Quelle  make_unicode.py   Sprache: Python

#!/usr/bin/env python
# -*- coding: utf-8 -*-
# Based upon makeunicodedata.py
# (http://hg.python.org/cpython/file/c8192197d23d/Tools/unicode/makeunicodedata.py)
# written by Fredrik Lundh (fredrik@pythonware.com)
#
#    Copyright (C) 2011 Tom Schuster <evilpies@gmail.com>
#
#    This program is free software: you can redistribute it and/or modify
#    it under the terms of the GNU General Public License as published by
#    the Free Software Foundation, either version 3 of the License, or
#    (at your option) any later version.
#
#    This program is distributed in the hope that it will be useful,
#    but WITHOUT ANY WARRANTY; without even the implied warranty of
#    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
#    GNU General Public License for more details.
#
#    You should have received a copy of the GNU General Public License
#    along with this program.  If not, see <http://www.gnu.org/licenses/>.

import csv
import io
import os
import re
import sys
from contextlib import closing
from functools import partial
from itertools import chain, tee
from operator import is_not, itemgetter
from zipfile import ZipFile

if sys.version_info.major == 2:
    from itertools import ifilter as filter
    from itertools import imap as map
    from itertools import izip_longest as zip_longest

    from urllib2 import urlopen

    range = xrange
else:
    from itertools import zip_longest
    from urllib.request import urlopen


class codepoint_dict(dict):
    def name(self, code_point):
        (_, _, name, alias) = self[code_point]
        return "{}{}".format(name, (" (" + alias + ")" if alias else ""))

    def full_name(self, code_point):
        (_, _, name, alias) = self[code_point]
        return "U+{:04X} {}{}".format(
            code_point, name, (" (" + alias + ")" if alias else "")
        )


# ECMAScript 2016
# §11.2 White Space
whitespace = [
    # python doesn't support using control character names :(
    0x9,  # CHARACTER TABULATION
    0xB,  # LINE TABULATION
    0xC,  # FORM FEED
    ord("\N{SPACE}"),
    ord("\N{NO-BREAK SPACE}"),
    ord("\N{ZERO WIDTH NO-BREAK SPACE}"),  # also BOM
]

# §11.3 Line Terminators
line_terminator = [
    0xA,  # LINE FEED
    0xD,  # CARRIAGE RETURN
    ord("\N{LINE SEPARATOR}"),
    ord("\N{PARAGRAPH SEPARATOR}"),
]

# These are also part of IdentifierPart §11.6 Names and Keywords
compatibility_identifier_part = [
    ord("\N{ZERO WIDTH NON-JOINER}"),
    ord("\N{ZERO WIDTH JOINER}"),
]

FLAG_SPACE = 1 << 0
FLAG_UNICODE_ID_START = 1 << 1
FLAG_UNICODE_ID_CONTINUE_ONLY = 1 << 2

MAX_BMP = 0xFFFF

public_domain = """
/*
 * Any copyright is dedicated to the Public Domain.
 * http://creativecommons.org/licenses/publicdomain/
 */
"""

mpl_license = """\
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 4 -*-
 * vim: set ts=8 sts=4 et sw=4 tw=99:
 * This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
"""

warning_message = """\
/* Generated by make_unicode.py DO NOT MODIFY */
"""

unicode_version_message = """\
/* Unicode version: {0} */
"""


def read_unicode_data(unicode_data):
    """
    If you want to understand how this wonderful file format works checkout
      Unicode Standard Annex #44 - Unicode Character Database
      http://www.unicode.org/reports/tr44/
    """

    reader = csv.reader(unicode_data, delimiter=str(";"))

    while True:
        row = next(reader, None)
        if row is None:
            return
        name = row[1]

        # We need to expand the UAX #44 4.2.3 Code Point Range
        if name.startswith("<") and name.endswith("First>"):
            next_row = next(reader)

            for i in range(int(row[0], 16), int(next_row[0], 16) + 1):
                row[0] = i
                row[1] = name[1:-8]

                yield row
        else:
            row[0] = int(row[0], 16)
            yield row


def read_case_folding(case_folding):
    """
    File format is:
    <code>; <status>; <mapping>; # <name>
    """

    for line in case_folding:
        if line == "\n" or line.startswith("#"):
            continue
        row = line.split("; ")
        if row[1] in ["F", "T"]:
            continue
        assert row[1] in ["C", "S"], "expect either (C)ommon or (S)imple case foldings"
        code = int(row[0], 16)
        mapping = int(row[2], 16)
        yield (code, mapping)


def read_derived_core_properties(derived_core_properties):
    for line in derived_core_properties:
        if line == "\n" or line.startswith("#"):
            continue
        row = line.split("#")[0].split(";")
        char_range = row[0].strip()
        char_property = row[1].strip()
        if ".." not in char_range:
            yield (int(char_range, 16), char_property)
        else:
            [start, end] = char_range.split("..")
            for char in range(int(start, 16), int(end, 16) + 1):
                yield (char, char_property)


def read_special_casing(special_casing):
    # Format:
    # <code>; <lower>; <title>; <upper>; (<condition_list>;)? # <comment>
    for line in special_casing:
        if line == "\n" or line.startswith("#"):
            continue
        row = line.split("#")[0].split(";")
        code = int(row[0].strip(), 16)
        lower = row[1].strip()
        lower = [int(c, 16) for c in lower.split(" ")] if lower else []
        upper = row[3].strip()
        upper = [int(c, 16) for c in upper.split(" ")] if upper else []
        languages = []
        contexts = []
        condition = row[4].strip()
        if condition:
            for cond in condition.split(" "):
                if cond[0].islower():
                    languages.append(cond)
                else:
                    contexts.append(cond)
            pass
        yield (code, lower, upper, languages, contexts)


def int_ranges(ints):
    """Yields consecutive ranges (inclusive) from integer values."""
    (a, b) = tee(sorted(ints))
    start = next(b)
    for curr, succ in zip_longest(a, b):
        if curr + 1 != succ:
            yield (start, curr)
            start = succ


def utf16_encode(code):
    NonBMPMin = 0x10000
    LeadSurrogateMin = 0xD800
    TrailSurrogateMin = 0xDC00

    lead = (code - NonBMPMin) // 1024 + LeadSurrogateMin
    trail = ((code - NonBMPMin) % 1024) + TrailSurrogateMin

    return lead, trail


def make_non_bmp_convert_macro(out_file, name, convert_map, codepoint_table):
    # Find continuous range in convert_map.
    convert_list = []
    entry = None
    for code in sorted(convert_map.keys()):
        lead, trail = utf16_encode(code)
        converted = convert_map[code]
        diff = converted - code

        if (
            entry
            and code == entry["code"] + entry["length"]
            and diff == entry["diff"]
            and lead == entry["lead"]
        ):
            entry["length"] += 1
            continue

        entry = {
            "code": code,
            "diff": diff,
            "length": 1,
            "lead": lead,
            "trail": trail,
        }
        convert_list.append(entry)

