Generate the icu data binaries at compile time instead of checking in binaries

build_tools/filter_data_for_size.py

+"""Rewrites data so that it becomes smaller after compilation."""
+
+from __future__ import print_function
+
+import argparse
+import os
+import shutil
+
+def _mergetree(source_dir, dest_dir):
+    files = os.listdir(source_dir)
+    for f in files:
+        src = os.path.join(source_dir, f)
+        dest = os.path.join(dest_dir, f)
+        if os.path.isdir(src):
+            copytree(src, dest)
+        else:
+            shutil.copy2(src, dest)
+
+def copytree(source, dest):
+    """Like shutil.copytree but can handle that dest exists and
+    merge/overwrite the files there in that case."""
+    if not os.path.isdir(source):
+        shutil.copy2(source, dest)
+    elif os.path.isdir(dest):
+        _mergetree(source, dest)
+    else:
+        shutil.copytree(source, dest)
+
+def main():
+    parser = argparse.ArgumentParser(
+        description=('Generates an icudata.lst file to be compiled by icu.'))
+
+    parser.add_argument('--mode',
+                        required=True,
+                        choices=['clean-copy', 'strip-for-size'],
+                        help='Whether the files should be changed or not.')
+
+    parser.add_argument('--in-word-txt',
+                        required=True,
+                        help='The word.txt to filter')
+
+    parser.add_argument('--out-word-txt',
+                        required=True,
+                        help='The word.txt to filter')
+
+    parser.add_argument('--in-brkitr-root-txt',
+                        required=True,
+                        help='The brkitr/root.txt to filter')
+
+    parser.add_argument('--out-brkitr-root-txt',
+                        required=True,
+                        help='The brkitr/root.txt to filter')
+
+    parser.add_argument('--in-brkitr-ja-txt',
+                        required=True,
+                        help='The brkitr/ja.txt to filter')
+
+    parser.add_argument('--out-brkitr-ja-txt',
+                        required=True,
+                        help='The brkitr/ja.txt to filter')
+
+    parser.add_argument('--currency-keep-list',
+                        required=True,
+                        help="The file with currencies to keep")
+
+    parser.add_argument('--in-curr-dir',
+                        required=True,
+                        help="The currency data files")
+
+    parser.add_argument('--out-curr-dir',
+                        required=True,
+                        help="The filtered/copied currency data files")
+
+    parser.add_argument('--minimize-language-list',
+                        required=True,
+                        help=("Comma separated list of languages (locales) " +
+                              "to minimize the data for."))
+
+    parser.add_argument('--in-locales-dir',
+                        required=True,
+                        help="The locales data files")
+
+    parser.add_argument('--out-locales-dir',
+                        required=True,
+                        help="The filtered/copied locales data files")
+
+    parser.add_argument('--in-lang-dir',
+                        required=True,
+                        help="The lang data files")
+
+    parser.add_argument('--out-lang-dir',
+                        required=True,
+                        help="The filtered/copied lang data files")
+
+    parser.add_argument('--remove-data-already-existing-in-android',
+                        action="store_true",
+                        help=("Removes data in lang/region/... that " +
+                              "can be fetched from Android APIs"))
+
+    parser.add_argument('--in-zone-dir',
+                        required=True,
+                        help="The zone data files")
+
+    parser.add_argument('--out-zone-dir',
+                        required=True,
+                        help="The filtered/copied zone data files")
+
+    args = parser.parse_args()
+
+    if args.mode == 'strip-for-size':
+        fix_word_txt(args.in_word_txt, args.out_word_txt)
+        fix_brkitr_root_txt(args.in_brkitr_root_txt, args.out_brkitr_root_txt)
+        fix_brkitr_ja_txt(args.in_brkitr_ja_txt, args.out_brkitr_ja_txt)
+        fix_currencies(args.currency_keep_list,
+                       args.in_curr_dir,
+                       args.out_curr_dir)
+        minimize_data_for_locales(args.in_locales_dir,
+                                  args.out_locales_dir,
+                                  args.minimize_language_list)
+        remove_zone_example_cities(args.in_zone_dir, args.out_zone_dir)
+    else:
+        assert args.mode == 'clean-copy'
+        shutil.copyfile(args.in_word_txt, args.out_word_txt)
+        shutil.copyfile(args.in_brkitr_root_txt, args.out_brkitr_root_txt)
+        shutil.copyfile(args.in_brkitr_root_txt, args.out_brkitr_ja_txt)
+        copytree(args.in_curr_dir, args.out_curr_dir)
+        copytree(args.in_locales_dir, args.out_locales_dir)
+        copytree(args.in_zone_dir, args.out_zone_dir)
+
+    if (args.mode == 'strip-for-size' and
+        args.remove_data_already_existing_in_android):
+        strip_android_langs(args.in_lang_dir, args.out_lang_dir)
+    else:
+        assert args.mode == 'clean-copy'
+        copytree(args.in_lang_dir, args.out_lang_dir)
+
+def fix_word_txt(source, dest):
+    """Strip all references to dictionaryCJK and KanaKanji and
+       HangulSymbal."""
