view i18n/hggettext @ 50400:95acba2c29f6

encoding: avoid quadratic time complexity when json-encoding non-UTF8 strings

Apparently the code uses "+=" with a bytes object, which is linear-time, so the
whole encoding is quadratic-time. This patch makes us use a bytearray object,
instead, which has a(n amortized-)constant-time append operation.

The encoding is still not particularly fast, but at least a 10MB file
takes tens of seconds, not many hours to encode.

#!/usr/bin/env python3
#
# hggettext - carefully extract docstrings for Mercurial
#
# Copyright 2009 Olivia Mackall <[e-mail]> and others
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

# The normalize function is taken from pygettext which is distributed
# with Python under the Python License, which is GPL compatible.

"""Extract docstrings from Mercurial commands.

Compared to pygettext, this script knows about the cmdtable and table
dictionaries used by Mercurial, and will only extract docstrings from
functions mentioned therein.

Use xgettext like normal to extract strings marked as translatable and
join the message cataloges to get the final catalog.
"""


import inspect
import os
import re
import sys


def escape(s):
    # The order is important, the backslash must be escaped first
    # since the other replacements introduce new backslashes
    # themselves.
    s = s.replace('\\', '\\\\')
    s = s.replace('\n', '\\n')
    s = s.replace('\r', '\\r')
    s = s.replace('\t', '\\t')
    s = s.replace('"', '\\"')
    return s


def normalize(s):
    # This converts the various Python string types into a format that
    # is appropriate for .po files, namely much closer to C style.
    lines = s.split('\n')
    if len(lines) == 1:
        s = '"' + escape(s) + '"'
    else:
        if not lines[-1]:
            del lines[-1]
            lines[-1] = lines[-1] + '\n'
        lines = map(escape, lines)
        lineterm = '\\n"\n"'
        s = '""\n"' + lineterm.join(lines) + '"'
    return s


def poentry(path, lineno, s):
    return (
        '#: %s:%d\n' % (path, lineno)
        + 'msgid %s\n' % normalize(s)
        + 'msgstr ""\n'
    )


doctestre = re.compile(r'^ +>>> ', re.MULTILINE)


def offset(src, doc, name, lineno, default):
    """Compute offset or issue a warning on stdout."""
    # remove doctest part, in order to avoid backslash mismatching
    m = doctestre.search(doc)
    if m:
        doc = doc[: m.start()]

    # Backslashes in doc appear doubled in src.
    end = src.find(doc.replace('\\', '\\\\'))
    if end == -1:
        # This can happen if the docstring contains unnecessary escape
        # sequences such as \" in a triple-quoted string. The problem
        # is that \" is turned into " and so doc wont appear in src.
        sys.stderr.write(
            "%s:%d:warning:"
            " unknown docstr offset, assuming %d lines\n"
            % (name, lineno, default)
        )
        return default
    else:
        return src.count('\n', 0, end)


def importpath(path):
    """Import a path like foo/bar/baz.py and return the baz module."""
    if path.endswith('.py'):
        path = path[:-3]
    if path.endswith('/__init__'):
        path = path[:-9]
    path = path.replace('/', '.')
    mod = __import__(path)
    for comp in path.split('.')[1:]:
        mod = getattr(mod, comp)
    return mod


def docstrings(path):
    """Extract docstrings from path.

    This respects the Mercurial cmdtable/table convention and will
    only extract docstrings from functions mentioned in these tables.
    """
    mod = importpath(path)
    if not path.startswith('mercurial/') and mod.__doc__:
        with open(path) as fobj:
            src = fobj.read()
        lineno = 1 + offset(src, mod.__doc__, path, 1, 7)
        print(poentry(path, lineno, mod.__doc__))

    functions = list(getattr(mod, 'i18nfunctions', []))
    functions = [(f, True) for f in functions]

    cmdtable = getattr(mod, 'cmdtable', {})
    if not cmdtable:
        # Maybe we are processing mercurial.commands?
        cmdtable = getattr(mod, 'table', {})
    functions.extend((c[0], False) for c in cmdtable.values())

    for func, rstrip in functions:
        if func.__doc__:
            docobj = func  # this might be a proxy to provide formatted doc
            func = getattr(func, '_origfunc', func)
            funcmod = inspect.getmodule(func)
            extra = ''
            if funcmod.__package__ == funcmod.__name__:
                extra = '/__init__'
            actualpath = '%s%s.py' % (funcmod.__name__.replace('.', '/'), extra)

            src = inspect.getsource(func)
            lineno = inspect.getsourcelines(func)[1]
            doc = docobj.__doc__
            origdoc = getattr(docobj, '_origdoc', '')
            if rstrip:
                doc = doc.rstrip()
                origdoc = origdoc.rstrip()
            if origdoc:
                lineno += offset(src, origdoc, actualpath, lineno, 1)
            else:
                lineno += offset(src, doc, actualpath, lineno, 1)
            print(poentry(actualpath, lineno, doc))


def rawtext(path):
    with open(path) as f:
        src = f.read()
    print(poentry(path, 1, src))


if __name__ == "__main__":
    # It is very important that we import the Mercurial modules from
    # the source tree where hggettext is executed. Otherwise we might
    # accidentally import and extract strings from a Mercurial
    # installation mentioned in PYTHONPATH.
    sys.path.insert(0, os.getcwd())
    from mercurial import demandimport

    demandimport.enable()
    for path in sys.argv[1:]:
        if path.endswith('.txt'):
            rawtext(path)
        else:
            docstrings(path)