    # Generate macro call for each range.
    lines = []
    comment = []
    for entry in convert_list:
        from_code = entry["code"]
        to_code = entry["code"] + entry["length"] - 1
        diff = entry["diff"]

        lead = entry["lead"]
        from_trail = entry["trail"]
        to_trail = entry["trail"] + entry["length"] - 1

        lines.append(
            " MACRO(0x{:x}, 0x{:x}, 0x{:x}, 0x{:x}, 0x{:x}, {:d})".format(
                from_code, to_code, lead, from_trail, to_trail, diff
            )
        )
        comment.append(
            "// {} .. {}".format(
                codepoint_table.full_name(from_code), codepoint_table.full_name(to_code)
            )
        )

    out_file.write("\n".join(comment))
    out_file.write("\n")
    out_file.write("#define FOR_EACH_NON_BMP_{}(MACRO) \\\n".format(name))
    out_file.write(" \\\n".join(lines))
    out_file.write("\n")


def process_derived_core_properties(derived_core_properties):
    id_start = set()
    id_continue = set()

    for char, prop in read_derived_core_properties(derived_core_properties):
        if prop == "ID_Start":
            id_start.add(char)
        if prop == "ID_Continue":
            id_continue.add(char)

    return (id_start, id_continue)


def process_unicode_data(unicode_data, derived_core_properties):
    dummy = (0, 0, 0)
    table = [dummy]
    cache = {dummy: 0}
    index = [0] * (MAX_BMP + 1)

    codepoint_table = codepoint_dict()
    test_space_table = []

    non_bmp_lower_map = {}
    non_bmp_upper_map = {}
    non_bmp_id_start_set = {}
    non_bmp_id_cont_set = {}
    non_bmp_space_set = {}

    (id_start, id_continue) = process_derived_core_properties(derived_core_properties)

    for row in read_unicode_data(unicode_data):
        code = row[0]
        name = row[1]
        category = row[2]
        alias = row[-5]
        uppercase = row[-3]
        lowercase = row[-2]

        if uppercase:
            upper = int(uppercase, 16)
        else:
            upper = code

        if lowercase:
            lower = int(lowercase, 16)
        else:
            lower = code

        codepoint_table[code] = (upper, lower, name, alias)

        if code > MAX_BMP:
            if code != lower:
                non_bmp_lower_map[code] = lower
            if code != upper:
                non_bmp_upper_map[code] = upper
            if category == "Zs":
                non_bmp_space_set[code] = 1
                test_space_table.append(code)
            if code in id_start:
                non_bmp_id_start_set[code] = 1
            if code in id_continue:
                non_bmp_id_cont_set[code] = 1
            continue

        assert lower <= MAX_BMP and upper <= MAX_BMP

        flags = 0

        # we combine whitespace and lineterminators because in pratice we don't need them separated
        if category == "Zs" or code in whitespace or code in line_terminator:
            flags |= FLAG_SPACE
            test_space_table.append(code)

        # §11.6 (IdentifierStart)
        if code in id_start:
            flags |= FLAG_UNICODE_ID_START

        # §11.6 (IdentifierPart)
        elif code in id_continue or code in compatibility_identifier_part:
            flags |= FLAG_UNICODE_ID_CONTINUE_ONLY

        up_d = upper - code
        low_d = lower - code

        assert up_d > -65535 and up_d < 65535
        assert low_d > -65535 and low_d < 65535

        upper = up_d & 0xFFFF
        lower = low_d & 0xFFFF

        item = (upper, lower, flags)

        i = cache.get(item)
        if i is None:
            assert item not in table
            cache[item] = i = len(table)
            table.append(item)
        index[code] = i

    return (
        table,
        index,
        non_bmp_lower_map,
        non_bmp_upper_map,
        non_bmp_space_set,
        non_bmp_id_start_set,
        non_bmp_id_cont_set,
        codepoint_table,
        test_space_table,
    )


def process_case_folding(case_folding):
    folding_map = {}
    rev_folding_map = {}
    folding_dummy = (0,)
    folding_table = [folding_dummy]
    folding_cache = {folding_dummy: 0}
    folding_index = [0] * (MAX_BMP + 1)

    folding_tests = []
    folding_codes = set()

    for code, mapping in read_case_folding(case_folding):
        folding_map[code] = mapping

        if mapping not in rev_folding_map:
            rev_folding_map[mapping] = [code]
        else:
            rev_folding_map[mapping].append(code)

        folding_codes.add(code)
        folding_codes.add(mapping)

    for code in sorted(folding_codes):
        if code in folding_map:
            folding = folding_map[code]
        else:
            folding = code

        if code in rev_folding_map:
            rev_folding = rev_folding_map[code]
        elif folding in rev_folding_map:
            rev_folding = [c for c in rev_folding_map[folding] if c != code]
        else:
            rev_folding = []

        if folding != code or len(rev_folding):
            item = [code]
            if folding != code:
                item.append(folding)
            folding_tests.append(item + rev_folding)

        if code > MAX_BMP:
            continue

        folding_d = folding - code

        assert folding_d > -65535 and folding_d < 65535

        folding = folding_d & 0xFFFF

        item = (folding,)

        i = folding_cache.get(item)
        if i is None:
            assert item not in folding_table
            folding_cache[item] = i = len(folding_table)
            folding_table.append(item)
        folding_index[code] = i
    return (folding_table, folding_index, folding_tests)


def process_special_casing(special_casing, table, index):
    # Unconditional special casing.
    unconditional_tolower = {}
    unconditional_toupper = {}

