hgext/mozext/__init__.py

# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

"""make Mozilla contributors more productive

This extension adds Mozilla-centric commands and functionality to Mercurial
to enable contributors to Firefox and related applications to be more
productive.

Included are commands that interface with Mozilla's automation infrastructure.
These allow developers to quickly check the status of repositories and builds
without having to leave the comfort of the terminal.

Pushlog Data
============

This extension supports downloading pushlog data from Mozilla's repositories.
Pushlog data records who pushed changesets where and when.

To use this functionality, you'll need to sync pushlog data from the server to
the local machine. This facilitates rapid data lookup and can be done by
running `hg pull` from the specified repo with this extension installed.

Once pushlog data is synced, you can use `hg changesetpushes` to look up push
information for a specific changeset.

Bug Info
========

Information about bugs is extracted from commit messages as changesets are
introduced into the repository. You can look up information about specific
bugs via `hg buginfo`.

Revisions Sets
==============

This extension adds the following revision set selectors functions.

bug(BUG)
   Retreive changesets that reference a specific bug. e.g. ``bug(784841)``.

dontbuild()
   Retrieve changesets that are marked as DONTBUILD.

me()
   Retrieve changesets that you are involved with.

   This retrieves changesets you authored (via ui.username) as well as
   changesets you reviewed (via mozext.ircnick).

nobug()
   Retrieve changesets that don't reference a bug in the commit message.

pushhead([TREE])
   Retrieve changesets that are push heads.

firstpushdate(DATE)
   Retrieve changesets that were first pushed according to the date spec.

firstpushtree(TREE)
   Retrieve changesets that initially landed on the specified tree.

pushdate(DATE)
   Retrieve changesets that were pushed according to the date spec.

reviewer(REVIEWER)
   Retrieve changesets there were reviewed by the specified person.

   The reviewer string matches the *r=* string specified in the commit. In
   the future, we may consult a database of known aliases, etc.

reviewed()
   Retrieve changesets that have a reviewer marked.

tree(TREE)
   Retrieve changesets that are currently in the specified tree.

   Trees are specified with a known alias, typically a tree from the `firefoxtrees`
   extension, e.g. ``tree(central)``

   It's possible to see which changesets are in some tree but not others.
   e.g. to find changesets in *inbound* that haven't merged to *central*
   yet, do ``tree(inbound) - tree(central)``. This is essentially the same
   as ``::inbound - ::central``.

Templates
=========

This extension provides keywords that can used with templates.

bug
   The bug primarily associated with a changeset.

bugs
   All the bugs associated with a changeset.

firstrelease
   The version number of the first release channel release a changeset
   was present in.

firstbeta
   The version number of the first beta channel release a changeset was
   present in.

firstnightly
   The version number of the first nightly channel release a changeset was
   present in.

nightlydate
   The date of the first Nightly a changeset was likely present in.

firstpushuser
   The username of the first person who pushed this changeset.

firstpushtree
   The name of the first tree this changeset was pushed to.

firstpushtreeherder
   The URL of the Treeherder results for the first push of this changeset.

firstpushdate
   The date of the first push of this changeset.

pushdates
   The list of dates a changeset was pushed.

pushheaddates
   The list of dates a changeset was pushed as a push head.

trees
   The list of trees a changeset has landed in.

reltrees
   The list of release trees a changeset has landed in.

This extension provides the following template functions:

dates(VALUES, [format, [sep]])
   Format a list of dates.

   If the second argument is defined, the specified date formatting
   string will be used. Else, it defaults to '%Y-%m-%d'.

   If the third argument is defined, elements will be separated by the
   specified string. Else, ',' is used.

Config Options
==============

This extension consults the following config options.

mozext.headless
   Indicates that this extension is running in *headless* mode. *headless*
   mode is intended for server operation, not local development.

mozext.ircnick
   Your Mozilla IRC nickname/Phabricator account name. This string value
   will be used to look for your reviews and patches, on top of the value
   set in ``ui.username``.

mozext.disable_local_database
   When this boolean flag is true, the local SQLite database indexing
   useful lookups (such as bugs and pushlog info) is not enabled.

mozext.reject_pushes_with_repo_names
   This boolean is used to enable a ``prepushkey`` hook that prevents
   pushes to keys (bookmarks, tags, etc) whose name is prefixed with that
   of an official Mozilla repository.

   This hook is useful for servers exposing a monolithic repository where
   each separate Mozilla repository is exposed through bookmarks and where
   the server does not want to allow external users from changing the
   *canonical* refs.

   For example, with this flag set, pushes to bookmarks ``central/default``
   and ``inbound/foobar`` will be rejected because they begin with the names
   of official Mozilla repositories. However, pushes to the bookmark
   ``gps/test`` will be allowed.
"""

import calendar
import datetime
import errno
import gc
import os
import re
import sys
from collections import (
    defaultdict,
    namedtuple,
)

from operator import methodcaller

from mercurial.i18n import _
from mercurial.error import (
    ParseError,
    RepoError,
)
from mercurial.localrepo import (
    repofilecache,
)
from mercurial.node import (
    bin,
    hex,
    nullid,
    short,
)
from mercurial import (
    cmdutil,
    commands,
    configitems,
    demandimport,
    encoding,
    error,
    exchange,
    extensions,
    hg,
    mdiff,
    patch,
    pycompat,
    registrar,
    revset,
    scmutil,
    sshpeer,
    templatefilters,
    templateutil,
    util,
    pathutil,
    url,
)
from mercurial.utils import (
    dateutil,
)


OUR_DIR = os.path.normpath(os.path.dirname(__file__))
with open(os.path.join(OUR_DIR, "..", "bootstrap.py")) as f:
    exec(f.read())

from mozhg.util import (
    import_module,
    get_backoutbynode,
)

logcmdutil = import_module("mercurial.logcmdutil")
getlogrevs = logcmdutil.getrevs

