rhodecode/model/pull_request.py

# -*- coding: utf-8 -*-

# Copyright (C) 2012-2020 RhodeCode GmbH
#
# This program is free software: you can redistribute it and/or modify
# it under the terms of the GNU Affero General Public License, version 3
# (only), as published by the Free Software Foundation.
#
# 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 Affero General Public License
# along with this program.  If not, see <http://www.gnu.org/licenses/>.
#
# This program is dual-licensed. If you wish to learn more about the
# RhodeCode Enterprise Edition, including its added features, Support services,
# and proprietary license terms, please see https://rhodecode.com/licenses/


"""
pull request model for RhodeCode
"""


import json
import logging
import os

import datetime
import urllib
import collections

from pyramid import compat
from pyramid.threadlocal import get_current_request

from rhodecode.lib.vcs.nodes import FileNode
from rhodecode.translation import lazy_ugettext
from rhodecode.lib import helpers as h, hooks_utils, diffs
from rhodecode.lib import audit_logger
from rhodecode.lib.compat import OrderedDict
from rhodecode.lib.hooks_daemon import prepare_callback_daemon
from rhodecode.lib.markup_renderer import (
    DEFAULT_COMMENTS_RENDERER, RstTemplateRenderer)
from rhodecode.lib.utils2 import (
    safe_unicode, safe_str, md5_safe, AttributeDict, safe_int,
    get_current_rhodecode_user)
from rhodecode.lib.vcs.backends.base import (
    Reference, MergeResponse, MergeFailureReason, UpdateFailureReason,
    TargetRefMissing, SourceRefMissing)
from rhodecode.lib.vcs.conf import settings as vcs_settings
from rhodecode.lib.vcs.exceptions import (
    CommitDoesNotExistError, EmptyRepositoryError)
from rhodecode.model import BaseModel
from rhodecode.model.changeset_status import ChangesetStatusModel
from rhodecode.model.comment import CommentsModel
from rhodecode.model.db import (
    or_, String, cast, PullRequest, PullRequestReviewers, ChangesetStatus,
    PullRequestVersion, ChangesetComment, Repository, RepoReviewRule, User)
from rhodecode.model.meta import Session
from rhodecode.model.notification import NotificationModel, \
    EmailNotificationModel
from rhodecode.model.scm import ScmModel
from rhodecode.model.settings import VcsSettingsModel


log = logging.getLogger(__name__)


# Data structure to hold the response data when updating commits during a pull
# request update.
class UpdateResponse(object):

    def __init__(self, executed, reason, new, old, common_ancestor_id,
                 commit_changes, source_changed, target_changed):

        self.executed = executed
        self.reason = reason
        self.new = new
        self.old = old
        self.common_ancestor_id = common_ancestor_id
        self.changes = commit_changes
        self.source_changed = source_changed
        self.target_changed = target_changed


def get_diff_info(
        source_repo, source_ref, target_repo, target_ref, get_authors=False,
        get_commit_authors=True):
    """
    Calculates detailed diff information for usage in preview of creation of a pull-request.
    This is also used for default reviewers logic
    """

    source_scm = source_repo.scm_instance()
    target_scm = target_repo.scm_instance()

    ancestor_id = target_scm.get_common_ancestor(target_ref, source_ref, source_scm)
    if not ancestor_id:
        raise ValueError(
            'cannot calculate diff info without a common ancestor. '
            'Make sure both repositories are related, and have a common forking commit.')

    # case here is that want a simple diff without incoming commits,
    # previewing what will be merged based only on commits in the source.
    log.debug('Using ancestor %s as source_ref instead of %s',
              ancestor_id, source_ref)

    # source of changes now is the common ancestor
    source_commit = source_scm.get_commit(commit_id=ancestor_id)
    # target commit becomes the source ref as it is the last commit
    # for diff generation this logic gives proper diff
    target_commit = source_scm.get_commit(commit_id=source_ref)

    vcs_diff = \
        source_scm.get_diff(commit1=source_commit, commit2=target_commit,
                            ignore_whitespace=False, context=3)

    diff_processor = diffs.DiffProcessor(
        vcs_diff, format='newdiff', diff_limit=None,
        file_limit=None, show_full_diff=True)

    _parsed = diff_processor.prepare()

    all_files = []
    all_files_changes = []
    changed_lines = {}
    stats = [0, 0]
    for f in _parsed:
        all_files.append(f['filename'])
        all_files_changes.append({
            'filename': f['filename'],
            'stats': f['stats']
        })
        stats[0] += f['stats']['added']
        stats[1] += f['stats']['deleted']

        changed_lines[f['filename']] = []
        if len(f['chunks']) < 2:
            continue
        # first line is "context" information
        for chunks in f['chunks'][1:]:
            for chunk in chunks['lines']:
                if chunk['action'] not in ('del', 'mod'):
                    continue
                changed_lines[f['filename']].append(chunk['old_lineno'])

    commit_authors = []
    user_counts = {}
    email_counts = {}
    author_counts = {}
    _commit_cache = {}