    # Conditional special casing, language independent.
    conditional_tolower = {}
    conditional_toupper = {}

    # Conditional special casing, language dependent.
    lang_conditional_tolower = {}
    lang_conditional_toupper = {}

    def caseInfo(code):
        (upper, lower, flags) = table[index[code]]
        return ((code + lower) & 0xFFFF, (code + upper) & 0xFFFF)

    for code, lower, upper, languages, contexts in read_special_casing(special_casing):
        assert code <= MAX_BMP, "Unexpected character outside of BMP: %s" % code
        assert len(languages) <= 1, "Expected zero or one language ids: %s" % languages
        assert len(contexts) <= 1, (
            "Expected zero or one casing contexts: %s" % languages
        )

        (default_lower, default_upper) = caseInfo(code)
        special_lower = len(lower) != 1 or lower[0] != default_lower
        special_upper = len(upper) != 1 or upper[0] != default_upper

        # Invariant: If |code| has casing per UnicodeData.txt, then it also has
        # casing rules in SpecialCasing.txt.
        assert code == default_lower or len(lower) != 1 or code != lower[0]
        assert code == default_upper or len(upper) != 1 or code != upper[0]

        language = languages[0] if languages else None
        context = contexts[0] if contexts else None

        if not language and not context:
            if special_lower:
                unconditional_tolower[code] = lower
            if special_upper:
                unconditional_toupper[code] = upper
        elif not language and context:
            if special_lower:
                conditional_tolower[code] = (lower, context)
            if special_upper:
                conditional_toupper[code] = (upper, context)
        else:
            if language not in lang_conditional_tolower:
                lang_conditional_tolower[language] = {}
                lang_conditional_toupper[language] = {}
            if special_lower:
                lang_conditional_tolower[language][code] = (lower, context)
            if special_upper:
                lang_conditional_toupper[language][code] = (upper, context)

    # Certain special casing rules are inlined in jsstr.cpp, ensure these cases
    # still match the current SpecialCasing.txt file.
    def lowerCase(code):
        (lower, _) = caseInfo(code)
        return lower

    def upperCase(code):
        (_, upper) = caseInfo(code)
        return upper

    def ascii(char_dict):
        return (ch for ch in char_dict.keys() if ch <= 0x7F)

    def latin1(char_dict):
        return (ch for ch in char_dict.keys() if ch <= 0xFF)

    def is_empty(iterable):
        return not any(True for _ in iterable)

    def is_equals(iter1, iter2):
        return all(x == y for (x, y) in zip_longest(iter1, iter2))

    # Ensure no ASCII characters have special case mappings.
    assert is_empty(ascii(unconditional_tolower))
    assert is_empty(ascii(unconditional_toupper))
    assert is_empty(ascii(conditional_tolower))
    assert is_empty(ascii(conditional_toupper))

    # Ensure no Latin1 characters have special lower case mappings.
    assert is_empty(latin1(unconditional_tolower))
    assert is_empty(latin1(conditional_tolower))

    # Ensure no Latin1 characters have conditional special upper case mappings.
    assert is_empty(latin1(conditional_toupper))

    # Ensure U+00DF is the only Latin1 character with a special upper case mapping.
    assert is_equals([0x00DF], latin1(unconditional_toupper))

    # Ensure U+0130 is the only character with a special lower case mapping.
    assert is_equals([0x0130], unconditional_tolower)

    # Ensure no characters have language independent conditional upper case mappings.
    assert is_empty(conditional_toupper)

    # Ensure U+03A3 is the only character with language independent conditional lower case mapping.
    assert is_equals([0x03A3], conditional_tolower)

    # Verify U+0130 and U+03A3 have simple lower case mappings.
    assert all(ch != lowerCase(ch) for ch in [0x0130, 0x03A3])

    # Ensure Azeri, Lithuanian, and Turkish are the only languages with conditional case mappings.
    assert is_equals(["az", "lt", "tr"], sorted(lang_conditional_tolower.keys()))
    assert is_equals(["az", "lt", "tr"], sorted(lang_conditional_toupper.keys()))

    # Maximum case mapping length is three characters.
    assert (
        max(
            map(
                len,
                chain(
                    unconditional_tolower.values(),
                    unconditional_toupper.values(),
                    map(itemgetter(0), conditional_tolower.values()),
                    map(itemgetter(0), conditional_toupper.values()),
                    map(
                        itemgetter(0),
                        chain.from_iterable(
                            d.values() for d in lang_conditional_tolower.values()
                        ),
                    ),
                    map(
                        itemgetter(0),
                        chain.from_iterable(
                            d.values() for d in lang_conditional_toupper.values()
                        ),
                    ),
                ),
            )
        )
        <= 3
    )

    # Ensure all case mapping contexts are known (see Unicode 9.0, §3.13 Default Case Algorithms).
    assert set(
        [
            "After_I",
            "After_Soft_Dotted",
            "Final_Sigma",
            "More_Above",
            "Not_Before_Dot",
        ]
    ).issuperset(
        set(
            filter(
                partial(is_not, None),
                chain(
                    map(itemgetter(1), conditional_tolower.values()),
                    map(itemgetter(1), conditional_toupper.values()),
                    map(
                        itemgetter(1),
                        chain.from_iterable(
                            d.values() for d in lang_conditional_tolower.values()
                        ),
                    ),
                    map(
                        itemgetter(1),
                        chain.from_iterable(
                            d.values() for d in lang_conditional_toupper.values()
                        ),
                    ),
                ),
            )
        )
    )

    # Special casing for U+00DF (LATIN SMALL LETTER SHARP S).
    assert upperCase(0x00DF) == 0x00DF and unconditional_toupper[0x00DF] == [
        0x0053,
        0x0053,
    ]

    # Special casing for U+0130 (LATIN CAPITAL LETTER I WITH DOT ABOVE).
    assert unconditional_tolower[0x0130] == [0x0069, 0x0307]