# Disable demand importing for mozautomation because "requests" doesn't
# play nice with the demand importer.
with demandimport.deactivated():
    from mozautomation.changetracker import (
        ChangeTracker,
    )

    from mozautomation.commitparser import (
        BUG_CONSERVATIVE_RE,
        parse_backouts,
        parse_bugs,
        parse_reviewers,
    )

    from mozautomation.repository import (
        MercurialRepository,
        RELEASE_TREES,
        REPOS,
        resolve_trees_to_official,
        resolve_trees_to_uris,
        resolve_uri_to_tree,
        treeherder_url,
        TREE_ALIASES,
    )

testedwith = b"4.7 4.8 4.9 5.0 5.1 5.2"
minimumhgversion = b"4.7"
buglink = b"https://bugzilla.mozilla.org/enter_bug.cgi?product=Developer%20Services&component=Mercurial%3A%20mozext"

cmdtable = {}

command = registrar.command(cmdtable)

revsetpredicate = registrar.revsetpredicate()
templatekeyword = registrar.templatekeyword()
templatefunc = registrar.templatefunc()

configtable = {}
configitem = registrar.configitem(configtable)

configitem(b"mozext", b"headless", default=None)
configitem(b"mozext", b"ircnick", default=None)
configitem(b"mozext", b"disable_local_database", default=False)
configitem(b"mozext", b"reject_pushes_with_repo_names", default=False)
configitem(b"mozext", b"backoutsearchlimit", default=configitems.dynamicdefault)
configitem(b"reviewers", b".*", generic=True, default=configitems.dynamicdefault)


colortable = {
    b"buildstatus.success": b"green",
    b"buildstatus.failed": b"red",
    b"buildstatus.testfailed": b"cyan",
}

backout_re = re.compile(
    rb"[bB]ack(?:ed)?(?: ?out) (?:(?:changeset|revision|rev) )?([a-fA-F0-9]{8,40})"
)
reapply_re = re.compile(
    rb"Reapplied (?:(?:changeset|revision|rev) )?([a-fA-F0-9]{8,40})"
)


def get_ircnick(ui):
    headless = ui.configbool(b"mozext", b"headless")
    ircnick = ui.config(b"mozext", b"ircnick")
    if not ircnick and not headless:
        raise error.Abort(
            _(
                b'Set "[mozext] ircnick" in your hgrc to your '
                b"Mozilla IRC nickname to enable additional functionality."
            )
        )
    return ircnick


def wrapped_peer(orig, uiorrepo, opts, path, **kwargs):
    """Wrap `hg.peer` to return peer repos for known tree names (autoland, central etc)."""
    try:
        return orig(uiorrepo, opts, path, **kwargs)
    except RepoError:
        tree, uri = resolve_trees_to_uris([path.rawloc])[0]
        if not uri:
            raise

        return orig(uiorrepo, opts, uri, **kwargs)


def wrapped_peerorrepo(orig, ui, path, *args, **kwargs):
    # Always try the old mechanism first. That way if there is a local
    # path that shares the name of a magic remote the local path is accessible.
    try:
        return orig(ui, path, *args, **kwargs)
    except RepoError:
        tree, uri = resolve_trees_to_uris([path])[0]

        if not uri:
            raise

        path = uri
        return orig(ui, path, *args, **kwargs)


def exchangepullpushlog(orig, pullop):
    res = orig(pullop)

    if not pullop.remote.capable(b"pushlog"):
        return res

    # stepsdone added in Mercurial 3.2.
    if util.safehasattr(pullop, "stepsdone") and b"pushlog" in pullop.stepsdone:
        return res

    repo = pullop.repo

    tree = resolve_uri_to_tree(pullop.remote.url())
    if not tree or not repo.changetracker or tree == b"try":
        return res

    # Calling wire protocol commands via SSH requires the server-side wire
    # protocol code to be known by the client. The server-side code is defined
    # by the pushlog extension, so we effectively need the pushlog extension
    # enabled to call the wire protocol method when pulling via SSH. We don't
    # (yet) recommend installing the pushlog extension locally. Furthermore,
    # pulls from hg.mozilla.org should be performed via https://, not ssh://.
    # So just bail on pushlog fetching if pulling via ssh://.
    if isinstance(pullop.remote, sshpeer.sshv1peer):
        pullop.repo.ui.warn(
            b"cannot fetch pushlog when pulling via ssh://; "
            b"you should be pulling via https://\n"
        )
        return res

    lastpushid = repo.changetracker.last_push_id(tree)
    fetchfrom = lastpushid + 1 if lastpushid is not None else 0

    lines = pullop.remote._call(b"pushlog", firstpush=str(fetchfrom))
    lines = iter(lines.splitlines())

    statusline = next(lines)
    if statusline[0] == b"0":
        raise error.Abort(b"remote error fetching pushlog: %s" % next(lines))
    elif statusline != b"1":
        raise error.Abort(
            b"error fetching pushlog: unexpected response: %s\n" % statusline
        )

    pushes = []
    for line in lines:
        pushid, who, when, nodes = line.split(b" ", 3)
        nodes = [bin(n) for n in nodes.split()]

        # Warn when incoming changesets are unknown. This can happen when we're
        # pulling a former head instead of all changesets, when there are obsolete
        # changesets that are not being pulled, or when relbranches are used.
        for node in nodes:
            try:
                repo[node]
            except error.RepoLookupError:
                repo.ui.warn(
                    b"received pushlog entry for unknown changeset %s; ignoring\n"
                    % (hex(node),)
                )

        pushes.append((int(pushid), who, int(when), nodes))

    if pushes:
        repo.changetracker.add_pushes(tree, pushes)
        repo.ui.status(b"added %d pushes\n" % len(pushes))

    return res


@command(b"treestatus", [], _(b"hg treestatus [TREE] ..."), norepo=True)
def treestatus(ui, *trees, **opts):
    """Show the status of the Mozilla repositories.

    If trees are open, it is OK to land code on them.

    If trees require approval, you'll need to obtain approval from
    release management to land changes.

    If trees are closed, you shouldn't push unless you are fixing the reason
    the tree is closed.
    """
    from mozautomation.treestatus import TreeStatusClient

    client = TreeStatusClient()
    status = client.all()

    trees = resolve_trees_to_official(trees)

    if trees:
        for k in set(status.keys()) - set(trees):
            del status[k]
    if not status:
        raise error.Abort(b"No status info found.")

    longest = max(len(s) for s in status)

    for tree in sorted(status):
        s = status[tree]
        if s.status == b"closed":
            ui.write(b"%s: %s (%s)\n" % (tree.rjust(longest), s.status, s.reason))
        else:
            ui.write(b"%s: %s\n" % (tree.rjust(longest), s.status))


@command(b"treeherder", [], _(b"hg treeherder [TREE] [REV]"))
def treeherder(ui, repo, tree=None, rev=None, **opts):
    """Open Treeherder showing build status for the specified revision.