    commits = []
    if get_commit_authors:
        log.debug('Obtaining commit authors from set of commits')
        _compare_data = target_scm.compare(
            target_ref, source_ref, source_scm, merge=True,
            pre_load=["author", "date", "message"]
        )

        for commit in _compare_data:
            # NOTE(marcink): we serialize here, so we don't produce more vcsserver calls on data returned
            # at this function which is later called via JSON serialization
            serialized_commit = dict(
                author=commit.author,
                date=commit.date,
                message=commit.message,
                commit_id=commit.raw_id,
                raw_id=commit.raw_id
            )
            commits.append(serialized_commit)
            user = User.get_from_cs_author(serialized_commit['author'])
            if user and user not in commit_authors:
                commit_authors.append(user)

    # lines
    if get_authors:
        log.debug('Calculating authors of changed files')
        target_commit = source_repo.get_commit(ancestor_id)

        for fname, lines in changed_lines.items():

            try:
                node = target_commit.get_node(fname, pre_load=["is_binary"])
            except Exception:
                log.exception("Failed to load node with path %s", fname)
                continue

            if not isinstance(node, FileNode):
                continue

            # NOTE(marcink): for binary node we don't do annotation, just use last author
            if node.is_binary:
                author = node.last_commit.author
                email = node.last_commit.author_email

                user = User.get_from_cs_author(author)
                if user:
                    user_counts[user.user_id] = user_counts.get(user.user_id, 0) + 1
                author_counts[author] = author_counts.get(author, 0) + 1
                email_counts[email] = email_counts.get(email, 0) + 1

                continue

            for annotation in node.annotate:
                line_no, commit_id, get_commit_func, line_text = annotation
                if line_no in lines:
                    if commit_id not in _commit_cache:
                        _commit_cache[commit_id] = get_commit_func()
                    commit = _commit_cache[commit_id]
                    author = commit.author
                    email = commit.author_email
                    user = User.get_from_cs_author(author)
                    if user:
                        user_counts[user.user_id] = user_counts.get(user.user_id, 0) + 1
                    author_counts[author] = author_counts.get(author, 0) + 1
                    email_counts[email] = email_counts.get(email, 0) + 1

    log.debug('Default reviewers processing finished')

    return {
        'commits': commits,
        'files': all_files_changes,
        'stats': stats,
        'ancestor': ancestor_id,
        # original authors of modified files
        'original_authors': {
            'users': user_counts,
            'authors': author_counts,
            'emails': email_counts,
        },
        'commit_authors': commit_authors
    }


class PullRequestModel(BaseModel):

    cls = PullRequest

    DIFF_CONTEXT = diffs.DEFAULT_CONTEXT

    UPDATE_STATUS_MESSAGES = {
        UpdateFailureReason.NONE: lazy_ugettext(
            'Pull request update successful.'),
        UpdateFailureReason.UNKNOWN: lazy_ugettext(
            'Pull request update failed because of an unknown error.'),
        UpdateFailureReason.NO_CHANGE: lazy_ugettext(
            'No update needed because the source and target have not changed.'),
        UpdateFailureReason.WRONG_REF_TYPE: lazy_ugettext(
            'Pull request cannot be updated because the reference type is '
            'not supported for an update. Only Branch, Tag or Bookmark is allowed.'),
        UpdateFailureReason.MISSING_TARGET_REF: lazy_ugettext(
            'This pull request cannot be updated because the target '
            'reference is missing.'),
        UpdateFailureReason.MISSING_SOURCE_REF: lazy_ugettext(
            'This pull request cannot be updated because the source '
            'reference is missing.'),
    }
    REF_TYPES = ['bookmark', 'book', 'tag', 'branch']
    UPDATABLE_REF_TYPES = ['bookmark', 'book', 'branch']

    def __get_pull_request(self, pull_request):
        return self._get_instance((
            PullRequest, PullRequestVersion), pull_request)

    def _check_perms(self, perms, pull_request, user, api=False):
        if not api:
            return h.HasRepoPermissionAny(*perms)(
                user=user, repo_name=pull_request.target_repo.repo_name)
        else:
            return h.HasRepoPermissionAnyApi(*perms)(
                user=user, repo_name=pull_request.target_repo.repo_name)

    def check_user_read(self, pull_request, user, api=False):
        _perms = ('repository.admin', 'repository.write', 'repository.read',)
        return self._check_perms(_perms, pull_request, user, api)

    def check_user_merge(self, pull_request, user, api=False):
        _perms = ('repository.admin', 'repository.write', 'hg.admin',)
        return self._check_perms(_perms, pull_request, user, api)

    def check_user_update(self, pull_request, user, api=False):
        owner = user.user_id == pull_request.user_id
        return self.check_user_merge(pull_request, user, api) or owner

    def check_user_delete(self, pull_request, user):
        owner = user.user_id == pull_request.user_id
        _perms = ('repository.admin',)
        return self._check_perms(_perms, pull_request, user) or owner