    # Special casing for U+03A3 (GREEK CAPITAL LETTER SIGMA).
    assert lowerCase(0x03A3) == 0x03C3 and conditional_tolower[0x03A3] == (
        [0x03C2],
        "Final_Sigma",
    )

    return (unconditional_tolower, unconditional_toupper)


def make_non_bmp_file(version, non_bmp_lower_map, non_bmp_upper_map, codepoint_table):
    file_name = "UnicodeNonBMP.h"
    with io.open(file_name, mode="w", encoding="utf-8") as non_bmp_file:
        non_bmp_file.write(mpl_license)
        non_bmp_file.write("\n")
        non_bmp_file.write(warning_message)
        non_bmp_file.write(unicode_version_message.format(version))
        non_bmp_file.write(
            """
#ifndef util_UnicodeNonBMP_h
#define util_UnicodeNonBMP_h

// |MACRO| receives the following arguments
//   MACRO(FROM, TO, LEAD, TRAIL_FROM, TRAIL_TO, DIFF)
//     FROM:       code point where the range starts
//     TO:         code point where the range ends
//     LEAD:       common lead surrogate of FROM and TO
//     TRAIL_FROM: trail surrogate of FROM
//     TRAIL_FROM: trail surrogate of TO
//     DIFF:       the difference between the code point in the range and
//                 converted code point

"""
        )

        make_non_bmp_convert_macro(
            non_bmp_file, "LOWERCASE", non_bmp_lower_map, codepoint_table
        )
        non_bmp_file.write("\n")
        make_non_bmp_convert_macro(
            non_bmp_file, "UPPERCASE", non_bmp_upper_map, codepoint_table
        )

        non_bmp_file.write(
            """
#endif /* util_UnicodeNonBMP_h */
"""
        )


def write_special_casing_methods(unconditional_toupper, codepoint_table, println):
    def hexlit(n):
        """Returns C++ hex-literal for |n|."""
        return "0x{:04X}".format(n)

    def describe_range(ranges, depth):
        indent = depth * " "
        for start, end in ranges:
            if start == end:
                println(indent, "// {}".format(codepoint_table.full_name(start)))
            else:
                println(
                    indent,
                    "// {} .. {}".format(
                        codepoint_table.full_name(start), codepoint_table.full_name(end)
                    ),
                )

    def out_range(start, end):
        """Tests if the input character isn't a member of the set {x | start <= x <= end}."""
        if start == end:
            return "ch != {}".format(hexlit(start))
        return "ch < {} || ch > {}".format(hexlit(start), hexlit(end))

    def in_range(start, end, parenthesize=False):
        """Tests if the input character is in the set {x | start <= x <= end}."""
        if start == end:
            return "ch == {}".format(hexlit(start))
        (left, right) = ("(", ")") if parenthesize else ("", "")
        return "{}ch >= {} && ch <= {}{}".format(
            left, hexlit(start), hexlit(end), right
        )

    def in_any_range(ranges, spaces):
        """Tests if the input character is included in any of the given ranges."""
        lines = [[]]
        for start, end in ranges:
            expr = in_range(start, end, parenthesize=True)
            line = " || ".join(lines[-1] + [expr])
            if len(line) < (100 - len(spaces) - len(" ||")):
                lines[-1].append(expr)
            else:
                lines.append([expr])
        return " ||\n{}".format(spaces).join(" || ".join(t) for t in lines)

    def write_range_accept(parent_list, child_list, depth):
        """Accepts the input character if it matches any code unit in |child_list|."""
        (min_parent, max_parent) = (parent_list[0], parent_list[-1])
        (min_child, max_child) = (child_list[0], child_list[-1])
        assert min_child >= min_parent
        assert max_child <= max_parent
        indent = depth * " "

        child_ranges = list(int_ranges(child_list))
        has_successor = max_child != max_parent

        # If |child_list| is a contiguous list of code units, emit a simple
        # range check: |min_child <= input <= max_child|.
        if len(child_ranges) == 1:
            describe_range(child_ranges, depth)
            if has_successor:
                println(indent, "if (ch <= {}) {{".format(hexlit(max_child)))
                println(indent, " return ch >= {};".format(hexlit(min_child)))
                println(indent, "}")
            else:
                println(indent, "return {};".format(in_range(min_child, max_child)))
            return

        # Otherwise create a disjunction over the subranges in |child_ranges|.
        if not has_successor:
            spaces = indent + len("return ") * " "
        else:
            spaces = indent + len(" return ") * " "
        range_test_expr = in_any_range(child_ranges, spaces)

        if min_child != min_parent:
            println(indent, "if (ch < {}) {{".format(hexlit(min_child)))
            println(indent, " return false;")
            println(indent, "}")

        # If there's no successor block, we can omit the |input <= max_child| check,
        # because it was already checked when we emitted the parent range test.
        if not has_successor:
            describe_range(child_ranges, depth)
            println(indent, "return {};".format(range_test_expr))
        else:
            println(indent, "if (ch <= {}) {{".format(hexlit(max_child)))
            describe_range(child_ranges, depth + 1)
            println(indent, " return {};".format(range_test_expr))
            println(indent, "}")

    def write_ChangesWhenUpperCasedSpecialCasing():
        """Checks if the input has a special upper case mapping."""
        println("bool")
        println("js::unicode::ChangesWhenUpperCasedSpecialCasing(char16_t ch)")
        println("{")

        assert unconditional_toupper, "|unconditional_toupper| is not empty"

        # Sorted list of code units with special upper case mappings.
        code_list = sorted(unconditional_toupper.keys())

        # Fail-fast if the input character isn't a special casing character.
        println(" if ({}) {{".format(out_range(code_list[0], code_list[-1])))
        println(" return false;")
        println(" }")

        for i in range(0, 16):
            # Check if the input characters is in the range:
            # |start_point <= input < end_point|.
            start_point = i << 12
            end_point = (i + 1) << 12
            matches = [cu for cu in code_list if start_point <= cu < end_point]