    The command receives a tree name and a revision to query. The tree is
    required because a revision/changeset may existing in multiple
    repositories.
    """
    if not tree:
        raise error.Abort(b"A tree must be specified.")

    if not rev:
        raise error.Abort(b"A revision must be specified.")

    tree, repo_url = resolve_trees_to_uris([tree])[0]
    if not repo_url:
        raise error.Abort(b"Don't know about tree: %s" % tree)

    r = MercurialRepository(repo_url)
    node = repo[rev].hex()
    push = r.push_info_for_changeset(node)

    if not push:
        raise error.Abort(b"Could not find push info for changeset %s" % node)

    push_node = push.last_node

    url = treeherder_url(tree, push_node)

    import webbrowser

    webbrowser.get("firefox").open(url)


def print_changeset_pushes(ui, repo, rev, all=False):
    if not repo.changetracker:
        ui.warn(b"Local database appears to be disabled.")
        return 1

    ctx = repo[rev]
    node = ctx.node()

    pushes = repo.changetracker.pushes_for_changeset(node)
    pushes = [p for p in pushes if all or p[0] in RELEASE_TREES]
    if not pushes:
        ui.warn(b"No pushes recorded for changeset: ", str(ctx), "\n")
        return 1

    longest_tree = max(len(p[0]) for p in pushes) + 2
    longest_user = max(len(p[3]) for p in pushes) + 2

    ui.write(str(ctx.rev()), b":", str(ctx), b" ", ctx.description(), b"\n")

    ui.write(
        b"Release ",
        b"Tree".ljust(longest_tree),
        b"Date".ljust(20),
        b"Username".ljust(longest_user),
        b"Build Info\n",
    )
    for tree, push_id, when, user, head_node in pushes:
        releases = set()
        release = ""
        versions = {}

        if tree == b"beta":
            versions = repo._beta_releases()
        elif tree == b"release":
            versions = repo._release_releases()

        for version, e in versions.items():
            vctx = repo[e[0]]
            if ctx.descendant(vctx):
                releases.add(version)

        if len(releases):
            release = sorted(releases)[0]

        url = treeherder_url(tree, hex(head_node))
        date = datetime.datetime.fromtimestamp(when)
        ui.write(
            release.ljust(8),
            tree.ljust(longest_tree),
            date.isoformat(),
            b" ",
            user.ljust(longest_user),
            url or b"",
            b"\n",
        )


@command(
    b"changesetpushes",
    [(b"a", b"all", False, _(b"Show all trees, not just release trees."), b"")],
    _(b"hg changesetpushes REV"),
)
def changesetpushes(ui, repo, rev, all=False, **opts):
    """Display pushlog information for a changeset.

    This command prints pushlog entries for a given changeset. It is used to
    answer the question: how did a changeset propagate to all the trees.
    """
    print_changeset_pushes(ui, repo, scmutil.revsingle(repo, rev), all=all)


@command(
    b"buginfo",
    [
        (b"a", b"all", False, _(b"Show all trees, not just release trees."), b""),
        (b"", b"reset", False, _(b"Wipe and repopulate the bug database."), b""),
        (b"", b"sync", False, _(b"Synchronize the bug database."), b""),
    ],
    _(b"hg buginfo [BUG] ..."),
)
def buginfo(ui, repo, *bugs, **opts):
    if not repo.changetracker:
        ui.warning(b"Local database appears to be disabled")
        return 1

    if opts["sync"]:
        repo.sync_bug_database()
        return

    if opts["reset"]:
        repo.reset_bug_database()
        return

    tracker = repo.changetracker

    nodes = set()
    for bug in bugs:
        nodes |= set(tracker.changesets_with_bug(bug))

    # Sorting by topological order would probably be preferred. This is quick
    # and easy.
    contexts = sorted([repo[node] for node in nodes], key=methodcaller("rev"))

    for ctx in contexts:
        print_changeset_pushes(ui, repo, ctx.rev(), all=opts["all"])
        ui.write(b"\n")


def reject_repo_names_hook(
    ui, repo, namespace=None, key=None, old=None, new=None, **kwargs
):
    """prepushkey hook that prevents changes to reserved names.

    Names that begin with the name of a repository identifier are rejected.
    """
    if key.lower().startswith(tuple(REPOS.keys())):
        ui.warn(
            b"You are not allowed to push tags or bookmarks that share "
            b"names with official Mozilla repositories: %s\n" % key
        )
        return True

    return False


def wrappedpull(orig, repo, remote, *args, **kwargs):
    """Wraps exchange.pull to add remote tracking refs."""
    if not util.safehasattr(repo, "changetracker"):
        return orig(repo, remote, *args, **kwargs)

    old_rev = len(repo)
    res = orig(repo, remote, *args, **kwargs)
    lock = repo.wlock()
    try:
        tree = resolve_uri_to_tree(remote.url())

        if tree:
            repo._update_remote_refs(remote, tree)

        # Sync bug info.
        for rev in repo.changelog.revs(old_rev + 1):
            ctx = repo[rev]
            bugs = parse_bugs(ctx.description())
            if bugs and repo.changetracker:
                repo.changetracker.associate_bugs_with_changeset(bugs, ctx.node())

    finally:
        lock.release()

    return res


def wrappedpush(orig, repo, remote, *args, **kwargs):
    if not util.safehasattr(repo, "changetracker"):
        return orig(repo, remote, *args, **kwargs)

    res = orig(repo, remote, *args, **kwargs)
    lock = repo.wlock()
    try:
        tree = resolve_uri_to_tree(remote.url())

        if tree:
            repo._update_remote_refs(remote, tree)

    finally:
        lock.release()

    return res


class remoterefs(dict):
    """Represents a remote refs file."""

    def __init__(self, repo):
        dict.__init__(self)
        self._repo = repo

        try:
            for line in repo.vfs(b"remoterefs"):
                line = line.strip()
                if not line:
                    continue

                sha, ref = line.split(None, 1)
                ref = encoding.tolocal(ref)
                try:
                    self[ref] = repo.changelog.lookup(sha)
                except LookupError:
                    pass

        except IOError as e:
            if e.errno != errno.ENOENT:
                raise

    def write(self):
        f = self._repo.vfs(b"remoterefs", b"w", atomictemp=True)
        for ref in sorted(self):
            f.write(b"%s %s\n" % (hex(self[ref]), encoding.fromlocal(ref)))
        f.close()


@revsetpredicate(b"bug(N)")
def revset_bug(repo, subset, x):
    """Changesets referencing a specified Bugzilla bug. e.g. bug(123456)."""
    err = _(b"bug() requires an integer argument.")
    bugstring = revset.getstring(x, err)

    try:
        bug = int(bugstring)
    except Exception:
        raise ParseError(err)

    def fltr(x):
        # We do a simple string test first because avoiding regular expressions
        # is good for performance.
        desc = repo[x].description()
        return bugstring in desc and bug in parse_bugs(desc)

    return subset.filter(fltr)


@revsetpredicate(b"dontbuild()")
def revset_dontbuild(repo, subset, x):
    if x:
        raise ParseError(_(b"dontbuild() does not take any arguments"))

    return subset.filter(lambda x: b"DONTBUILD" in repo[x].description())


@revsetpredicate(b"me()")
def revset_me(repo, subset, x):
    """Changesets that you are involved in."""
    if x:
        raise ParseError(_(b"me() does not take any arguments"))

    me = repo.ui.config(b"ui", b"username")
    if not me:
        raise error.Abort(_(b'"[ui] username" must be set to use me()'))

    ircnick = get_ircnick(repo.ui)

    n = encoding.lower(me)
    kind, pattern, matcher = revset._substringmatcher(n)

    def fltr(x):
        ctx = repo[x]
        if matcher(encoding.lower(ctx.user())):
            return True

        return ircnick in parse_reviewers(ctx.description())

    return subset.filter(fltr)


@revsetpredicate(b"nobug()")
def revset_nobug(repo, subset, x):
    if x:
        raise ParseError(_(b"nobug() does not take any arguments"))

    return subset.filter(lambda x: not parse_bugs(repo[x].description()))


def revset_tree(repo, subset, x):
    """``tree(X)``
    Changesets currently in the specified Mozilla tree.