    def is_user_reviewer(self, pull_request, user):
        return user.user_id in [
            x.user_id for x in
            pull_request.get_pull_request_reviewers(PullRequestReviewers.ROLE_REVIEWER)
            if x.user
        ]

    def check_user_change_status(self, pull_request, user, api=False):
        return self.check_user_update(pull_request, user, api) \
               or self.is_user_reviewer(pull_request, user)

    def check_user_comment(self, pull_request, user):
        owner = user.user_id == pull_request.user_id
        return self.check_user_read(pull_request, user) or owner

    def get(self, pull_request):
        return self.__get_pull_request(pull_request)

    def _prepare_get_all_query(self, repo_name, search_q=None, source=False,
                               statuses=None, opened_by=None, order_by=None,
                               order_dir='desc', only_created=False):
        repo = None
        if repo_name:
            repo = self._get_repo(repo_name)

        q = PullRequest.query()

        if search_q:
            like_expression = u'%{}%'.format(safe_unicode(search_q))
            q = q.join(User)
            q = q.filter(or_(
                cast(PullRequest.pull_request_id, String).ilike(like_expression),
                User.username.ilike(like_expression),
                PullRequest.title.ilike(like_expression),
                PullRequest.description.ilike(like_expression),
            ))

        # source or target
        if repo and source:
            q = q.filter(PullRequest.source_repo == repo)
        elif repo:
            q = q.filter(PullRequest.target_repo == repo)

        # closed,opened
        if statuses:
            q = q.filter(PullRequest.status.in_(statuses))

        # opened by filter
        if opened_by:
            q = q.filter(PullRequest.user_id.in_(opened_by))

        # only get those that are in "created" state
        if only_created:
            q = q.filter(PullRequest.pull_request_state == PullRequest.STATE_CREATED)

        if order_by:
            order_map = {
                'name_raw': PullRequest.pull_request_id,
                'id': PullRequest.pull_request_id,
                'title': PullRequest.title,
                'updated_on_raw': PullRequest.updated_on,
                'target_repo': PullRequest.target_repo_id
            }
            if order_dir == 'asc':
                q = q.order_by(order_map[order_by].asc())
            else:
                q = q.order_by(order_map[order_by].desc())

        return q

    def count_all(self, repo_name, search_q=None, source=False, statuses=None,
                  opened_by=None):
        """
        Count the number of pull requests for a specific repository.

        :param repo_name: target or source repo
        :param search_q: filter by text
        :param source: boolean flag to specify if repo_name refers to source
        :param statuses: list of pull request statuses
        :param opened_by: author user of the pull request
        :returns: int number of pull requests
        """
        q = self._prepare_get_all_query(
            repo_name, search_q=search_q, source=source, statuses=statuses,
            opened_by=opened_by)

        return q.count()

    def get_all(self, repo_name, search_q=None, source=False, statuses=None,
                opened_by=None, offset=0, length=None, order_by=None, order_dir='desc'):
        """
        Get all pull requests for a specific repository.

        :param repo_name: target or source repo
        :param search_q: filter by text
        :param source: boolean flag to specify if repo_name refers to source
        :param statuses: list of pull request statuses
        :param opened_by: author user of the pull request
        :param offset: pagination offset
        :param length: length of returned list
        :param order_by: order of the returned list
        :param order_dir: 'asc' or 'desc' ordering direction
        :returns: list of pull requests
        """
        q = self._prepare_get_all_query(
            repo_name, search_q=search_q, source=source, statuses=statuses,
            opened_by=opened_by, order_by=order_by, order_dir=order_dir)

        if length:
            pull_requests = q.limit(length).offset(offset).all()
        else:
            pull_requests = q.all()

        return pull_requests

    def count_awaiting_review(self, repo_name, search_q=None, source=False, statuses=None,
                              opened_by=None):
        """
        Count the number of pull requests for a specific repository that are
        awaiting review.

        :param repo_name: target or source repo
        :param search_q: filter by text
        :param source: boolean flag to specify if repo_name refers to source
        :param statuses: list of pull request statuses
        :param opened_by: author user of the pull request
        :returns: int number of pull requests
        """
        pull_requests = self.get_awaiting_review(
            repo_name, search_q=search_q, source=source, statuses=statuses, opened_by=opened_by)

        return len(pull_requests)

    def get_awaiting_review(self, repo_name, search_q=None, source=False, statuses=None,
                            opened_by=None, offset=0, length=None,
                            order_by=None, order_dir='desc'):
        """
        Get all pull requests for a specific repository that are awaiting
        review.