            # Skip empty ranges.
            if not matches:
                continue

            # If |matches| consists of only a few characters, directly check
            # the input against the characters in |matches|.
            if len(matches) <= 8:
                write_range_accept(code_list, matches, depth=1)
                continue

            # Otherwise split into further subranges.

            # Only enter the if-block if the input is less-or-equals to the
            # largest value in the current range.
            is_last_block = matches[-1] == code_list[-1]
            if not is_last_block:
                println(" if (ch <= {}) {{".format(hexlit(matches[-1])))
            else:
                println(" if (ch < {}) {{".format(hexlit(matches[0])))
                println(" return false;")
                println(" }")

            for j in range(0, 16):
                inner_start = start_point + (j << 8)
                inner_end = start_point + ((j + 1) << 8)
                inner_matches = [cu for cu in matches if inner_start <= cu < inner_end]

                if inner_matches:
                    d = 1 if is_last_block else 2
                    write_range_accept(matches, inner_matches, depth=d)

            if not is_last_block:
                println(" }")

        println("}")

    def write_LengthUpperCaseSpecialCasing():
        """Slow case: Special casing character was found, returns its mapping length."""
        println("size_t")
        println("js::unicode::LengthUpperCaseSpecialCasing(char16_t ch)")
        println("{")

        println(" switch(ch) {")
        for code, converted in sorted(unconditional_toupper.items(), key=itemgetter(0)):
            println(
                " case {}: return {}; // {}".format(
                    hexlit(code), len(converted), codepoint_table.name(code)
                )
            )
        println(" }")
        println("")
        println(' MOZ_ASSERT_UNREACHABLE("Bad character input.");')
        println(" return 0;")

        println("}")

    def write_AppendUpperCaseSpecialCasing():
        """Slow case: Special casing character was found, append its mapping characters."""
        println("void")
        println(
            "js::unicode::AppendUpperCaseSpecialCasing(char16_t ch, char16_t* elements, size_t* index)"  # NOQA: E501
        )
        println("{")

        println(" switch(ch) {")
        for code, converted in sorted(unconditional_toupper.items(), key=itemgetter(0)):
            println(
                " case {}: // {}".format(hexlit(code), codepoint_table.name(code))
            )
            for ch in converted:
                println(
                    " elements[(*index)++] = {}; // {}".format(
                        hexlit(ch), codepoint_table.name(ch)
                    )
                )
            println(" return;")
        println(" }")
        println("")
        println(' MOZ_ASSERT_UNREACHABLE("Bad character input.");')

        println("}")

    write_ChangesWhenUpperCasedSpecialCasing()
    println("")
    write_LengthUpperCaseSpecialCasing()
    println("")
    write_AppendUpperCaseSpecialCasing()


def write_ascii_lookup_tables(table, index, write, println):
    def is_id_compat(code):
        return code == ord("\N{DOLLAR SIGN}") or code == ord("\N{LOW LINE}")

    def is_id_start(code):
        (upper, lower, flags) = table[index[code]]
        return (flags & FLAG_UNICODE_ID_START) or is_id_compat(code)

    def is_id_continue(code):
        (upper, lower, flags) = table[index[code]]
        return (flags & FLAG_UNICODE_ID_CONTINUE_ONLY) or is_id_start(code)

    def is_space(code):
        (upper, lower, flags) = table[index[code]]
        return flags & FLAG_SPACE

    def write_entries(name, predicate):
        println("const bool unicode::{}[] = {{".format(name))
        header = "".join("{0: <6}".format(x) for x in range(0, 10)).rstrip()
        println("/* {} */".format(header))
        for i in range(0, 13):
            write("/* {0: >2} */".format(i))
            for j in range(0, 10):
                code = i * 10 + j
                if code <= 0x7F:
                    write(" {},".format("true" if predicate(code) else "____"))
            println("")
        println("};")

    println("")
    println("#define ____ false")

    println(
        """
/*
 * Identifier start chars:
 * -      36:    $
 * -  65..90: A..Z
 * -      95:    _
 * - 97..122: a..z
 */"""
    )
    write_entries("js_isidstart", is_id_start)

    println(
        """
/*
 * Identifier chars:
 * -      36:    $
 * -  48..57: 0..9
 * -  65..90: A..Z
 * -      95:    _
 * - 97..122: a..z
 */"""
    )
    write_entries("js_isident", is_id_continue)

    println(
        """
/* Whitespace chars: '\\t', '\\n', '\\v', '\\f', '\\r', ' '. */"""
    )
    write_entries("js_isspace", is_space)

    println("")
    println("#undef ____")


def write_latin1_lookup_tables(table, index, write, println):
    def case_info(code):
        assert 0 <= code and code <= MAX_BMP
        (upper, lower, flags) = table[index[code]]
        return ((code + upper) & 0xFFFF, (code + lower) & 0xFFFF, flags)

    def toLowerCase(code):
        (_, lower, _) = case_info(code)
        assert lower <= 0xFF, "lower-case of Latin-1 is always Latin-1"
        return lower

    def write_entries(name, mapper):
        println("const JS::Latin1Char unicode::{}[] = {{".format(name))
        header = "".join("{0: <6}".format(x) for x in range(0, 16)).rstrip()
        println("/* {} */".format(header))
        for i in range(0, 16):
            write("/* {0: >2} */".format(i))
            for j in range(0, 16):
                code = i * 16 + j
                if code <= 0xFF:
                    write(" 0x{:02X},".format(mapper(code)))
            println("")
        println("};")

    println("")
    write_entries("latin1ToLowerCaseTable", toLowerCase)


def make_bmp_mapping_test(
    version, codepoint_table, unconditional_tolower, unconditional_toupper
):
    def unicodeEsc(n):
        return "\\u{:04X}".format(n)

    file_name = "../tests/non262/String/string-upper-lower-mapping.js"
    with io.open(file_name, mode="w", encoding="utf-8") as output:
        write = partial(print, file=output, sep="", end="")
        println = partial(print, file=output, sep="", end="\n")

        write(warning_message)
        write(unicode_version_message.format(version))
        write(public_domain)
        println("var mapping = [")
        for code in range(0, MAX_BMP + 1):
            entry = codepoint_table.get(code)