    A tree is the name of a repository. e.g. ``central``.
    """
    err = _(b"tree() requires a string argument.")
    tree = revset.getstring(x, err)

    tree, uri = resolve_trees_to_uris([tree])[0]
    if not uri:
        raise error.Abort(_(b"Don't know about tree: %s") % tree)

    ref = b"%s/default" % tree

    head = repo[ref].rev()
    ancestors = set(repo.changelog.ancestors([head], inclusive=True))

    return subset & revset.baseset(ancestors)


def revset_firstpushdate(repo, subset, x):
    """``firstpushdate(DATE)``
    Changesets that were initially pushed according to the date spec provided.
    """
    ds = revset.getstring(x, _(b"firstpushdate() requires a string"))
    dm = dateutil.matchdate(ds)

    def fltr(x):
        pushes = list(repo.changetracker.pushes_for_changeset(repo[x].node()))

        if not pushes:
            return False

        when = pushes[0][2]

        return dm(when)

    return subset.filter(fltr)


def revset_firstpushtree(repo, subset, x):
    """``firstpushtree(X)``
    Changesets that were initially pushed to tree X.
    """
    tree = revset.getstring(x, _(b"firstpushtree() requires a string argument."))

    tree, uri = resolve_trees_to_uris([tree])[0]
    if not uri:
        raise error.Abort(_(b"Don't know about tree: %s") % tree)

    def fltr(x):
        pushes = list(repo.changetracker.pushes_for_changeset(repo[x].node()))

        if not pushes:
            return False

        return pushes[0][0] == tree

    return subset.filter(fltr)


def revset_pushdate(repo, subset, x):
    """``pushdate(DATE)``
    Changesets that were pushed according to the date spec provided.

    All pushes are examined.
    """
    ds = revset.getstring(x, _(b"pushdate() requires a string"))
    dm = dateutil.matchdate(ds)

    def fltr(x):
        for push in repo.changetracker.pushes_for_changeset(repo[x].node()):
            when = push[2]

            if dm(when):
                return True

        return False

    return subset.filter(fltr)


def revset_pushhead(repo, subset, x):
    """``pushhead([TREE])``
    Changesets that are push heads.

    A push head is a changeset that was a head when it was pushed to a
    repository. In other words, the automation infrastructure likely
    kicked off a build using this changeset.

    If an argument is given, we limit ourselves to pushes on the specified
    tree.

    If no argument is given, we return all push heads for all trees. Note that
    a changeset can be a push head multiple times. This function doesn't care
    where the push was made if no argument was given.
    """
    # We have separate code paths because the single tree path uses a single
    # query and is faster.
    if x:
        tree = revset.getstring(x, _(b"pushhead() requires a string argument."))
        tree, uri = resolve_trees_to_uris([tree])[0]

        if not uri:
            raise error.Abort(_(b"Don't know about tree: %s") % tree)

        def pushheads():
            for push_id, head_changeset in repo.changetracker.tree_push_head_changesets(
                tree
            ):
                head_changeset = pycompat.bytestr(head_changeset)
                try:
                    head = repo[head_changeset].rev()
                    yield head
                except error.RepoLookupError:
                    # There are some malformed pushes.  Ignore them.
                    continue

        # Push heads are returned in order of ascending push ID, which
        # corresponds to ascending commit order in hg.
        return subset & revset.generatorset(pushheads(), iterasc=True)
    else:

        def is_pushhead(r):
            node = repo[r].node()
            for push in repo.changetracker.pushes_for_changeset(node):
                if pycompat.bytestr(push[4]) == node:
                    return True
            return False

        return subset.filter(is_pushhead)


@revsetpredicate(b"reviewer(REVIEWER)")
def revset_reviewer(repo, subset, x):
    """Changesets reviewed by a specific person."""
    n = revset.getstring(x, _(b"reviewer() requires a string argument."))

    return subset.filter(lambda x: n in parse_reviewers(repo[x].description()))


@revsetpredicate(b"reviewed()")
def revset_reviewed(repo, subset, x):
    """Changesets that were reviewed."""
    if x:
        raise ParseError(_(b"reviewed() does not take an argument"))

    return subset.filter(lambda x: list(parse_reviewers(repo[x].description())))


@templatekeyword(b"bug", requires={b"ctx"})
def template_bug(context, mapping):
    """:bug: String. The bug this changeset is most associated with."""
    ctx = context.resource(mapping, b"ctx")

    bugs = parse_bugs(ctx.description())
    return bugs[0] if bugs else None


@templatekeyword(b"backedoutby", requires={b"repo", b"ctx"})
def template_backedoutby(context, mapping):
    repo = context.resource(mapping, b"repo")
    ctx = context.resource(mapping, b"ctx")

    return get_backoutbynode(b"mozext", repo, ctx)


@templatekeyword(b"bugs", requires={b"ctx"})
def template_bugs(context, mapping, **args):
    """:bugs: List of ints. The bugs associated with this changeset."""
    ctx = context.resource(mapping, b"ctx")

    bugs = parse_bugs(ctx.description())

    return templateutil.hybridlist(bugs, b"bugs")


@templatekeyword(b"backsoutnodes", requires={b"ctx"})
def template_backsoutnodes(context, mapping):
    ctx = context.resource(mapping, b"ctx")

    description = encoding.fromlocal(ctx.description())
    backouts = parse_backouts(description)
    # return just the nodes, not the bug numbers
    if backouts and backouts[0]:
        return templateutil.hybridlist(backouts[0], b"backouts")


@templatekeyword(b"reviewer", requires={b"ctx"})
def template_reviewer(context, mapping):
    """:reviewer: String. The first reviewer of this changeset."""
    ctx = context.resource(mapping, b"ctx")
    reviewers = parse_reviewers(ctx.description())
    try:
        first_reviewer = next(reviewers)
        return first_reviewer
    except StopIteration:
        return None


