view mercurial/templatefilters.py @ 47010:76ae43d5b1db stable

repoview: fix memory leak of filtered repo classes

The leak occurs in long-running server processes with
extensions, and is measured at 110kB per request.

Before this change, the contents of the `_filteredrepotypes`
cache are not properly garbage collected, despite it begin
a `WeakKeyDictionary`.

Extensions have a tendency to generate a new repository class
for each `localrepo` instantiation. Server processes based
on `hgwebdir_mod` will instantiate a new `localrepo` for each
HTTP request that involves a repository.

As a result, with a testing process that repeatedly opens a
repository with several extensions activated
(`topic` notably among them), we see a steady increase in
resident memory of 110kB per repository instantiation before this
change. This is also true, if we call `gc.collect()` at each
instantiation, like `hgwebdir_mod` does, or not.

The cause of the leak is that the *values* aren't weak references.
This change uses `weakref.ref` for the values, and this makes
in our measurements the resident size increase drop to 5kB per
repository instantiation, with no explicit call of `gc.collect()`
at all.
There is currently no reason to believe that this remaining leak
of 5kB is related to or even due to Mercurial core.

We've also seen evidence that `ui.ui` instances weren't properly
garbage collected before the change (with the change, they are).
This could explain why the figures are relatively high.

In theory, the collection of weak references could lead to
much more misses in the cache, so we measured the impact on
the original case that was motivation for introducing that cache
in 7e89bd0cfb86 (see also issue5043): `hg convert` of the
mozilla-central repository. The bad news here is that there is a
major memory leak there, both with and without the present changeset.
There were no more cache misses, and we could see no
more memory leak with this change: the resident size after importing
roughly 100000 changesets was at 12.4GB before, and 12.5GB after.
The small increase is mentioned for completeness only, and we
believe that it should be ignored, at least as long as the main
leak isn't fixed. At less than 1% of the main leak, even finding out
whether it is merely noise would be wasteful.

Original context where this was spotted and first mitigated:
  https://foss.heptapod.net/heptapod/heptapod/-/issues/466

The leak reduction was also obtained in Heptapod inner HTTP server,
which amounts to the same as `hgwebdir_mod` for these questions.

The measurements done with Python 3.9, similar figures seen with 3.8.
More work on our side would be needed to give measurements with 2.7,
because of testing server process does not support it.

# templatefilters.py - common template expansion filters
#
# Copyright 2005-2008 Olivia Mackall <[e-mail]>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

from __future__ import absolute_import

import os
import re
import time

from .i18n import _
from .node import hex
from . import (
    encoding,
    error,
    pycompat,
    registrar,
    smartset,
    templateutil,
    url,
    util,
)
from .utils import (
    cborutil,
    dateutil,
    stringutil,
)

urlerr = util.urlerr
urlreq = util.urlreq

# filters are callables like:
#   fn(obj)
# with:
#   obj - object to be filtered (text, date, list and so on)
filters = {}

templatefilter = registrar.templatefilter(filters)


@templatefilter(b'addbreaks', intype=bytes)
def addbreaks(text):
    """Any text. Add an XHTML "<br />" tag before the end of
    every line except the last.
    """
    return text.replace(b'\n', b'<br/>\n')


agescales = [
    (b"year", 3600 * 24 * 365, b'Y'),
    (b"month", 3600 * 24 * 30, b'M'),
    (b"week", 3600 * 24 * 7, b'W'),
    (b"day", 3600 * 24, b'd'),
    (b"hour", 3600, b'h'),
    (b"minute", 60, b'm'),
    (b"second", 1, b's'),
]


@templatefilter(b'age', intype=templateutil.date)
def age(date, abbrev=False):
    """Date. Returns a human-readable date/time difference between the
    given date/time and the current date/time.
    """

    def plural(t, c):
        if c == 1:
            return t
        return t + b"s"

    def fmt(t, c, a):
        if abbrev:
            return b"%d%s" % (c, a)
        return b"%d %s" % (c, plural(t, c))

    now = time.time()
    then = date[0]
    future = False
    if then > now:
        future = True
        delta = max(1, int(then - now))
        if delta > agescales[0][1] * 30:
            return b'in the distant future'
    else:
        delta = max(1, int(now - then))
        if delta > agescales[0][1] * 2:
            return dateutil.shortdate(date)

    for t, s, a in agescales:
        n = delta // s
        if n >= 2 or s == 1:
            if future:
                return b'%s from now' % fmt(t, n, a)
            return b'%s ago' % fmt(t, n, a)