            if entry:
                (upper, lower, _, _) = entry
                upper = (
                    unconditional_toupper[code]
                    if code in unconditional_toupper
                    else [upper]
                )
                lower = (
                    unconditional_tolower[code]
                    if code in unconditional_tolower
                    else [lower]
                )
                println(
                    ' ["{}", "{}"], /* {} */'.format(
                        "".join(map(unicodeEsc, upper)),
                        "".join(map(unicodeEsc, lower)),
                        codepoint_table.name(code),
                    )
                )
            else:
                println(' ["{0}", "{0}"],'.format(unicodeEsc(code)))
        println("];")
        write(
            """
assertEq(mapping.length, 0x10000);
for (var i = 0; i <= 0xffff; i++) {
    var char = String.fromCharCode(i);
    var info = mapping[i];

    assertEq(char.toUpperCase(), info[0]);
    assertEq(char.toLowerCase(), info[1]);
}

if (typeof reportCompare === "function")
    reportCompare(true, true);
"""
        )


def make_non_bmp_mapping_test(
    version, non_bmp_upper_map, non_bmp_lower_map, codepoint_table
):
    file_name = "../tests/non262/String/string-code-point-upper-lower-mapping.js"
    with io.open(file_name, mode="w", encoding="utf-8") as test_non_bmp_mapping:
        test_non_bmp_mapping.write(warning_message)
        test_non_bmp_mapping.write(unicode_version_message.format(version))
        test_non_bmp_mapping.write(public_domain)

        for code in sorted(non_bmp_upper_map.keys()):
            test_non_bmp_mapping.write(
                """\
assertEq(String.fromCodePoint(0x{:04X}).toUpperCase().codePointAt(0), 0x{:04X}); // {}, {}
""".format(
                    code,
                    non_bmp_upper_map[code],
                    codepoint_table.name(code),
                    codepoint_table.name(non_bmp_upper_map[code]),
                )
            )

        for code in sorted(non_bmp_lower_map.keys()):
            test_non_bmp_mapping.write(
                """\
assertEq(String.fromCodePoint(0x{:04X}).toLowerCase().codePointAt(0), 0x{:04X}); // {}, {}
""".format(
                    code,
                    non_bmp_lower_map[code],
                    codepoint_table.name(code),
                    codepoint_table.name(non_bmp_lower_map[code]),
                )
            )

        test_non_bmp_mapping.write(
            """
if (typeof reportCompare === "function")
    reportCompare(true, true);
"""
        )


def make_space_test(version, test_space_table, codepoint_table):
    def hex_and_name(c):
        return " 0x{:04X} /* {} */".format(c, codepoint_table.name(c))

    file_name = "../tests/non262/String/string-space-trim.js"
    with io.open(file_name, mode="w", encoding="utf-8") as test_space:
        test_space.write(warning_message)
        test_space.write(unicode_version_message.format(version))
        test_space.write(public_domain)
        test_space.write("var onlySpace = String.fromCharCode(\n")
        test_space.write(",\n".join(map(hex_and_name, test_space_table)))
        test_space.write("\n);\n")
        test_space.write(
            """
assertEq(onlySpace.trim(), "");
assertEq((onlySpace + 'aaaa').trim(), 'aaaa');
assertEq(('aaaa' + onlySpace).trim(), 'aaaa');
assertEq((onlySpace + 'aaaa' + onlySpace).trim(), 'aaaa');

if (typeof reportCompare === "function")
    reportCompare(true, true);
"""
        )


def make_regexp_space_test(version, test_space_table, codepoint_table):
    def hex_and_name(c):
        return " 0x{:04X} /* {} */".format(c, codepoint_table.name(c))

    file_name = "../tests/non262/RegExp/character-class-escape-s.js"
    with io.open(file_name, mode="w", encoding="utf-8") as test_space:
        test_space.write(warning_message)
        test_space.write(unicode_version_message.format(version))
        test_space.write(public_domain)
        test_space.write("var onlySpace = String.fromCodePoint(\n")
        test_space.write(",\n".join(map(hex_and_name, test_space_table)))
        test_space.write("\n);\n")
        test_space.write(
            r"""
assertEq(/^\s+$/.exec(onlySpace) !== null, true);
assertEq(/^[\s]+$/.exec(onlySpace) !== null, true);
assertEq(/^[^\s]+$/.exec(onlySpace) === null, true);

assertEq(/^\S+$/.exec(onlySpace) === null, true);
assertEq(/^[\S]+$/.exec(onlySpace) === null, true);
assertEq(/^[^\S]+$/.exec(onlySpace) !== null, true);

// Also test with Unicode RegExps.
assertEq(/^\s+$/u.exec(onlySpace) !== null, true);
assertEq(/^[\s]+$/u.exec(onlySpace) !== null, true);
assertEq(/^[^\s]+$/u.exec(onlySpace) === null, true);

assertEq(/^\S+$/u.exec(onlySpace) === null, true);
assertEq(/^[\S]+$/u.exec(onlySpace) === null, true);
assertEq(/^[^\S]+$/u.exec(onlySpace) !== null, true);

if (typeof reportCompare === "function")
    reportCompare(true, true);
"""
        )


def make_icase_test(version, folding_tests, codepoint_table):
    def char_hex(c):
        return "0x{:04X}".format(c)

    file_name = "../tests/non262/RegExp/unicode-ignoreCase.js"
    with io.open(file_name, mode="w", encoding="utf-8") as test_icase:
        test_icase.write("// |reftest| skip-if(!this.hasOwnProperty('Intl'))\n\n")
        test_icase.write(warning_message)
        test_icase.write(unicode_version_message.format(version))
        test_icase.write(public_domain)
        test_icase.write(
            """
var BUGNUMBER = 1135377;
var summary = "Implement RegExp unicode flag -- ignoreCase flag.";

print(BUGNUMBER + ": " + summary);

function test(code, ...equivs) {
  var codeRe = new RegExp(String.fromCodePoint(code) + "+", "iu");
  var ans = String.fromCodePoint(code) + equivs.map(c => String.fromCodePoint(c)).join("");
  assertEqArray(codeRe.exec("<" + ans + ">"), [ans]);
  codeRe = new RegExp("[" + String.fromCodePoint(code) + "]+", "iu");
  assertEqArray(codeRe.exec("<" + ans + ">"), [ans]);
}
"""
        )
        for args in folding_tests:
            test_icase.write(
                "test({}); // {}\n".format(
                    ", ".join(map(char_hex, args)),
                    ", ".join(map(codepoint_table.name, args)),
                )
            )
        test_icase.write(
            """
if (typeof reportCompare === "function")
    reportCompare(true, true);
"""
        )


def make_unicode_file(
    version,
    table,
    index,
    folding_table,
    folding_index,
    non_bmp_space_set,
    non_bmp_id_start_set,
    non_bmp_id_cont_set,
    unconditional_toupper,
    codepoint_table,
):
    index1, index2, shift = splitbins(index)