@templatekeyword(b"reviewers", requires={b"ctx"})
def template_reviewers(context, mapping):
    """:reviewers: List of strings. The reviewers associated with tis
    changeset."""
    ctx = context.resource(mapping, b"ctx")

    reviewers = parse_reviewers(ctx.description())

    return templateutil.hybridlist(reviewers, b"reviewers")


def _compute_first_version(repo, ctx, what, cache):
    rev = ctx.rev()
    cache_key = "%s_ancestors" % what

    if cache_key not in cache:
        versions = getattr(repo, "_%s_releases" % what)()
        cache[cache_key] = repo._earliest_version_ancestors(versions)

    for version, ancestors in cache[cache_key].items():
        if rev in ancestors:
            return version

    return None


def template_firstrelease(context, mapping):
    """:firstrelease: String. The version of the first release channel
    release with this changeset.
    """
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")
    cache = context.resource(mapping, b"cache")

    return _compute_first_version(repo, ctx, "release", cache)


def template_firstbeta(context, mapping):
    """:firstbeta: String. The version of the first beta release with this
    changeset.
    """
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")
    cache = context.resource(mapping, b"cache")

    return _compute_first_version(repo, ctx, "beta", cache)


def _calculate_push_milestone(repo, ctx, tree):
    # This function appears to be slow. Consider caching results.
    pushes = repo.changetracker.pushes_for_changeset(ctx.node())
    pushes = [p for p in pushes if p[0] == tree]

    if not pushes:
        return None

    push = pushes[0]

    return repo._revision_milestone(pycompat.bytestr(push[4]))


def template_firstnightly(context, mapping):
    """:firstnightly: String. The version of the first nightly release
    with this changeset.
    """
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    return _calculate_push_milestone(repo, ctx, b"central")


def _calculate_next_daily_release(repo, ctx, tree):
    pushes = repo.changetracker.pushes_for_changeset(ctx.node())
    pushes = [p for p in pushes if p[0] == tree]

    if not pushes:
        return None

    push = pushes[0]
    when = push[2]

    dt = datetime.datetime.utcfromtimestamp(when)

    # Daily builds kick off at 3 AM in Mountain View. This is -7 hours
    # from UTC during daylight savings and -8 during regular.
    # Mercurial nor core Python have timezone info built-in, so we
    # hack this calculation together here. This calculation is wrong
    # for date before 2007, when the DST start/end days changed. It
    # may not always be correct in the future. We should use a real
    # timezone database.
    dst_start = None
    dst_end = None

    c = calendar.Calendar(calendar.SUNDAY)
    sunday_count = 0
    for day in c.itermonthdates(dt.year, 3):
        if day.month != 3:
            continue

        if day.weekday() == 6:
            sunday_count += 1
            if sunday_count == 2:
                dst_start = day
                break

    for day in c.itermonthdates(dt.year, 11):
        if day.month != 11:
            if day.weekday() == 6:
                dst_end = day
                break

    dst_start = datetime.datetime(dst_start.year, dst_start.month, dst_start.day, 2)
    dst_end = datetime.datetime(dst_end.year, dst_end.month, dst_end.day, 2)

    is_dst = dt >= dst_start and dt <= dst_end
    utc_offset = 11 if is_dst else 10

    if dt.hour > 3 + utc_offset:
        dt += datetime.timedelta(days=1)

    return pycompat.bytestr(dt.date().isoformat())


def template_nightlydate(context, mapping):
    """:nightlydate: Date information. The date of the first Nightly this
    changeset was likely first active in as a YYYY-MM-DD value.
    """
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    return _calculate_next_daily_release(repo, ctx, b"central")


def template_firstpushuser(context, mapping):
    """:firstpushuser: String. The first person who pushed this changeset."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    pushes = list(repo.changetracker.pushes_for_changeset(ctx.node()))

    if not pushes:
        return None

    return pushes[0][3]


def template_firstpushtree(context, mapping):
    """:firstpushtree: String. The first tree this changeset was pushed to."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    pushes = list(repo.changetracker.pushes_for_changeset(ctx.node()))

    if not pushes:
        return None

    return pushes[0][0]


def template_firstpushtreeherder(context, mapping):
    """:firstpushtreeherder: String. Treeherder URL for the first push of this changeset."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    pushes = list(repo.changetracker.pushes_for_changeset(ctx.node()))
    if not pushes:
        return None

    push = pushes[0]
    tree, node = push[0], push[4]

    return treeherder_url(tree, hex(node))


def template_firstpushdate(context, mapping):
    """:firstpushdate: Date information. The date of the first push of this
    changeset."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    pushes = list(repo.changetracker.pushes_for_changeset(ctx.node()))
    if not pushes:
        return None

    return dateutil.makedate(pushes[0][2])


def template_pushdates(context, mapping):
    """:pushdates: List of date information. The dates this changeset was
    pushed to various trees."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    pushes = repo.changetracker.pushes_for_changeset(ctx.node())
    pushdates = [dateutil.makedate(p[2]) for p in pushes]

    return templateutil.hybridlist(pushdates, b"pushdates")


def template_pushheaddates(context, mapping):
    """:pushheaddates: List of date information. The dates this changeset
    was pushed to various trees as a push head."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    node = ctx.node()
    pushes = repo.changetracker.pushes_for_changeset(ctx.node())
    pushheaddates = [
        dateutil.makedate(p[2]) for p in pushes if pycompat.bytestr(p[4]) == node
    ]

    return templateutil.hybridlist(pushheaddates, b"pushheaddates")


def _trees(repo, ctx):
    """Returns a list of trees the changeset has landed in"""
    return [p[0] for p in repo.changetracker.pushes_for_changeset(ctx.node())]


def template_trees(context, mapping):
    """:trees: List of strings. Trees this changeset has landed in."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    trees = _trees(repo, ctx)

    return templateutil.hybridlist(trees, b"trees")


def template_reltrees(context, mapping):
    """:reltrees: List of strings. Release trees this changeset has landed in."""
    ctx = context.resource(mapping, b"ctx")
    repo = context.resource(mapping, b"repo")

    reltrees = [t for t in _trees(repo, ctx) if t in RELEASE_TREES]

    return templateutil.hybridlist(reltrees, b"reltrees")