@templatefilter(b'basename', intype=bytes)
def basename(path):
    """Any text. Treats the text as a path, and returns the last
    component of the path after splitting by the path separator.
    For example, "foo/bar/baz" becomes "baz" and "foo/bar//" becomes "".
    """
    return os.path.basename(path)


def _tocborencodable(obj):
    if isinstance(obj, smartset.abstractsmartset):
        return list(obj)
    return obj


@templatefilter(b'cbor')
def cbor(obj):
    """Any object. Serializes the object to CBOR bytes."""
    # cborutil is stricter about type than json() filter
    obj = pycompat.rapply(_tocborencodable, obj)
    return b''.join(cborutil.streamencode(obj))


@templatefilter(b'commondir')
def commondir(filelist):
    """List of text. Treats each list item as file name with /
    as path separator and returns the longest common directory
    prefix shared by all list items.
    Returns the empty string if no common prefix exists.

    The list items are not normalized, i.e. "foo/../bar" is handled as
    file "bar" in the directory "foo/..". Leading slashes are ignored.

    For example, ["foo/bar/baz", "foo/baz/bar"] becomes "foo" and
    ["foo/bar", "baz"] becomes "".
    """

    def common(a, b):
        if len(a) > len(b):
            a = b[: len(a)]
        elif len(b) > len(a):
            b = b[: len(a)]
        if a == b:
            return a
        for i in pycompat.xrange(len(a)):
            if a[i] != b[i]:
                return a[:i]
        return a

    try:
        if not filelist:
            return b""
        dirlist = [f.lstrip(b'/').split(b'/')[:-1] for f in filelist]
        if len(dirlist) == 1:
            return b'/'.join(dirlist[0])
        a = min(dirlist)
        b = max(dirlist)
        # The common prefix of a and b is shared with all
        # elements of the list since Python sorts lexicographical
        # and [1, x] after [1].
        return b'/'.join(common(a, b))
    except TypeError:
        raise error.ParseError(_(b'argument is not a list of text'))


@templatefilter(b'count')
def count(i):
    """List or text. Returns the length as an integer."""
    try:
        return len(i)
    except TypeError:
        raise error.ParseError(_(b'not countable'))


@templatefilter(b'dirname', intype=bytes)
def dirname(path):
    """Any text. Treats the text as a path, and strips the last
    component of the path after splitting by the path separator.
    """
    return os.path.dirname(path)


@templatefilter(b'domain', intype=bytes)
def domain(author):
    """Any text. Finds the first string that looks like an email
    address, and extracts just the domain component. Example: ``User
    <[e-mail]>`` becomes ``example.com``.
    """
    f = author.find(b'@')
    if f == -1:
        return b''
    author = author[f + 1 :]
    f = author.find(b'>')
    if f >= 0:
        author = author[:f]
    return author


@templatefilter(b'email', intype=bytes)
def email(text):
    """Any text. Extracts the first string that looks like an email
    address. Example: ``User <[e-mail]>`` becomes
    ``[e-mail]``.
    """
    return stringutil.email(text)


@templatefilter(b'escape', intype=bytes)
def escape(text):
    """Any text. Replaces the special XML/XHTML characters "&", "<"
    and ">" with XML entities, and filters out NUL characters.
    """
    return url.escape(text.replace(b'\0', b''), True)


para_re = None
space_re = None


def fill(text, width, initindent=b'', hangindent=b''):
    '''fill many paragraphs with optional indentation.'''
    global para_re, space_re
    if para_re is None:
        para_re = re.compile(b'(\n\n|\n\\s*[-*]\\s*)', re.M)
        space_re = re.compile(br'  +')

    def findparas():
        start = 0
        while True:
            m = para_re.search(text, start)
            if not m:
                uctext = encoding.unifromlocal(text[start:])
                w = len(uctext)
                while w > 0 and uctext[w - 1].isspace():
                    w -= 1
                yield (
                    encoding.unitolocal(uctext[:w]),
                    encoding.unitolocal(uctext[w:]),
                )
                break
            yield text[start : m.start(0)], m.group(1)
            start = m.end(1)

    return b"".join(
        [
            stringutil.wrap(
                space_re.sub(b' ', stringutil.wrap(para, width)),
                width,
                initindent,
                hangindent,
            )
            + rest
            for para, rest in findparas()
        ]
    )


@templatefilter(b'fill68', intype=bytes)
def fill68(text):
    """Any text. Wraps the text to fit in 68 columns."""
    return fill(text, 68)


