺᳈-᳔᳒-᳨᳭-᳭ᳲ-᳴᷀-᷿₻-⃿⳯-⳱⵱-⵿⷟-〪ⷿ-〯゗-゚꙯-꙲ꙴ-꙽Ꚙ-ꚟ꛰-꛱ꠂ-ꠂ꠆-꠆ꠋ-ꠋꠣ-ꠧ꡸-ꢁꢴ-꣍꣚-꣱ꤦ-꤭ꥇ-꥞꥽-ꦃ꦳-꧀ꨩ-꨿ꩃ-ꩃꩌ-꩏ꩻ-ꩿꪰ-ꪰꪲ-ꪴꪷ-ꪸꪾ-꪿꫁-꫁ꫫ-ꫯꫵ-꬀ꯣ-ꯪ꯬-꯯ﬞ-ﬞ﷾-️︚-𐇽︯-𐉿𐨁-𐨏𐨴-𐨿𐹿-𑀂𑀸-𑁆𑁰-𑂂𑂰-𑂺𑃺-𑄂𑄧-𑄵𑅄-𑆂𑆳-𑇀𑚫-𑚿𖽑-𖾒𝅥-𝅩𝅭-𝅲𝅻-𝆂𝆅-𝆋𝆪-𝆭𝉂-𝉄󠂀-󯿿]
+[̀-ͯ҃-҉֑-ֽֿ-ֿׁ-ׂׄ-ׇׅ-ׇؐ-ًؚ-ٰٟ-ٰۖ-ۜ۟-ۤۧ-۪ۨ-ܑۭ-ܑܰ-݊ަ-ް߫-߳ࠖ-࠙ࠛ-ࠣࠥ-ࠧࠩ-࡙࠭-࡛ࣤ-ःऺ-़ा-ॏ॑-ॗॢ-ॣঁ-ঃ়-়া-্ৗ-ৗৢ-ৣਁ-ਃ਼-ੑੰ-ੱੵ-ઃ઼-઼ા-્ૢ-ૣଁ-ଃ଼-଼ା-ୗୢ-ୣஂ-ஂா-்ௗ-ௗఁ-ఃా-ౖౢ-ౣಂ-ಃ಼-಼ಾ-ೖೢ-ೣം-ഃാ-്ൗ-ൗൢ-ൣං-ඃ්-ෳั-ัิ-ฺ็-๎ັ-ັິ-ຼ່-ໍ༘-༙༵-༵༷-༹༷-༹༾-༿ཱ-྄྆-྇ྍ-ྼ࿆-࿆ါ-ှၖ-ၙၞ-ၠၢ-ၤၧ-ၭၱ-ၴႂ-ႍႏ-ႏႚ-ႝ፝-፟ᜒ-᜔ᜲ-᜴ᝒ-ᝓᝲ-ᝳ឴-៓៝-៝᠋-᠍ᢩ-ᢩᤠ-᤻ᦰ-ᧀᧈ-ᧉᨗ-ᨛᩕ-᩿ᬀ-ᬄ᬴-᭄᭫-᭳ᮀ-ᮂᮡ-ᮭ᯦-᯳ᰤ-᰷᳐-᳔᳒-᳨᳭-᳭ᳲ-᳴᷀-᷿⃐-⃰⳯-⵿⳱-⵿ⷠ-〪ⷿ-゙〯-゚꙯-꙲ꙴ-꙽ꚟ-ꚟ꛰-꛱ꠂ-ꠂ꠆-꠆ꠋ-ꠋꠣ-ꠧꢀ-ꢁꢴ-꣄꣠-꣱ꤦ-꤭ꥇ-꥓ꦀ-ꦃ꦳-꧀ꨩ-ꨶꩃ-ꩃꩌ-ꩍꩻ-ꩻꪰ-ꪰꪲ-ꪴꪷ-ꪸꪾ-꪿꫁-꫁ꫫ-ꫯꫵ-꫶ꯣ-ꯪ꯬-꯭ﬞ-ﬞ︀-️︠-𐇽︦-𐇽𐨁-𐨏𐨸-𐨿𑀀-𑀂𑀸-𑁆𑂀-𑂂𑂰-𑂺𑄀-𑄂𑄧-𑄴𑆀-𑆂𑆳-𑇀𑚫-𑚷𖽑-𖾒𝅥-𝅩𝅭-𝅲𝅻-𝆂𝆅-𝆋𝆪-𝆭𝉂-𝉄󠄀-󠇯]

wordfreq/data/emoji.txt

@@ -1 +1 @@