# [gps] This function may not be necessary. However, I was unable to figure out
# how to do the equivalent with regular template syntax. Yes, I tried the
# list operator.
@templatefunc(b"dates(VALUES, [fmt, [sep]])")
def template_dates(context, mapping, args):
    """Format a list of dates"""
    if not (1 <= len(args) <= 3):
        raise ParseError(_(b"dates expects one, two, or three arguments"))

    fmt = b"%Y-%m-%d"
    sep = b","

    if len(args) > 1:
        fmt = templatefilters.stringify(args[1][0](context, mapping, args[1][1]))
    if len(args) > 2:
        sep = templatefilters.stringify(args[2][0](context, mapping, args[2][1]))

    return sep.join(
        util.datestr(d, fmt) for d in args[0][0](context, mapping, args[0][1])
    )


def do_backout(
    ui, repo, rev, handle_change, commit_change, reverse_order=False, **opts
):
    if not opts.get("force"):
        ui.status(b"checking for uncommitted changes\n")
        cmdutil.bailifchanged(repo)
    backout = not opts.get("apply")
    desc = {
        "action": b"backout",
        "Actioned": b"Backed out",
        "actioning": b"backing out",
        "name": b"backout",
    }
    if not backout:
        desc = {
            "action": b"apply",
            "Actioned": b"Reapplied",
            "actioning": b"Reapplying",
            "name": b"patch",
        }

    rev = scmutil.revrange(repo, rev)
    if len(rev) == 0:
        raise error.Abort(b"at least one revision required")

    csets = [repo[r] for r in rev]
    csets.sort(reverse=reverse_order, key=lambda cset: cset.rev())

    new_opts = opts.copy()

    def bugs_suffix(bugs):
        if len(bugs) == 0:
            return b""
        elif len(bugs) == 1:
            return b" (bug " + list(bugs)[0] + b")"
        else:
            return b" (" + b", ".join(map(lambda b: b"bug %s" % b, bugs)) + b")"

    def parse_bugs(msg):
        bugs = set()
        m = BUG_CONSERVATIVE_RE.search(msg)
        if m:
            bugs.add(m.group(2))
        return bugs

    def apply_change(node, reverse, push_patch=True, name=None):
        p1, p2 = repo.changelog.parents(node)
        if p2 != nullid:
            raise error.Abort(b"cannot %s a merge changeset" % desc["action"])

        opts = mdiff.defaultopts
        opts.git = True
        rpatch = util.stringio()
        orig, mod = (node, p1) if reverse else (p1, node)
        for chunk in patch.diff(repo, node1=orig, node2=mod, opts=opts):
            rpatch.write(chunk)
        rpatch.seek(0)

        saved_stdin = None
        try:
            save_fin = ui.fin
            ui.fin = rpatch
        except:
            # Old versions of hg did not use the ui.fin mechanism
            saved_stdin = sys.stdin
            sys.stdin = rpatch

        handle_change()

        if saved_stdin is None:
            ui.fin = save_fin
        else:
            sys.stdin = saved_stdin

    allbugs = set()
    messages = []
    for cset in csets:
        # Hunt down original description if we might want to use it
        orig_desc = None
        orig_desc_cset = None
        orig_author = None
        r = cset
        while len(csets) == 1 or not opts.get("single"):
            ui.debug(b"Parsing message for %s\n" % short(r.node()))
            m = backout_re.match(r.description())
            if m:
                ui.debug(b"  looks like a backout of %s\n" % m.group(1))
            else:
                m = reapply_re.match(r.description())
                if m:
                    ui.debug(b"  looks like a reapply of %s\n" % m.group(1))
                else:
                    ui.debug(b"  looks like the original description\n")
                    orig_desc = r.description()
                    orig_desc_cset = r
                    orig_author = r.user()
                    break
            r = scmutil.revsingle(repo, m.group(1))

        bugs = parse_bugs(cset.description())
        allbugs.update(bugs)
        node = cset.node()
        shortnode = short(node)
        ui.status(b"%s %s\n" % (desc["actioning"], shortnode))

        apply_change(node, backout, push_patch=(not opts.get("nopush")))

        msg = (b"%s changeset %s" % (desc["Actioned"], shortnode)) + bugs_suffix(bugs)
        user = None

        if backout:
            # If backing out a backout, reuse the original commit message & author.
            if orig_desc_cset is not None and orig_desc_cset != cset:
                msg = orig_desc
                user = orig_author
        else:
            # If reapplying the original change, reuse the original commit message & author.
            if orig_desc_cset is not None and orig_desc_cset == cset:
                msg = orig_desc
                user = orig_author

        messages.append(msg)
        if not opts.get("single") and not opts.get("nopush"):
            new_opts["message"] = messages[-1]
            # Override the user to that of the original patch author in the case of --apply
            if user is not None:
                new_opts["user"] = user
            commit_change(
                ui,
                repo,
                desc["name"],
                node=node,
                force_name=opts.get("name"),
                **new_opts
            )

        # Iterations of this loop appear to leak memory for unknown reasons.
        # Work around it by forcing a gc.
        gc.collect()

    msg = (
        (b"%s %d changesets" % (desc["Actioned"], len(rev)))
        + bugs_suffix(allbugs)
        + b"\n"
    )
    messages.insert(0, msg)
    new_opts["message"] = b"\n".join(messages)
    if opts.get("single"):

        commit_change(
            ui,
            repo,
            desc["name"],
            revisions=rev,
            force_name=opts.get("name"),
            **new_opts
        )


@command(
    b"oops",
    [
        (b"r", b"rev", [], _(b"revisions to backout")),
        (
            b"s",
            b"single",
            None,
            _(b"fold all backed out changes into a single changeset"),
        ),
        (b"f", b"force", None, _(b"skip check for outstanding uncommitted changes")),
        (b"e", b"edit", None, _(b"edit commit messages")),
        (b"m", b"message", b"", _(b"use text as commit message"), _(b"TEXT")),
        (b"U", b"currentuser", None, _(b'add "From: <current user>" to patch')),
        (b"u", b"user", b"", _(b'add "From: <USER>" to patch'), _(b"USER")),
        (b"D", b"currentdate", None, _(b'add "Date: <current date>" to patch')),
        (b"d", b"date", b"", _(b'add "Date: <DATE>" to patch'), _(b"DATE")),
    ],
    _(b"hg oops -r REVS [-f] [commit options]"),
)
def oops(ui, repo, rev, **opts):
    """backout a change or set of changes

    oops commits a changeset or set of changesets by undoing existing changesets.
    If the -s/--single option is set, then all backed-out changesets
    will be rolled up into a single backout changeset. Otherwise, there will
    be one changeset queued up for each backed-out changeset.

    Note that if you want to reapply a previously backed out patch, use
    hg graft -f.