        :param repo_name: target or source repo
        :param search_q: filter by text
        :param source: boolean flag to specify if repo_name refers to source
        :param statuses: list of pull request statuses
        :param opened_by: author user of the pull request
        :param offset: pagination offset
        :param length: length of returned list
        :param order_by: order of the returned list
        :param order_dir: 'asc' or 'desc' ordering direction
        :returns: list of pull requests
        """
        pull_requests = self.get_all(
            repo_name, search_q=search_q, source=source, statuses=statuses,
            opened_by=opened_by, order_by=order_by, order_dir=order_dir)

        _filtered_pull_requests = []
        for pr in pull_requests:
            status = pr.calculated_review_status()
            if status in [ChangesetStatus.STATUS_NOT_REVIEWED,
                          ChangesetStatus.STATUS_UNDER_REVIEW]:
                _filtered_pull_requests.append(pr)
        if length:
            return _filtered_pull_requests[offset:offset+length]
        else:
            return _filtered_pull_requests

    def count_awaiting_my_review(self, repo_name, search_q=None, source=False, statuses=None,
                                 opened_by=None, user_id=None):
        """
        Count the number of pull requests for a specific repository that are
        awaiting review from a specific user.

        :param repo_name: target or source repo
        :param search_q: filter by text
        :param source: boolean flag to specify if repo_name refers to source
        :param statuses: list of pull request statuses
        :param opened_by: author user of the pull request
        :param user_id: reviewer user of the pull request
        :returns: int number of pull requests
        """
        pull_requests = self.get_awaiting_my_review(
            repo_name, search_q=search_q, source=source, statuses=statuses,
            opened_by=opened_by, user_id=user_id)

        return len(pull_requests)

    def get_awaiting_my_review(self, repo_name, search_q=None, source=False, statuses=None,
                               opened_by=None, user_id=None, offset=0,
                               length=None, order_by=None, order_dir='desc'):
        """
        Get all pull requests for a specific repository that are awaiting
        review from a specific user.

        :param repo_name: target or source repo
        :param search_q: filter by text
        :param source: boolean flag to specify if repo_name refers to source
        :param statuses: list of pull request statuses
        :param opened_by: author user of the pull request
        :param user_id: reviewer user of the pull request
        :param offset: pagination offset
        :param length: length of returned list
        :param order_by: order of the returned list
        :param order_dir: 'asc' or 'desc' ordering direction
        :returns: list of pull requests
        """
        pull_requests = self.get_all(
            repo_name, search_q=search_q, source=source, statuses=statuses,
            opened_by=opened_by, order_by=order_by, order_dir=order_dir)

        _my = PullRequestModel().get_not_reviewed(user_id)
        my_participation = []
        for pr in pull_requests:
            if pr in _my:
                my_participation.append(pr)
        _filtered_pull_requests = my_participation
        if length:
            return _filtered_pull_requests[offset:offset+length]
        else:
            return _filtered_pull_requests

    def get_not_reviewed(self, user_id):
        return [
            x.pull_request for x in PullRequestReviewers.query().filter(
                PullRequestReviewers.user_id == user_id).all()
        ]

    def _prepare_participating_query(self, user_id=None, statuses=None, query='',
                                     order_by=None, order_dir='desc'):
        q = PullRequest.query()
        if user_id:
            reviewers_subquery = Session().query(
                PullRequestReviewers.pull_request_id).filter(
                PullRequestReviewers.user_id == user_id).subquery()
            user_filter = or_(
                PullRequest.user_id == user_id,
                PullRequest.pull_request_id.in_(reviewers_subquery)
            )
            q = PullRequest.query().filter(user_filter)

        # closed,opened
        if statuses:
            q = q.filter(PullRequest.status.in_(statuses))

        if query:
            like_expression = u'%{}%'.format(safe_unicode(query))
            q = q.join(User)
            q = q.filter(or_(
                cast(PullRequest.pull_request_id, String).ilike(like_expression),
                User.username.ilike(like_expression),
                PullRequest.title.ilike(like_expression),
                PullRequest.description.ilike(like_expression),
            ))
        if order_by:
            order_map = {
                'name_raw': PullRequest.pull_request_id,
                'title': PullRequest.title,
                'updated_on_raw': PullRequest.updated_on,
                'target_repo': PullRequest.target_repo_id
            }
            if order_dir == 'asc':
                q = q.order_by(order_map[order_by].asc())
            else:
                q = q.order_by(order_map[order_by].desc())

        return q

    def count_im_participating_in(self, user_id=None, statuses=None, query=''):
        q = self._prepare_participating_query(user_id, statuses=statuses, query=query)
        return q.count()

    def get_im_participating_in(
            self, user_id=None, statuses=None, query='', offset=0,
            length=None, order_by=None, order_dir='desc'):
        """
        Get all Pull requests that i'm participating in, or i have opened
        """

        q = self._prepare_participating_query(
            user_id, statuses=statuses, query=query, order_by=order_by,
            order_dir=order_dir)

        if length:
            pull_requests = q.limit(length).offset(offset).all()
        else:
            pull_requests = q.all()

        return pull_requests

    def get_versions(self, pull_request):
        """
        returns version of pull request sorted by ID descending
        """
        return PullRequestVersion.query()\
            .filter(PullRequestVersion.pull_request == pull_request)\
            .order_by(PullRequestVersion.pull_request_version_id.asc())\
            .all()