@templatefilter(b'fill76', intype=bytes)
def fill76(text):
    """Any text. Wraps the text to fit in 76 columns."""
    return fill(text, 76)


@templatefilter(b'firstline', intype=bytes)
def firstline(text):
    """Any text. Returns the first line of text."""
    try:
        return text.splitlines(True)[0].rstrip(b'\r\n')
    except IndexError:
        return b''


@templatefilter(b'hex', intype=bytes)
def hexfilter(text):
    """Any text. Convert a binary Mercurial node identifier into
    its long hexadecimal representation.
    """
    return hex(text)


@templatefilter(b'hgdate', intype=templateutil.date)
def hgdate(text):
    """Date. Returns the date as a pair of numbers: "1157407993
    25200" (Unix timestamp, timezone offset).
    """
    return b"%d %d" % text


@templatefilter(b'isodate', intype=templateutil.date)
def isodate(text):
    """Date. Returns the date in ISO 8601 format: "2009-08-18 13:00
    +0200".
    """
    return dateutil.datestr(text, b'%Y-%m-%d %H:%M %1%2')


@templatefilter(b'isodatesec', intype=templateutil.date)
def isodatesec(text):
    """Date. Returns the date in ISO 8601 format, including
    seconds: "2009-08-18 13:00:13 +0200". See also the rfc3339date
    filter.
    """
    return dateutil.datestr(text, b'%Y-%m-%d %H:%M:%S %1%2')


def indent(text, prefix, firstline=b''):
    '''indent each non-empty line of text after first with prefix.'''
    lines = text.splitlines()
    num_lines = len(lines)
    endswithnewline = text[-1:] == b'\n'

    def indenter():
        for i in pycompat.xrange(num_lines):
            l = lines[i]
            if l.strip():
                yield prefix if i else firstline
            yield l
            if i < num_lines - 1 or endswithnewline:
                yield b'\n'

    return b"".join(indenter())


@templatefilter(b'json')
def json(obj, paranoid=True):
    """Any object. Serializes the object to a JSON formatted text."""
    if obj is None:
        return b'null'
    elif obj is False:
        return b'false'
    elif obj is True:
        return b'true'
    elif isinstance(obj, (int, pycompat.long, float)):
        return pycompat.bytestr(obj)
    elif isinstance(obj, bytes):
        return b'"%s"' % encoding.jsonescape(obj, paranoid=paranoid)
    elif isinstance(obj, type(u'')):
        raise error.ProgrammingError(
            b'Mercurial only does output with bytes: %r' % obj
        )
    elif util.safehasattr(obj, b'keys'):
        out = [
            b'"%s": %s'
            % (encoding.jsonescape(k, paranoid=paranoid), json(v, paranoid))
            for k, v in sorted(pycompat.iteritems(obj))
        ]
        return b'{' + b', '.join(out) + b'}'
    elif util.safehasattr(obj, b'__iter__'):
        out = [json(i, paranoid) for i in obj]
        return b'[' + b', '.join(out) + b']'
    raise error.ProgrammingError(b'cannot encode %r' % obj)


@templatefilter(b'lower', intype=bytes)
def lower(text):
    """Any text. Converts the text to lowercase."""
    return encoding.lower(text)


@templatefilter(b'nonempty', intype=bytes)
def nonempty(text):
    """Any text. Returns '(none)' if the string is empty."""
    return text or b"(none)"


@templatefilter(b'obfuscate', intype=bytes)
def obfuscate(text):
    """Any text. Returns the input text rendered as a sequence of
    XML entities.
    """
    text = pycompat.unicode(
        text, pycompat.sysstr(encoding.encoding), r'replace'
    )
    return b''.join([b'&#%d;' % ord(c) for c in text])


@templatefilter(b'permissions', intype=bytes)
def permissions(flags):
    if b"l" in flags:
        return b"lrwxrwxrwx"
    if b"x" in flags:
        return b"-rwxr-xr-x"
    return b"-rw-r--r--"


@templatefilter(b'person', intype=bytes)
def person(author):
    """Any text. Returns the name before an email address,
    interpreting it as per RFC 5322.
    """
    return stringutil.person(author)


@templatefilter(b'revescape', intype=bytes)
def revescape(text):
    """Any text. Escapes all "special" characters, except @.
    Forward slashes are escaped twice to prevent web servers from prematurely
    unescaping them. For example, "@foo bar/baz" becomes "@foo%20bar%252Fbaz".
    """
    return urlreq.quote(text, safe=b'/@').replace(b'/', b'%252F')