    Examples:
      hg oops -r 20 -r 30    # backout revisions 20 and 30

      hg oops -r 20+30       # backout revisions 20 and 30

      hg oops -r 20+30:32    # backout revisions 20, 30, 31, and 32

      hg oops -r a3a81775    # the usual revision syntax is available

    See "hg help revisions" and "hg help revsets" for more about specifying
    revisions.
    """

    def handle_change():
        commands.import_(
            ui,
            repo,
            b"-",
            force=True,
            no_commit=True,
            strip=1,
            base=b"",
            prefix=b"",
            obsolete=[],
        )

    def commit_change(
        ui, repo, action, force_name=None, node=None, revisions=None, **opts
    ):
        commands.commit(ui, repo, **opts)

    do_backout(
        ui,
        repo,
        rev,
        handle_change,
        commit_change,
        reverse_order=(not opts.get("apply")),
        **opts
    )


def extsetup(ui):
    extensions.wrapfunction(exchange, "pull", wrappedpull)
    extensions.wrapfunction(exchange, "push", wrappedpush)
    extensions.wrapfunction(exchange, "_pullobsolete", exchangepullpushlog)

    # TRACKING hg64 - `_peerorrepo` is removed, wrap `hg.peer` directly.
    if util.versiontuple() >= (6, 4):
        extensions.wrapfunction(hg, "peer", wrapped_peer)
    else:
        extensions.wrapfunction(hg, "_peerorrepo", wrapped_peerorrepo)

    if ui.configbool(b"mozext", b"disable_local_database"):
        return

    revsetpredicate(b"pushhead([TREE])")(revset_pushhead)
    revsetpredicate(b"tree(X)")(revset_tree)
    revsetpredicate(b"firstpushdate(DATE)")(revset_firstpushdate)
    revsetpredicate(b"firstpushtree(X)")(revset_firstpushtree)
    revsetpredicate(b"pushdate(DATE)")(revset_pushdate)

    templatekeyword(b"firstrelease", requires={b"ctx", b"repo", b"cache"})(
        template_firstrelease
    )
    templatekeyword(b"firstbeta", requires={b"ctx", b"repo", b"cache"})(
        template_firstbeta
    )
    templatekeyword(b"firstnightly", requires={b"ctx", b"repo"})(template_firstnightly)
    templatekeyword(b"nightlydate", requires={b"ctx", b"repo"})(template_nightlydate)
    templatekeyword(b"firstpushuser", requires={b"ctx", b"repo"})(
        template_firstpushuser
    )
    templatekeyword(b"firstpushtree", requires={b"ctx", b"repo"})(
        template_firstpushtree
    )
    templatekeyword(b"firstpushtreeherder", requires={b"ctx", b"repo"})(
        template_firstpushtreeherder
    )
    templatekeyword(b"firstpushdate", requires={b"ctx", b"repo"})(
        template_firstpushdate
    )
    templatekeyword(b"pushdates", requires={b"ctx", b"repo"})(template_pushdates)
    templatekeyword(b"pushheaddates", requires={b"ctx", b"repo"})(
        template_pushheaddates
    )
    templatekeyword(b"trees", requires={b"ctx", b"repo"})(template_trees)
    templatekeyword(b"reltrees", requires={b"ctx", b"repo"})(template_reltrees)


def reposetup(ui, repo):
    """Custom repository implementation.

    Our custom repository class tracks remote tree references so users can
    reference specific revisions on remotes.
    """

    if not repo.local():
        return

    orig_findtags = repo._findtags
    orig_lookup = repo.lookup

    class remotestrackingrepo(repo.__class__):
        @repofilecache(b"remoterefs")
        def remoterefs(self):
            return remoterefs(self)

        @util.propertycache
        def changetracker(self):
            if ui.configbool(b"mozext", b"disable_local_database"):
                return None
            try:
                return ChangeTracker(
                    self.vfs.join(b"changetracker.db"), bytestype=pycompat.bytestr
                )
            except Exception as e:
                raise error.Abort(e.message)

        def _update_remote_refs(self, remote, tree):
            existing_refs = set()
            incoming_refs = set()

            for ref in self.remoterefs:
                if ref.startswith(b"%s/" % tree):
                    existing_refs.add(ref)

            for branch, nodes in remote.branchmap().items():
                # Don't store RELBRANCH refs for non-release trees, as they are
                # meaningless and cruft from yesteryear.
                if branch.endswith(b"RELBRANCH"):
                    if tree not in TREE_ALIASES[b"releases"]:
                        continue

                ref = b"%s/%s" % (tree, branch)
                incoming_refs.add(ref)

                for node in nodes:
                    self.remoterefs[ref] = node

            # Prune old refs.
            for ref in existing_refs - incoming_refs:
                try:
                    del self.remoterefs[ref]
                except KeyError:
                    pass

            with self.wlock():
                self.remoterefs.write()

        def _revision_milestone(self, rev):
            """Look up the Gecko milestone of a revision."""
            fctx = self.filectx(b"config/milestone.txt", changeid=rev)
            lines = fctx.data().splitlines()
            lines = [l for l in lines if not l.startswith(b"#") and l.strip()]

            if not lines:
                return None

            return lines[0]

        def _beta_releases(self):
            """Obtain information for each beta release."""
            return self._release_versions(b"beta/")

        def _release_releases(self):
            return self._release_versions(b"release/")

        def _release_versions(self, prefix):
            d = {}

            for key, node in self.remoterefs.items():
                if not key.startswith(prefix):
                    continue

                key = key[len(prefix) :]

                if not key.startswith(b"GECKO") or not key.endswith(b"RELBRANCH"):
                    continue

                version, date, _relbranch = key.split(b"_")
                version = version[5:]
                after = b""
                marker = b""

                if b"b" in version:
                    marker = b"b"
                    version, after = version.split(b"b")

                if len(version) > 2:
                    major, minor = version[0:2], version[2:]
                else:
                    major, minor = version

                version = b"%s.%s" % (major, minor)
                if marker:
                    version += b"%s%s" % (marker, after)

                d[version] = (key, node, major, minor, marker or None, after or None)

            return d

        def _earliest_version_ancestors(self, versions):
            """Take a set of versions and generate earliest version ancestors.

            This function takes the output of _release_versions as an input
            and calculates the set of revisions corresponding to each version's
            introduced ancestors. Put another way, it returns a dict of version
            to revision set where each set is disjoint and presence in a
            version's set indicates that particular version introduced that
            revision.

            This computation is computational expensive. Callers are encouraged
            to cache it.
            """
            d = {}
            seen = set()
            for version, e in sorted(versions.items()):
                version_rev = self[e[1]].rev()
                ancestors = set(
                    self.changelog.findmissingrevs(common=seen, heads=[version_rev])
                )
                d[version] = ancestors
                seen |= ancestors

            return d

        def reset_bug_database(self):
            if not self.changetracker:
                return

            self.changetracker.wipe_bugs()
            self.sync_bug_database()

        def sync_bug_database(self):
            if not self.changetracker:
                return

            for rev in self:
                ui.makeprogress(b"changeset", rev, total=len(self))
                ctx = self[rev]
                bugs = parse_bugs(ctx.description())
                if bugs:
                    self.changetracker.associate_bugs_with_changeset(bugs, ctx.node())

            ui.makeprogress(b"changeset", None)

    repo.__class__ = remotestrackingrepo

    if ui.configbool(b"mozext", b"reject_pushes_with_repo_names"):
        ui.setconfig(b"hooks", b"prepushkey.reject_repo_names", reject_repo_names_hook)


class DropoffCounter(object):
    """Maintain a mapping from values to counts and weights, where the weight
    drops off exponentially as "time" passes. This is useful when more recent
    contributions should be weighted higher than older ones."""