    # Don't forget to update CharInfo in Unicode.h if you need to change this
    assert shift == 6

    folding_index1, folding_index2, folding_shift = splitbins(folding_index)

    # Don't forget to update CaseFoldInfo in Unicode.h if you need to change this
    assert folding_shift == 5

    # verify correctness
    for char in index:
        test = table[index[char]]

        idx = index1[char >> shift]
        idx = index2[(idx << shift) + (char & ((1 << shift) - 1))]

        assert test == table[idx]

    # verify correctness
    for char in folding_index:
        test = folding_table[folding_index[char]]

        idx = folding_index1[char >> folding_shift]
        idx = folding_index2[
            (idx << folding_shift) + (char & ((1 << folding_shift) - 1))
        ]

        assert test == folding_table[idx]

    comment = """
/*
 * So how does indexing work?
 * First let's have a look at a char16_t, 16-bits:
 *              [................]
 * Step 1:
 *  Extracting the upper 11 bits from the char16_t.
 *   upper = char >>  5 ([***********.....])
 * Step 2:
 *  Using these bits to get an reduced index from index1.
 *   index = index1[upper]
 * Step 3:
 *  Combining the index and the bottom 5 bits of the original char16_t.
 *   real_index = index2[(index << 5) + (char & ((1 << 5) - 1))] ([...********+++++])
 *
 * The advantage here is that the biggest number in index1 doesn't need 10 bits,
 * but 7 and we save some memory.
 *
 * Step 4:
 *  Get the character informations by looking up real_index in js_charinfo.
 *
 * Pseudocode of generation:
 *
 * let table be the mapping of char16_t => js_charinfo_index
 * let index1 be an empty array
 * let index2 be an empty array
 * let cache be a hash map
 *
 * while shift is less then maximal amount you can shift 0xffff before it's 0
 *  let chunks be table split in chunks of size 2**shift
 *
 *  for every chunk in chunks
 *   if chunk is in cache
 *    let index be cache[chunk]
 *   else
 *    let index be the max key of index2 + 1
 *    for element in chunk
 *     push element to index2
 *    put index as chunk in cache
 *
 *   push index >> shift to index1
 *
 *  increase shift
 *  stop if you found the best shift
 */
"""

    def dump(data, name, println):
        println("const uint8_t unicode::{}[] = {{".format(name))

        line = pad = " " * 4
        lines = []
        for entry in data:
            assert entry < 256
            s = str(entry)
            s = s.rjust(3)

            if len(line + s) + 5 > 99:
                lines.append(line.rstrip())
                line = pad + s + ", "
            else:
                line = line + s + ", "
        lines.append(line.rstrip())

        println("\n".join(lines))
        println("};")

    def write_table(data_type, name, tbl, idx1_name, idx1, idx2_name, idx2, println):
        println("const {} unicode::{}[] = {{".format(data_type, name))
        for d in tbl:
            println(" {{ {} }},".format(", ".join(str(e) for e in d)))
        println("};")
        println("")

        dump(idx1, idx1_name, println)
        println("")
        dump(idx2, idx2_name, println)
        println("")

    def write_supplemental_identifier_method(name, group_set, println):
        println("bool")
        println("js::unicode::{}(char32_t codePoint)".format(name))
        println("{")
        for from_code, to_code in int_ranges(group_set.keys()):
            println(
                " if (codePoint >= 0x{:X} && codePoint <= 0x{:X}) {{ // {} .. {}".format(
                    from_code,
                    to_code,
                    codepoint_table.name(from_code),
                    codepoint_table.name(to_code),
                )
            )
            println(" return true;")
            println(" }")
        println(" return false;")
        println("}")
        println("")

    file_name = "Unicode.cpp"
    with io.open(file_name, "w", encoding="utf-8") as data_file:
        write = partial(print, file=data_file, sep="", end="")
        println = partial(print, file=data_file, sep="", end="\n")

        write(warning_message)
        write(unicode_version_message.format(version))
        write(public_domain)
        println('#include "util/Unicode.h"')
        println("")
        println("using namespace js;")
        println("using namespace js::unicode;")
        write(comment)

        write_table(
            "CharacterInfo",
            "js_charinfo",
            table,
            "index1",
            index1,
            "index2",
            index2,
            println,
        )

        write_table(
            "FoldingInfo",
            "js_foldinfo",
            folding_table,
            "folding_index1",
            folding_index1,
            "folding_index2",
            folding_index2,
            println,
        )

        # If the following assert fails, it means space character is added to
        # non-BMP area.  In that case the following code should be uncommented
        # and the corresponding code should be added to frontend.  (At least
        # unicode::IsSpace will require updating to handle this.)
        assert len(non_bmp_space_set.keys()) == 0

        write_supplemental_identifier_method(
            "IsIdentifierStartNonBMP", non_bmp_id_start_set, println
        )

        write_supplemental_identifier_method(
            "IsIdentifierPartNonBMP", non_bmp_id_cont_set, println
        )

        write_special_casing_methods(unconditional_toupper, codepoint_table, println)

        write_ascii_lookup_tables(table, index, write, println)

        write_latin1_lookup_tables(table, index, write, println)


def getsize(data):
    """return smallest possible integer size for the given array"""
    maxdata = max(data)
    assert maxdata < 2**32

    if maxdata < 256:
        return 1
    elif maxdata < 65536:
        return 2
    else:
        return 4


def splitbins(t):
    """t -> (t1, t2, shift). Split a table to save space.