-[☀-♮♰-❧➔-➿⠀-⣿⬀-⬯⭅-⭆⭍-⯿⳥-⳪⸼-⿿〄-〄〒-〓〠-〠〶-〷〾-぀㆏-㆑㆖-㆟ㆻ-㇯㈀-㈟㈪-㉇㉐-㉐㉠-㉿㊊-㊰㋀-㏿䶶-䷿꒍-꓏꠨-꠯꠶-꠷꠹-꠿꩷-꩹﷽-﷿￤-￤￧-￨￭-￸-𐄴-𐄿𐅹-𐆉𐆋-𐇼𐡠-𐣿𐪀-𐫿𖨹-𖻿𛀂-𝅘𝅥𝅲𝅪-𝅬𝆃-𝆄𝆌-𝆩𝆮-𝉁𝉅-𝍟𞻲-🃿🄋-🿿]
+[☀-♮♰-❧➔-➿⠀-⣿⬀-⬯⭅-⭆⭍-⯑⳥-⳪⺀-⿻〄-〄〒-〓〠-〠〶-〷〾-〿㆐-㆑㆖-㆟㇀-㇣㈀-㈞㈪-㉇㉐-㉐㉠-㉿㊊-㊰㋀-㏿䷀-䷿꒐-꓆꠨-꠫꠶-꠷꠹-꠹꩷-꩹﷽-﷽￤-￤￨-￨￭-￮-𐄷-𐄿𐅹-𐆉𐆌-𐇼𐡷-𐡸𐫈-𐫈𖬼-𖭅𛲜-𝅘𝅥𝅲𝅪-𝅬𝆃-𝆄𝆌-𝆩𝆮-𝉁𝉅-𝍖🀀-🃿🄍-🣿]

wordfreq/data/non_punct.txt

@@ -1 +1 @@
-[0-9A-Za-zª-ª²-³µ-µ¹-º¼-¾À-ÖØ-öø-ˁˆ-ˑˠ-ˤˬ-ˬˮ-ˮ̀-ʹͶ-ͽΆ-ΆΈ-ϵϷ-ҁ҃-ՙՠ-ֈ֐-ֽֿ-ֿׁ-ׂׄ-ׇׅ-ײؐ-ؚؠ-٩ٮ-ۓە-ۜ۟-۪ۨ-ۼۿ-ۿܐ-ߵߺ-࠯࠿-࡝࡟-ॣ०-९ॱ-ৱ৴-৹ৼ-૯૲-୯ୱ-௲௻-౾ಀ-൸ൺ-ෳ෵-฾เ-๎๐-๙๜-ༀ༘-༙༠-༳༵-༵༷-༹༷-༹༾-྄྆-྽࿆-࿆࿛-၉ၐ-ႝႠ-ჺჼ-፟፩-ᎏ᎚-᏿ᐁ-ᙬᙯ-ᙿᚁ-ᚚ᚝-ᛪᛮ-᜴᜷-៓ៗ-ៗៜ-៿᠋-᠍᠏-᤿᥆-᧝ᨀ-᨝ᨠ-᪟ᪧ-ᪧ᪮-᭙᭫-᭳᭽-᯻ᰀ-᰺᱀-ᱽ᳈-᳔᳒-ᾼι-ιῂ-ῌῐ-῜ῠ-Ῥ῰-ῼ⁰-⁹ⁿ-₉₏-₟₻-⃿ℂ-ℂℇ-ℇℊ-ℓℕ-ℕℙ-ℝℤ-ℤΩ-Ωℨ-ℨK-ℭℯ-ℹℼ-ℿⅅ-ⅉⅎ-ⅎ⅐-↏⑋-⒛⓪-⓿❶-➓⭚-ⳤⳫ-⳸⳽-⳽ⴀ-ⵯ⵱-ⷿⸯ-ⸯ々-〇〡-〯〱-〵〸-〼぀-゚ゝ-ゟァ-ヺー-㆏㆒-㆕ㆠ-ㆿ㇤-ㇿ㈟-㈩㉈-㉏㉑-㉟㊀-㊉㊱-㊿㐀-䶿一-꒏꓇-ꓽꔀ-ꘌꘐ-꙲ꙴ-꙽ꙿ-꛱ꜗ-ꜟꜢ-ꞈꞋ-ꠧ꠬-꠵꠺-ꡳ꡸-꣍꣐-ꣷꣻ-꤭ꤰ-꥞ꥠ-꧀꧎-꧝ꧠ-꩛ꩠ-ꩶꩺ-ꫝꫠ-ꫯꫲ-ꯪ꯬-퟿豈