    def get_pr_version(self, pull_request_id, version=None):
        at_version = None

        if version and version == 'latest':
            pull_request_ver = PullRequest.get(pull_request_id)
            pull_request_obj = pull_request_ver
            _org_pull_request_obj = pull_request_obj
            at_version = 'latest'
        elif version:
            pull_request_ver = PullRequestVersion.get_or_404(version)
            pull_request_obj = pull_request_ver
            _org_pull_request_obj = pull_request_ver.pull_request
            at_version = pull_request_ver.pull_request_version_id
        else:
            _org_pull_request_obj = pull_request_obj = PullRequest.get_or_404(
                pull_request_id)

        pull_request_display_obj = PullRequest.get_pr_display_object(
            pull_request_obj, _org_pull_request_obj)

        return _org_pull_request_obj, pull_request_obj, \
               pull_request_display_obj, at_version

    def pr_commits_versions(self, versions):
        """
        Maps the pull-request commits into all known PR versions. This way we can obtain
        each pr version the commit was introduced in.
        """
        commit_versions = collections.defaultdict(list)
        num_versions = [x.pull_request_version_id for x in versions]
        for ver in versions:
            for commit_id in ver.revisions:
                ver_idx = ChangesetComment.get_index_from_version(
                    ver.pull_request_version_id, num_versions=num_versions)
                commit_versions[commit_id].append(ver_idx)
        return commit_versions

    def create(self, created_by, source_repo, source_ref, target_repo,
               target_ref, revisions, reviewers, observers, title, description=None,
               common_ancestor_id=None,
               description_renderer=None,
               reviewer_data=None, translator=None, auth_user=None):
        translator = translator or get_current_request().translate

        created_by_user = self._get_user(created_by)
        auth_user = auth_user or created_by_user.AuthUser()
        source_repo = self._get_repo(source_repo)
        target_repo = self._get_repo(target_repo)

        pull_request = PullRequest()
        pull_request.source_repo = source_repo
        pull_request.source_ref = source_ref
        pull_request.target_repo = target_repo
        pull_request.target_ref = target_ref
        pull_request.revisions = revisions
        pull_request.title = title
        pull_request.description = description
        pull_request.description_renderer = description_renderer
        pull_request.author = created_by_user
        pull_request.reviewer_data = reviewer_data
        pull_request.pull_request_state = pull_request.STATE_CREATING
        pull_request.common_ancestor_id = common_ancestor_id

        Session().add(pull_request)
        Session().flush()

        reviewer_ids = set()
        # members / reviewers
        for reviewer_object in reviewers:
            user_id, reasons, mandatory, role, rules = reviewer_object
            user = self._get_user(user_id)

            # skip duplicates
            if user.user_id in reviewer_ids:
                continue

            reviewer_ids.add(user.user_id)

            reviewer = PullRequestReviewers()
            reviewer.user = user
            reviewer.pull_request = pull_request
            reviewer.reasons = reasons
            reviewer.mandatory = mandatory
            reviewer.role = role

            # NOTE(marcink): pick only first rule for now
            rule_id = list(rules)[0] if rules else None
            rule = RepoReviewRule.get(rule_id) if rule_id else None
            if rule:
                review_group = rule.user_group_vote_rule(user_id)
                # we check if this particular reviewer is member of a voting group
                if review_group:
                    # NOTE(marcink):
                    # can be that user is member of more but we pick the first same,
                    # same as default reviewers algo
                    review_group = review_group[0]

                    rule_data = {
                        'rule_name':
                            rule.review_rule_name,
                        'rule_user_group_entry_id':
                            review_group.repo_review_rule_users_group_id,
                        'rule_user_group_name':
                            review_group.users_group.users_group_name,
                        'rule_user_group_members':
                            [x.user.username for x in review_group.users_group.members],
                        'rule_user_group_members_id':
                            [x.user.user_id for x in review_group.users_group.members],
                    }
                    # e.g {'vote_rule': -1, 'mandatory': True}
                    rule_data.update(review_group.rule_data())

                    reviewer.rule_data = rule_data

            Session().add(reviewer)
            Session().flush()

        for observer_object in observers:
            user_id, reasons, mandatory, role, rules = observer_object
            user = self._get_user(user_id)

            # skip duplicates from reviewers
            if user.user_id in reviewer_ids:
                continue

            #reviewer_ids.add(user.user_id)

            observer = PullRequestReviewers()
            observer.user = user
            observer.pull_request = pull_request
            observer.reasons = reasons
            observer.mandatory = mandatory
            observer.role = role

            # NOTE(marcink): pick only first rule for now
            rule_id = list(rules)[0] if rules else None
            rule = RepoReviewRule.get(rule_id) if rule_id else None
            if rule:
                # TODO(marcink): do we need this for observers ??
                pass

            Session().add(observer)
            Session().flush()