+    with open(source) as in_file:
+        with open(dest, "w") as out:
+            _process_word_txt(in_file, out)
+
+WORD_TXT_REPLACEMENTS = (
+    ("$ALetter-$dictionaryCJK", "$ALetter"),
+    ("$ComplexContext $dictionaryCJK", "$ComplexContext"),
+    ("$dictionaryCJK", None),
+    ("special handling for CJK", None),
+    ("$HangulSyllable $HangulSyllable", None),
+    ("$KanaKanji $KanaKanji", None),
+)
+
+def _process_word_txt(in_file, out):
+    for line in in_file:
+        for replacement_pattern, replace_with in WORD_TXT_REPLACEMENTS:
+            if replacement_pattern in line:
+                if replace_with is None:
+                    # Delete line.
+                    line = None
+                else:
+                    line = line.replace(replacement_pattern,
+                                        replace_with)
+                break
+        if line is not None:
+            out.write(line)
+
+def fix_brkitr_root_txt(source, dest):
+    """Strip all references to cjdict."""
+    with open(source) as in_file:
+        with open(dest, "w") as out:
+            trigger1 = False
+            for line in in_file:
+                if "cjdict.dict" in line:
+                    trigger1 = True
+                    continue
+                out.write(line)
+            assert trigger1
+
+def fix_brkitr_ja_txt(source, dest):
+    """Strip all references to line_ja.brk."""
+    with open(source) as in_file:
+        with open(dest, "w") as out:
+            trigger1 = False
+            for line in in_file:
+                if 'line:process(dependency){"line_ja.brk"}' in line:
+                    line = line.replace(
+                        'line:process(dependency){"line_ja.brk"}',
+                        'word:process(dependency){"word_ja.brk"}')
+                    trigger1 = True
+                out.write(line)
+            assert trigger1
+
+def _print_block(line, in_file, out):
+    indentation_count = 0
+    for char in line:
+        if char != " ":
+            break
+        indentation_count += 1
+    indentation = " " * indentation_count
+
+    if out is not None:
+        out.write(line)
+    if "}" in line:
+        return
+
+    line = in_file.next()
+    while not line.startswith(indentation + "}"):
+        if out is not None:
+            out.write(line)
+        line = in_file.next()
+    if out is not None:
+        out.write(line)
+
+def _skip_block(line, in_file):
+    _print_block(line, in_file, None)
+
+def _print_or_skip_block(line, in_file, out, keep):
+    if not keep:
+        out = None
+    _print_block(line, in_file, out)
+
+def _block_name(line):
+    assert "{" in line
+    return line.split("{")[0].lstrip()
+
+def _copy_start_of_file(in_file, out, locale_name):
+    line = in_file.next()
+    while not line.startswith("%s{" % locale_name):
+        out.write(line)
+        line = in_file.next()
+    out.write(line)
+
+def fix_currencies(keep_list_file, in_curr_dir, out_curr_dir):
+    """Remove currencies that are not globally important."""