-ﬨשׁ-ﮱ﯂-ﴽ﵀-ﷻ﷾-️︚-︯﹬-﻾０-９Ａ-Ｚａ-ｚｦ-￟￾-𐃿𐄃-𐄶𐅀-𐅸𐆊-𐆏𐇽-𐎞𐎠-𐏏𐏑-𐡖𐡘-𐤞𐤠-𐤾𐥀-𐩏𐩙-𐩾𐪀-𐬸𐭀-𑁆𑁎-𑂺𑃂-𑄿𑅄-𑇄𑇉-𒑯𒑴-𜿿𝅥-𝅩𝅭-𝅲𝅻-𝆂𝆅-𝆋𝆪-𝆭𝉂-𝉄𝍗-𝛀𝛂-𝛚𝛜-𝛺𝛼-𝜔𝜖-𝜴𝜶-𝝎𝝐-𝝮𝝰-𝞈𝞊-𝞨𝞪-𝟂𝟄-𞻯🃠-🄏🝴-󠀀󠂀-󯿿]
+[0-9A-Za-zª-ª²-³µ-µ¹-º¼-¾À-ÖØ-öø-ˁˆ-ˑˠ-ˤˬ-ˬˮ-ˮ̀-ʹͶ-ͽΆ-ΆΈ-ϵϷ-ҁ҃-ՙա-և֑-ֽֿ-ֿׁ-ׂׄ-ׇׅ-ײؐ-ؚؠ-٩ٮ-ۓە-ۜ۟-۪ۨ-ۼۿ-ۿܐ-ߵߺ-࠭ࡀ-࡛ࢠ-ॣ०-९ॱ-ৱ৴-৹ਁ-૯ଁ-୯ୱ-௲ఁ-౾ಂ-൵ൺ-ෳก-ฺเ-๎๐-๙ກ-ༀ༘-༙༠-༳༵-༵༷-༹༷-༹༾-྄྆-ྼ࿆-࿆က-၉ၐ-ႝႠ-ჺჼ-፟፩-ᎏᎠ-Ᏼᐁ-ᙬᙯ-ᙿᚁ-ᚚᚠ-ᛪᛮ-᜴ᝀ-៓ៗ-ៗៜ-៹᠋-᠍᠐-᤻᥆-᧚ᨀ-ᨛᨠ-᪙ᪧ-ᪧᬀ-᭙᭫-᭳ᮀ-᯳ᰀ-᰷᱀-ᱽ᳐-᳔᳒-ᾼι-ιῂ-ῌῐ-Ίῠ-Ῥῲ-ῼ⁰-⁹ⁿ-₉ₐ-ₜ⃐-⃰ℂ-ℂℇ-ℇℊ-ℓℕ-ℕℙ-ℝℤ-ℤΩ-Ωℨ-ℨK-ℭℯ-ℹℼ-ℿⅅ-ⅉⅎ-ⅎ⅐-↉①-⒛⓪-⓿❶-➓Ⰰ-ⳤⳫ-ⳳ⳽-⳽ⴀ-ⵯ⵿-ⷿⸯ-ⸯ々-〇〡-〯〱-〵〸-〼ぁ-゚ゝ-ゟァ-ヺー-ㆎ㆒-㆕ㆠ-ㆺㇰ-ㇿ㈠-㈩㉈-㉏㉑-㉟㊀-㊉㊱-㊿㐀-䶵一-ꒌꓐ-ꓽꔀ-ꘌꘐ-꙲ꙴ-꙽ꙿ-꛱ꜗ-ꜟꜢ-ꞈꞋ-ꠧ꠰-꠵ꡀ-ꡳꢀ-꣄꣐-ꣷꣻ-꤭ꤰ-꥓ꥠ-꧀ꧏ-꧙ꨀ-꩙ꩠ-ꩶꩺ-ꫝꫠ-ꫯꫲ-ꯪ꯬-ퟻ豈-ﬨשׁ-ﮱﯓ-ﴽﵐ-ﷻ︀-️︠-︦ﹰ-ﻼ０-９Ａ-Ｚａ-ｚｦ-ￜ𐀀-𐃺𐄇-𐄳𐅀-𐅸𐆊-𐆊𐇽-𐎝𐎠-𐏏𐏑-𐡕𐡘-𐤛𐤠-𐤹𐦀-𐩇𐩠-𐩾𐬀-𐬵𐭀-𑁆𑁒-𑂺𑃐-𑄿𑆀-𑇄𑇐-𒑢𓀀-𛀁𝅥-𝅩𝅭-𝅲𝅻-𝆂𝆅-𝆋𝆪-𝆭𝉂-𝉄𝍠-𝛀𝛂-𝛚𝛜-𝛺𝛼-𝜔𝜖-𝜴𝜶-𝝎𝝐-𝝮𝝰-𝞈𝞊-𝞨𝞪-𝟂𝟄-𞺻🄀-🄊𠀀-𪘀󠄀-󠇯]