@templatefilter(b'rfc3339date', intype=templateutil.date)
def rfc3339date(text):
    """Date. Returns a date using the Internet date format
    specified in RFC 3339: "2009-08-18T13:00:13+02:00".
    """
    return dateutil.datestr(text, b"%Y-%m-%dT%H:%M:%S%1:%2")


@templatefilter(b'rfc822date', intype=templateutil.date)
def rfc822date(text):
    """Date. Returns a date using the same format used in email
    headers: "Tue, 18 Aug 2009 13:00:13 +0200".
    """
    return dateutil.datestr(text, b"%a, %d %b %Y %H:%M:%S %1%2")


@templatefilter(b'short', intype=bytes)
def short(text):
    """Changeset hash. Returns the short form of a changeset hash,
    i.e. a 12 hexadecimal digit string.
    """
    return text[:12]


@templatefilter(b'shortbisect', intype=bytes)
def shortbisect(label):
    """Any text. Treats `label` as a bisection status, and
    returns a single-character representing the status (G: good, B: bad,
    S: skipped, U: untested, I: ignored). Returns single space if `text`
    is not a valid bisection status.
    """
    if label:
        return label[0:1].upper()
    return b' '


@templatefilter(b'shortdate', intype=templateutil.date)
def shortdate(text):
    """Date. Returns a date like "2006-09-18"."""
    return dateutil.shortdate(text)


@templatefilter(b'slashpath', intype=bytes)
def slashpath(path):
    """Any text. Replaces the native path separator with slash."""
    return util.pconvert(path)


@templatefilter(b'splitlines', intype=bytes)
def splitlines(text):
    """Any text. Split text into a list of lines."""
    return templateutil.hybridlist(text.splitlines(), name=b'line')


@templatefilter(b'stringescape', intype=bytes)
def stringescape(text):
    return stringutil.escapestr(text)


@templatefilter(b'stringify', intype=bytes)
def stringify(thing):
    """Any type. Turns the value into text by converting values into
    text and concatenating them.
    """
    return thing  # coerced by the intype


@templatefilter(b'stripdir', intype=bytes)
def stripdir(text):
    """Treat the text as path and strip a directory level, if
    possible. For example, "foo" and "foo/bar" becomes "foo".
    """
    dir = os.path.dirname(text)
    if dir == b"":
        return os.path.basename(text)
    else:
        return dir


@templatefilter(b'tabindent', intype=bytes)
def tabindent(text):
    """Any text. Returns the text, with every non-empty line
    except the first starting with a tab character.
    """
    return indent(text, b'\t')


@templatefilter(b'upper', intype=bytes)
def upper(text):
    """Any text. Converts the text to uppercase."""
    return encoding.upper(text)


@templatefilter(b'urlescape', intype=bytes)
def urlescape(text):
    """Any text. Escapes all "special" characters. For example,
    "foo bar" becomes "foo%20bar".
    """
    return urlreq.quote(text)


@templatefilter(b'user', intype=bytes)
def userfilter(text):
    """Any text. Returns a short representation of a user name or email
    address."""
    return stringutil.shortuser(text)


@templatefilter(b'emailuser', intype=bytes)
def emailuser(text):
    """Any text. Returns the user portion of an email address."""
    return stringutil.emailuser(text)


@templatefilter(b'utf8', intype=bytes)
def utf8(text):
    """Any text. Converts from the local character encoding to UTF-8."""
    return encoding.fromlocal(text)


@templatefilter(b'xmlescape', intype=bytes)
def xmlescape(text):
    text = (
        text.replace(b'&', b'&amp;')
        .replace(b'<', b'&lt;')
        .replace(b'>', b'&gt;')
        .replace(b'"', b'&quot;')
        .replace(b"'", b'&#39;')
    )  # &apos; invalid in HTML
    return re.sub(b'[\x00-\x08\x0B\x0C\x0E-\x1F]', b' ', text)


def websub(text, websubtable):
    """:websub: Any text. Only applies to hgweb. Applies the regular
    expression replacements defined in the websub section.
    """
    if websubtable:
        for regexp, format in websubtable:
            text = regexp.sub(format, text)
    return text


def loadfilter(ui, extname, registrarobj):
    """Load template filter from specified registrarobj"""
    for name, func in pycompat.iteritems(registrarobj._table):
        filters[name] = func


# tell hggettext to extract docstrings from these functions:
i18nfunctions = filters.values()