+
+    with open(keep_list_file) as list_file:
+        currencies_to_keep = set(list_file.read().split())
+    currency_files = os.listdir(in_curr_dir)
+    for currency_file in currency_files:
+        in_file_name = os.path.join(in_curr_dir, currency_file)
+        out_file_name = os.path.join(out_curr_dir, currency_file)
+        if (not currency_file.endswith(".txt") or
+            currency_file == "supplementalData.txt"):
+            # For instance pool.res.
+            shutil.copyfile(in_file_name, out_file_name)
+            continue
+        locale_name = os.path.splitext(currency_file)[0]
+        with open(in_file_name) as in_file:
+            with open(out_file_name, "w") as out:
+                _process_currency_file(in_file, out, locale_name,
+                                       currencies_to_keep)
+
+def _process_currency_file(in_file, out, locale_name, currencies_to_keep):
+    try:
+        _copy_start_of_file(in_file, out, locale_name)
+        while True:
+            line = in_file.next()
+            if "{" in line:
+                block_name = _block_name(line)
+                if block_name in ("Currencies",
+                                  "Currencies%narrow",
+                                  "CurrencyPlurals"):
+                    # Keep only certain currencies.
+                    out.write(line)
+                    line = in_file.next()
+                    while not line.startswith("    }"):
+                        if "{" in line:
+                            currency_name = _block_name(line)
+                            keep_curr = currency_name in currencies_to_keep
+                            _print_or_skip_block(line, in_file, out, keep_curr)
+                        else:
+                            out.write(line)
+                        line = in_file.next()
+                    out.write(line)
+                else:
+                    keep = block_name in ('"%%ALIAS"',
+                                          "%%Parent",
+                                          "currencyMap",
+                                          "CurrencyMap",
+                                          "currencyMeta",
+                                          "CurrencyMeta",
+                                          "currencySpacing",
+                                          "CurrencySpacing",
+                                          "currencyUnitPatterns",
+                                          "CurrencyUnitPatterns",
+                                          "Version")
+                    _print_or_skip_block(line, in_file, out, keep)
+            else:
+                out.write(line)
+    except StopIteration:
+        pass
+
+def is_interesting_calendar(locale_name, calendar_name):
+    if calendar_name in ("generic", "gregorian"):
+        return True
+    interesting_calendar_map = {
+        "th": "buddhist",
+        "zh": "chinese",
+        "zh_Hant": "roc",
+        "ko": "dangi",
+        "am": "ethiopic",
+        "he": "hebrew",
+        "ar": "arabic",
+        "fa": "persian",
+        "ja": "japanese",
+    }
+    locale_parts = locale_name.split("_")
+    for i in range(len(locale_parts)):
+        base_locale = "_".join(locale_parts[:(i+1)])
+        if calendar_name == interesting_calendar_map.get(base_locale):
+            return True
+    return False
+
+def minimize_data_for_locales(locales_in_dir, locales_out_dir,
+                              languages_to_minimize_str):
+    languages_to_minimize = set(languages_to_minimize_str.split(","))
+
+    locale_files = os.listdir(locales_in_dir)
+    for locale_file in locale_files:
+        in_file_name = os.path.join(locales_in_dir, locale_file)
+        out_file_name = os.path.join(locales_out_dir, locale_file)
+        if not locale_file.endswith(".txt"):
+            # For instance pool.res.
+            shutil.copyfile(in_file_name, out_file_name)
+            continue
+        locale_name = os.path.splitext(locale_file)[0]
+
+        with open(in_file_name) as in_file:
+            with open(out_file_name, "w") as out:
+                _process_locale_file(in_file, out, locale_name,
+                                     languages_to_minimize)
+
+def _process_locale_file(in_file, out, locale_name, languages_to_minimize):
+    try:
+        _copy_start_of_file(in_file, out, locale_name)
+        while True: # Until StopIteration exception.
+            # Keep all blocks after "Version" but only a few
+            # blocks before that.