    Item = namedtuple("Item", ["name", "count", "weight"])

    def __init__(self, factor):
        self.factor = factor
        self.counts = defaultdict(int)
        self.weights = defaultdict(float)
        self.age = 0

    def add(self, value):
        self.counts[value] += 1
        self.weights[value] += pow(self.factor, self.age)

    def advance(self):
        self.age += 1

    def most_weighted(self, n):
        top = sorted(self.weights, key=lambda k: self.weights[k], reverse=True)
        if len(top) > n:
            top = top[:n]
        return [self[key] for key in top]

    def count_values(self):
        """Return number of distinct values stored."""
        return len(self.weights)

    def __getitem__(self, key):
        return DropoffCounter.Item(key, self.counts[key], self.weights[key])


def fullpaths(repo, paths):
    cwd = repo.getcwd()
    return [pathutil.canonpath(repo.root, cwd, path) for path in paths]


def get_logrevs_for_files(ui, repo, files, opts):
    limit = opts["limit"] or 1000000
    wopts = logcmdutil.parseopts(ui, files, {b"follow": True, b"limit": limit})
    revs = getlogrevs(repo, wopts)[0]
    for rev in revs:
        yield rev


def choose_changes(ui, repo, patchfile, opts):
    if opts.get("file"):
        changed_files = fullpaths(repo, opts["file"])
        return (changed_files, b"file", opts["file"])

    if opts.get("dir"):
        changed_files = opts["dir"]  # For --debug printout only
        return (changed_files, b"dir", opts["dir"])

    if opts.get("rev"):
        revs = scmutil.revrange(repo, opts["rev"])
        if not revs:
            raise error.Abort(b"no changes found")
        files_in_revs = set()
        for rev in revs:
            for f in repo[rev].files():
                files_in_revs.add(f)
        changed_files = sorted(files_in_revs)
        return (changed_files, b"rev", opts["rev"])

    changed_files = None
    if patchfile is not None:
        source = None
        if util.safehasattr(patchfile, "getvalue"):
            diff = patchfile.getvalue()
            source = (b"patchdata", None)
        else:
            try:
                diff = url.open(ui, patchfile).read()
                source = (b"patch", patchfile)
            except IOError:
                raise error.Abort(b'Could not find patchfile called "%s"' % patchfile)

    else:
        # try using:
        #  1. current diff (if nonempty)
        #  2. parent of working directory
        ui.pushbuffer()
        commands.diff(ui, repo, git=True)
        diff = ui.popbuffer()
        changed_files = fileRe.findall(diff)
        if len(changed_files) > 0:
            source = (b"current diff", None)
        else:
            changed_files = None
            diff = None

        if diff is None:
            changed_files = sorted(repo[b"."].files())
            source = (b"rev", b".")

    if changed_files is None:
        changed_files = fileRe.findall(diff)

    return (changed_files, source[0], source[1])


def patch_changes(ui, repo, patchfile=None, **opts):
    """Given a patch, look at what files it changes, and map a function over
    the changesets that touch overlapping files.

    Scan through the last LIMIT commits to find the relevant changesets

    The patch may be given as a file or a URL. If no patch is specified,
    the changes in the working directory will be used.

    Alternatively, the -f option may be used to pass in one or more files
    that will be used directly.
    """
    (changedFiles, source, source_info) = choose_changes(ui, repo, patchfile, opts)
    if ui.verbose:
        ui.write(b"Patch source: %s" % source)
        if source_info is not None:
            ui.write(b" %r" % (source_info,))
        ui.write(b"\n")

    if len(changedFiles) == 0:
        ui.write(
            b"Warning: no modified files found in patch. Did you mean to use the -f option?\n"
        )

    if ui.verbose:
        ui.write(b"Using files:\n")
        if len(changedFiles) == 0:
            ui.write(b"  (none)\n")
        else:
            for changedFile in changedFiles:
                ui.write(b"  %s\n" % changedFile)

    # Expand files out to their current full paths
    if opts.get("dir"):
        exact_files = [b"glob:" + opts["dir"] + b"/**"]
    else:
        paths = [p + b"/**" if os.path.isdir(p) else p for p in changedFiles]
        matchfn = scmutil.match(repo[b"."], paths, default=b"relglob")
        exact_files = [b"path:" + path for path in repo[b"."].walk(matchfn)]
        if len(exact_files) == 0:
            return

    for rev in get_logrevs_for_files(ui, repo, exact_files, opts):
        yield repo[rev]


fileRe = re.compile(rb"^\+\+\+ (?:b/)?([^\s]*)", re.MULTILINE)
suckerRe = re.compile(rb"[^s-]r=(\w+)")


@command(
    b"reviewers",
    [
        (b"f", b"file", [], b"see reviewers for FILE", b"FILE"),
        (b"r", b"rev", [], b"see reviewers for revisions", b"REVS"),
        (b"l", b"limit", 200, b"how many revisions back to scan", b"LIMIT"),
    ],
    _(b"hg reviewers [-f FILE1 -f FILE2...] [-r REVS] [-l LIMIT] [PATCH]"),
)
def reviewers(ui, repo, patchfile=None, **opts):
    """Suggest a reviewer for a patch

    Scan through the last LIMIT commits to find candidate reviewers for a
    patch (or set of files).

    The patch may be given as a file or a URL. If no patch is specified,
    the changes in the working directory will be used.

    Alternatively, the -f option may be used to pass in one or more files
    that will be used to infer the reviewers instead.

    The [reviewers] section of your .hgrc may be used to specify reviewer
    aliases in case reviewers are specified multiple ways.

    Written by Blake Winton http://weblog.latte.ca/blake/
    """

    def canon(reviewer):
        reviewer = reviewer.lower()
        return ui.config(b"reviewers", reviewer, reviewer)

    suckers = DropoffCounter(0.95)
    enough_suckers = 100
    for change in patch_changes(ui, repo, patchfile, **opts):
        for raw in suckerRe.findall(change.description()):
            suckers.add(canon(raw))
        if suckers.count_values() >= enough_suckers:
            break
        suckers.advance()

    if suckers.age == 0:
        ui.write(b"no matching files found\n")
        return

    if suckers.count_values() == 0:
        ui.write(b"No reviewers found in range (try higher --limit?)\n")
        return

    reviewers = suckers.most_weighted(5)
    ui.write(b'Ranking reviewers by "frecency"...\n')
    name_column_length = max(
        [len(reviewer.name) for reviewer in reviewers] + [len(b"Reviewer:")]
    )
    ui.write(
        b"%-*s    Recently reviewed commits:\n" % (name_column_length, b"Reviewer:")
    )
    for reviewer in reviewers:
        ui.write(b" %-*s    %d\n" % (name_column_length, reviewer.name, reviewer.count))
    ui.write(b"\n")