        # Set approval status to "Under Review" for all commits which are
        # part of this pull request.
        ChangesetStatusModel().set_status(
            repo=target_repo,
            status=ChangesetStatus.STATUS_UNDER_REVIEW,
            user=created_by_user,
            pull_request=pull_request
        )
        # we commit early at this point. This has to do with a fact
        # that before queries do some row-locking. And because of that
        # we need to commit and finish transaction before below validate call
        # that for large repos could be long resulting in long row locks
        Session().commit()

        # prepare workspace, and run initial merge simulation. Set state during that
        # operation
        pull_request = PullRequest.get(pull_request.pull_request_id)

        # set as merging, for merge simulation, and if finished to created so we mark
        # simulation is working fine
        with pull_request.set_state(PullRequest.STATE_MERGING,
                                    final_state=PullRequest.STATE_CREATED) as state_obj:
            MergeCheck.validate(
                pull_request, auth_user=auth_user, translator=translator)

        self.notify_reviewers(pull_request, reviewer_ids, created_by_user)
        self.trigger_pull_request_hook(pull_request, created_by_user, 'create')

        creation_data = pull_request.get_api_data(with_merge_state=False)
        self._log_audit_action(
            'repo.pull_request.create', {'data': creation_data},
            auth_user, pull_request)

        return pull_request

    def trigger_pull_request_hook(self, pull_request, user, action, data=None):
        pull_request = self.__get_pull_request(pull_request)
        target_scm = pull_request.target_repo.scm_instance()
        if action == 'create':
            trigger_hook = hooks_utils.trigger_create_pull_request_hook
        elif action == 'merge':
            trigger_hook = hooks_utils.trigger_merge_pull_request_hook
        elif action == 'close':
            trigger_hook = hooks_utils.trigger_close_pull_request_hook
        elif action == 'review_status_change':
            trigger_hook = hooks_utils.trigger_review_pull_request_hook
        elif action == 'update':
            trigger_hook = hooks_utils.trigger_update_pull_request_hook
        elif action == 'comment':
            trigger_hook = hooks_utils.trigger_comment_pull_request_hook
        elif action == 'comment_edit':
            trigger_hook = hooks_utils.trigger_comment_pull_request_edit_hook
        else:
            return

        log.debug('Handling pull_request %s trigger_pull_request_hook with action %s and hook: %s',
                  pull_request, action, trigger_hook)
        trigger_hook(
            username=user.username,
            repo_name=pull_request.target_repo.repo_name,
            repo_type=target_scm.alias,
            pull_request=pull_request,
            data=data)

    def _get_commit_ids(self, pull_request):
        """
        Return the commit ids of the merged pull request.

        This method is not dealing correctly yet with the lack of autoupdates
        nor with the implicit target updates.
        For example: if a commit in the source repo is already in the target it
        will be reported anyways.
        """
        merge_rev = pull_request.merge_rev
        if merge_rev is None:
            raise ValueError('This pull request was not merged yet')

        commit_ids = list(pull_request.revisions)
        if merge_rev not in commit_ids:
            commit_ids.append(merge_rev)

        return commit_ids

    def merge_repo(self, pull_request, user, extras):
        log.debug("Merging pull request %s", pull_request.pull_request_id)
        extras['user_agent'] = 'internal-merge'
        merge_state = self._merge_pull_request(pull_request, user, extras)
        if merge_state.executed:
            log.debug("Merge was successful, updating the pull request comments.")
            self._comment_and_close_pr(pull_request, user, merge_state)

            self._log_audit_action(
                'repo.pull_request.merge',
                {'merge_state': merge_state.__dict__},
                user, pull_request)

        else:
            log.warn("Merge failed, not updating the pull request.")
        return merge_state

    def _merge_pull_request(self, pull_request, user, extras, merge_msg=None):
        target_vcs = pull_request.target_repo.scm_instance()
        source_vcs = pull_request.source_repo.scm_instance()

        message = safe_unicode(merge_msg or vcs_settings.MERGE_MESSAGE_TMPL).format(
            pr_id=pull_request.pull_request_id,
            pr_title=pull_request.title,
            source_repo=source_vcs.name,
            source_ref_name=pull_request.source_ref_parts.name,
            target_repo=target_vcs.name,
            target_ref_name=pull_request.target_ref_parts.name,
        )

        workspace_id = self._workspace_id(pull_request)
        repo_id = pull_request.target_repo.repo_id
        use_rebase = self._use_rebase_for_merging(pull_request)
        close_branch = self._close_branch_before_merging(pull_request)
        user_name = self._user_name_for_merging(pull_request, user)

        target_ref = self._refresh_reference(
            pull_request.target_ref_parts, target_vcs)

        callback_daemon, extras = prepare_callback_daemon(
            extras, protocol=vcs_settings.HOOKS_PROTOCOL,
            host=vcs_settings.HOOKS_HOST,
            use_direct_calls=vcs_settings.HOOKS_DIRECT_CALLS)

        with callback_daemon:
            # TODO: johbo: Implement a clean way to run a config_override
            # for a single call.
            target_vcs.config.set(
                'rhodecode', 'RC_SCM_DATA', json.dumps(extras))

            merge_state = target_vcs.merge(
                repo_id, workspace_id, target_ref, source_vcs,
                pull_request.source_ref_parts,
                user_name=user_name, user_email=user.email,
                message=message, use_rebase=use_rebase,
                close_branch=close_branch)
        return merge_state

    def _comment_and_close_pr(self, pull_request, user, merge_state, close_msg=None):
        pull_request.merge_rev = merge_state.merge_ref.commit_id
        pull_request.updated_on = datetime.datetime.now()
        close_msg = close_msg or 'Pull request merged and closed'