    t is a sequence of ints.  This function can be useful to save space if
    many of the ints are the same.  t1 and t2 are lists of ints, and shift
    is an int, chosen to minimize the combined size of t1 and t2 (in C
    code), and where for each i in range(len(t)),
        t[i] == t2[(t1[i >> shift] << shift) + (i & mask)]
    where mask is a bitmask isolating the last "shift" bits.
    """

    def dump(t1, t2, shift, bytes):
        print(
            "%d+%d bins at shift %d; %d bytes" % (len(t1), len(t2), shift, bytes),
            file=sys.stderr,
        )
        print("Size of original table:", len(t) * getsize(t), "bytes", file=sys.stderr)

    n = len(t) - 1  # last valid index
    maxshift = 0  # the most we can shift n and still have something left
    if n > 0:
        while n >> 1:
            n >>= 1
            maxshift += 1
    del n
    bytes = sys.maxsize  # smallest total size so far
    t = tuple(t)  # so slices can be dict keys
    for shift in range(maxshift + 1):
        t1 = []
        t2 = []
        size = 2**shift
        bincache = {}

        for i in range(0, len(t), size):
            bin = t[i : i + size]

            index = bincache.get(bin)
            if index is None:
                index = len(t2)
                bincache[bin] = index
                t2.extend(bin)
            t1.append(index >> shift)

        # determine memory size
        b = len(t1) * getsize(t1) + len(t2) * getsize(t2)
        if b < bytes:
            best = t1, t2, shift
            bytes = b
    t1, t2, shift = best

    print("Best:", end=" ", file=sys.stderr)
    dump(t1, t2, shift, bytes)

    # exhaustively verify that the decomposition is correct
    mask = 2**shift - 1
    for i in range(len(t)):
        assert t[i] == t2[(t1[i >> shift] << shift) + (i & mask)]
    return best


def update_unicode(args):
    base_path = os.getcwd()

    version = args.version
    if version is not None:
        baseurl = "https://unicode.org/Public"
        if version == "UNIDATA":
            url = "%s/%s" % (baseurl, version)
        else:
            url = "%s/%s/ucd" % (baseurl, version)

    print("Arguments:")
    if version is not None:
        print("\tVersion: %s" % version)
        print("\tDownload url: %s" % url)

        request_url = "{}/UCD.zip".format(url)
        with closing(urlopen(request_url)) as downloaded_file:
            downloaded_data = io.BytesIO(downloaded_file.read())

        with ZipFile(downloaded_data) as zip_file:
            for fname in [
                "UnicodeData.txt",
                "CaseFolding.txt",
                "DerivedCoreProperties.txt",
                "SpecialCasing.txt",
            ]:
                zip_file.extract(fname, path=base_path)
    else:
        print("\tUsing local files.")
        print("\tAlways make sure you have the newest Unicode files!")
    print("")

    def version_from_file(f, fname):
        pat_version = re.compile(r"# %s-(?P\d+\.\d+\.\d+).txt" % fname)
        return pat_version.match(f.readline()).group("version")

    with io.open(
        os.path.join(base_path, "UnicodeData.txt"), "r", encoding="utf-8"
    ) as unicode_data, io.open(
        os.path.join(base_path, "CaseFolding.txt"), "r", encoding="utf-8"
    ) as case_folding, io.open(
        os.path.join(base_path, "DerivedCoreProperties.txt"), "r", encoding="utf-8"
    ) as derived_core_properties, io.open(
        os.path.join(base_path, "SpecialCasing.txt"), "r", encoding="utf-8"
    ) as special_casing:
        unicode_version = version_from_file(
            derived_core_properties, "DerivedCoreProperties"
        )

        print("Processing...")
        (
            table,
            index,
            non_bmp_lower_map,
            non_bmp_upper_map,
            non_bmp_space_set,
            non_bmp_id_start_set,
            non_bmp_id_cont_set,
            codepoint_table,
            test_space_table,
        ) = process_unicode_data(unicode_data, derived_core_properties)
        (folding_table, folding_index, folding_tests) = process_case_folding(
            case_folding
        )
        (unconditional_tolower, unconditional_toupper) = process_special_casing(
            special_casing, table, index
        )

    print("Generating...")
    make_unicode_file(
        unicode_version,
        table,
        index,
        folding_table,
        folding_index,
        non_bmp_space_set,
        non_bmp_id_start_set,
        non_bmp_id_cont_set,
        unconditional_toupper,
        codepoint_table,
    )
    make_non_bmp_file(
        unicode_version, non_bmp_lower_map, non_bmp_upper_map, codepoint_table
    )

    make_bmp_mapping_test(
        unicode_version, codepoint_table, unconditional_tolower, unconditional_toupper
    )
    make_non_bmp_mapping_test(
        unicode_version, non_bmp_upper_map, non_bmp_lower_map, codepoint_table
    )
    make_space_test(unicode_version, test_space_table, codepoint_table)
    make_regexp_space_test(unicode_version, test_space_table, codepoint_table)
    make_icase_test(unicode_version, folding_tests, codepoint_table)


if __name__ == "__main__":
    import argparse

    # This script must be run from js/src/util to work correctly.
    if "/".join(os.path.normpath(os.getcwd()).split(os.sep)[-3:]) != "js/src/util":
        raise RuntimeError("%s must be run from js/src/util" % sys.argv[0])

    parser = argparse.ArgumentParser(description="Update Unicode data.")

    parser.add_argument(
        "--version",
        help='Optional Unicode version number. If specified, downloads the\
                              selected version from <https://unicode.org/Public>. If not specified\
                              uses the existing local files to generate the Unicode data. The\
                              number must match a published Unicode version, e.g. use\
                              "--version=8.0.0" to download Unicode 8 files. Alternatively use\
                              "--version=UNIDATA" to download the latest published version.',
    )

    parser.set_defaults(func=update_unicode)

    args = parser.parse_args()
    args.func(args)