+            line = in_file.next()
+            if "{" in line:
+                block_name = _block_name(line)
+                if (not locale_name in languages_to_minimize and
+                    block_name == "calendar"):
+                    out.write(line)
+                    line = in_file.next()
+                    while not line.startswith("    }"):
+                        if "{" in line:
+                            calendar_name = _block_name(line)
+                            keep_cal = is_interesting_calendar(locale_name,
+                                                               calendar_name)
+                            _print_or_skip_block(line, in_file, out, keep_cal)
+                        else:
+                            out.write(line)
+                        line = in_file.next()
+                    out.write(line)
+                    continue
+
+                # Do not include '%%Parent' line on purpose.
+                keep = (not locale_name in languages_to_minimize or
+                        block_name in (
+                            '"%%ALIAS"',
+                            "AuxExemplarCharacters",
+                            "ExemplarCharacters",
+                            "LocaleScript",
+                            "layout",
+                            "Version"))
+                _print_or_skip_block(line, in_file, out, keep)
+            else:
+                out.write(line)
+    except StopIteration:
+        pass
+
+    # Note: patch_locale.sh also has code to strip only some calendars
+    # but since the calendar block is removed above that makes no
+    # difference.
+
+def strip_android_langs(in_lang_dir, out_lang_dir):
+    lang_files = os.listdir(in_lang_dir)
+    for lang_file in lang_files:
+        in_file_name = os.path.join(in_lang_dir, lang_file)
+        out_file_name = os.path.join(out_lang_dir, lang_file)
+        if not lang_file.endswith(".txt"):
+            # For instance pool.res
+            shutil.copyfile(in_file_name, out_file_name)
+            continue
+        locale_name = os.path.splitext(lang_file)[0]
+        with open(in_file_name) as in_file:
+            with open(out_file_name, "w") as out:
+                _process_lang_file(in_file, out, locale_name)
+
+def _process_lang_file(in_file, out, locale_name):
+    try:
+        _copy_start_of_file(in_file, out, locale_name)
+        while True: # Until StopIteration exception.
+            line = in_file.next()
+            if "{" in line:
+                block_name = _block_name(line)
+                if block_name in ("Languages", "Scripts"):
+                    out.write(line)
+                    parts_to_keep = {
+                        "Languages": ("zh{",),
+                        "Scripts": ("Hans{", "Hant{") }[block_name]
+                    line = in_file.next()
+                    while not line.startswith("    }"):
+                        if line.strip().startswith(parts_to_keep):
+                            out.write(line)
+                        line = in_file.next()
+                    out.write(line)
+                else:
+                    keep = True
+                    if block_name in ("Keys",
+                                      "LanguagesShort",
+                                      "Types",
+                                      "Variants",
+                                      "calendar",
+                                      "codePatterns",
+                                      "localeDisplayPattern"):
+                        # Delete the whole block.
+                        keep = False
+                    _print_or_skip_block(line, in_file, out, keep)
+            else:
+                out.write(line)
+    except StopIteration:
+        pass
+
+def remove_zone_example_cities(in_zone_dir, out_zone_dir):
+    zone_files = os.listdir(in_zone_dir)
+    for zone_file in zone_files:
+        locale_name = os.path.splitext(zone_file)[0]
+        with open(os.path.join(in_zone_dir, zone_file)) as in_file:
+            with open(os.path.join(out_zone_dir, zone_file), "w") as out:
+                # Strip everything from zoneStrings to the first line with
+                # a "meta:" name.
+                pos = 0 # 0 = start -> keep, 1 = mid -> delete, 2 = end -> keep
+                if locale_name == "root":
+                    pos = 2 # Keep everything in root.txt.
+                for line in in_file:
+                    if pos == 0:
+                        out.write(line)
+                        if line.startswith("    zoneStrings"):
+                            pos = 1
+                    if pos == 1 and line.startswith("        \"meta:"):
+                        pos = 2
+                    if pos == 2:
+                        out.write(line)
+
+if __name__ == '__main__':
+    main()
+