        CommentsModel().create(
            text=safe_unicode(close_msg),
            repo=pull_request.target_repo.repo_id,
            user=user.user_id,
            pull_request=pull_request.pull_request_id,
            f_path=None,
            line_no=None,
            closing_pr=True
        )

        Session().add(pull_request)
        Session().flush()
        # TODO: paris: replace invalidation with less radical solution
        ScmModel().mark_for_invalidation(
            pull_request.target_repo.repo_name)
        self.trigger_pull_request_hook(pull_request, user, 'merge')

    def has_valid_update_type(self, pull_request):
        source_ref_type = pull_request.source_ref_parts.type
        return source_ref_type in self.REF_TYPES

    def get_flow_commits(self, pull_request):

        # source repo
        source_ref_name = pull_request.source_ref_parts.name
        source_ref_type = pull_request.source_ref_parts.type
        source_ref_id = pull_request.source_ref_parts.commit_id
        source_repo = pull_request.source_repo.scm_instance()

        try:
            if source_ref_type in self.REF_TYPES:
                source_commit = source_repo.get_commit(source_ref_name)
            else:
                source_commit = source_repo.get_commit(source_ref_id)
        except CommitDoesNotExistError:
            raise SourceRefMissing()

        # target repo
        target_ref_name = pull_request.target_ref_parts.name
        target_ref_type = pull_request.target_ref_parts.type
        target_ref_id = pull_request.target_ref_parts.commit_id
        target_repo = pull_request.target_repo.scm_instance()

        try:
            if target_ref_type in self.REF_TYPES:
                target_commit = target_repo.get_commit(target_ref_name)
            else:
                target_commit = target_repo.get_commit(target_ref_id)
        except CommitDoesNotExistError:
            raise TargetRefMissing()

        return source_commit, target_commit

    def update_commits(self, pull_request, updating_user):
        """
        Get the updated list of commits for the pull request
        and return the new pull request version and the list
        of commits processed by this update action

        updating_user is the user_object who triggered the update
        """
        pull_request = self.__get_pull_request(pull_request)
        source_ref_type = pull_request.source_ref_parts.type
        source_ref_name = pull_request.source_ref_parts.name
        source_ref_id = pull_request.source_ref_parts.commit_id

        target_ref_type = pull_request.target_ref_parts.type
        target_ref_name = pull_request.target_ref_parts.name
        target_ref_id = pull_request.target_ref_parts.commit_id

        if not self.has_valid_update_type(pull_request):
            log.debug("Skipping update of pull request %s due to ref type: %s",
                      pull_request, source_ref_type)
            return UpdateResponse(
                executed=False,
                reason=UpdateFailureReason.WRONG_REF_TYPE,
                old=pull_request, new=None, common_ancestor_id=None, commit_changes=None,
                source_changed=False, target_changed=False)

        try:
            source_commit, target_commit = self.get_flow_commits(pull_request)
        except SourceRefMissing:
            return UpdateResponse(
                executed=False,
                reason=UpdateFailureReason.MISSING_SOURCE_REF,
                old=pull_request, new=None, common_ancestor_id=None, commit_changes=None,
                source_changed=False, target_changed=False)
        except TargetRefMissing:
            return UpdateResponse(
                executed=False,
                reason=UpdateFailureReason.MISSING_TARGET_REF,
                old=pull_request, new=None, common_ancestor_id=None, commit_changes=None,
                source_changed=False, target_changed=False)

        source_changed = source_ref_id != source_commit.raw_id
        target_changed = target_ref_id != target_commit.raw_id

        if not (source_changed or target_changed):
            log.debug("Nothing changed in pull request %s", pull_request)
            return UpdateResponse(
                executed=False,
                reason=UpdateFailureReason.NO_CHANGE,
                old=pull_request, new=None, common_ancestor_id=None, commit_changes=None,
                source_changed=target_changed, target_changed=source_changed)

        change_in_found = 'target repo' if target_changed else 'source repo'
        log.debug('Updating pull request because of change in %s detected',
                  change_in_found)

        # Finally there is a need for an update, in case of source change
        # we create a new version, else just an update
        if source_changed:
            pull_request_version = self._create_version_from_snapshot(pull_request)
            self._link_comments_to_version(pull_request_version)
        else:
            try:
                ver = pull_request.versions[-1]
            except IndexError:
                ver = None

            pull_request.pull_request_version_id = \
                ver.pull_request_version_id if ver else None
            pull_request_version = pull_request

        source_repo = pull_request.source_repo.scm_instance()
        target_repo = pull_request.target_repo.scm_instance()

        # re-compute commit ids
        old_commit_ids = pull_request.revisions
        pre_load = ["author", "date", "message", "branch"]
        commit_ranges = target_repo.compare(
            target_commit.raw_id, source_commit.raw_id, source_repo, merge=True,
            pre_load=pre_load)

        target_ref = target_commit.raw_id
        source_ref = source_commit.raw_id
        ancestor_commit_id = target_repo.get_common_ancestor(
            target_ref, source_ref, source_repo)

        if not ancestor_commit_id:
            raise ValueError(
                'cannot calculate diff info without a common ancestor. '
                'Make sure both repositories are related, and have a common forking commit.')

        pull_request.common_ancestor_id = ancestor_commit_id

        pull_request.source_ref = '%s:%s:%s' % (
            source_ref_type, source_ref_name, source_commit.raw_id)
        pull_request.target_ref = '%s:%s:%s' % (
            target_ref_type, target_ref_name, ancestor_commit_id)

        pull_request.revisions = [
            commit.raw_id for commit in reversed(commit_ranges)]
        pull_request.updated_on = datetime.datetime.now()
        Session().add(pull_request)
        new_commit_ids = pull_request.revisions

        old_diff_data, new_diff_data = self._generate_update_diffs(
            pull_request, pull_request_version)

        # calculate commit and file changes
        commit_changes = self._calculate_commit_id_changes(
            old_commit_ids, new_commit_ids)
        file_changes = self._calculate_file_changes(
            old_diff_data, new_diff_data)

        # set comments as outdated if DIFFS changed
        CommentsModel().outdate_comments(
            pull_request, old_diff_data=old_diff_data,
            new_diff_data=new_diff_data)

        valid_commit_changes = (commit_changes.added or commit_changes.removed)
        file_node_changes = (
            file_changes.added or file_changes.modified or file_changes.removed)
        pr_has_changes = valid_commit_changes or file_node_changes

        # Add an automatic comment to the pull request, in case
        # anything has changed
        if pr_has_changes:
            update_comment = CommentsModel().create(
                text=self._render_update_message(ancestor_commit_id, commit_changes, file_changes),
                repo=pull_request.target_repo,
                user=pull_request.author,
                pull_request=pull_request,
                send_email=False, renderer=DEFAULT_COMMENTS_RENDERER)

            # Update status to "Under Review" for added commits
            for commit_id in commit_changes.added:
                ChangesetStatusModel().set_status(
                    repo=pull_request.source_repo,
                    status=ChangesetStatus.STATUS_UNDER_REVIEW,
                    comment=update_comment,
                    user=pull_request.author,
                    pull_request=pull_request,
                    revision=commit_id)

            # send update email to users
            try:
                self.notify_users(pull_request=pull_request, updating_user=updating_user,
                                  ancestor_commit_id=ancestor_commit_id,
                                  commit_changes=commit_changes,
                                  file_changes=file_changes)
            except Exception:
                log.exception('Failed to send email notification to users')

        log.debug(
            'Updated pull request %s, added_ids: %s, common_ids: %s, '
            'removed_ids: %s', pull_request.pull_request_id,
            commit_changes.added, commit_changes.common, commit_changes.removed)
        log.debug(
            'Updated pull request with the following file changes: %s',
            file_changes)

        log.info(
            "Updated pull request %s from commit %s to commit %s, "
            "stored new version %s of this pull request.",
            pull_request.pull_request_id, source_ref_id,
            pull_request.source_ref_parts.commit_id,
            pull_request_version.pull_request_version_id)
        Session().commit()
        self.trigger_pull_request_hook(pull_request, pull_request.author, 'update')

        return UpdateResponse(
            executed=True, reason=UpdateFailureReason.NONE,
            old=pull_request, new=pull_request_version,
            common_ancestor_id=ancestor_commit_id, commit_changes=commit_changes,
            source_changed=source_changed, target_changed=target_changed)

    def _create_version_from_snapshot(self, pull_request):
        version = PullRequestVersion()
        version.title = pull_request.title
        version.description = pull_request.description
        version.status = pull_request.status
        version.pull_request_state = pull_request.pull_request_state
        version.created_on = datetime.datetime.now()
        version.updated_on = pull_request.updated_on
        version.user_id = pull_request.user_id
        version.source_repo = pull_request.source_repo
        version.source_ref = pull_request.source_ref
        version.target_repo = pull_request.target_repo
        version.target_ref = pull_request.target_ref

        version._last_merge_source_rev = pull_request._last_merge_source_rev
        version._last_merge_target_rev = pull_request._last_merge_target_rev
        version.last_merge_status = pull_request.last_merge_status
        version.last_merge_metadata = pull_request.last_merge_metadata
        version.shadow_merge_ref = pull_request.shadow_merge_ref
        version.merge_rev = pull_request.merge_rev
        version.reviewer_data = pull_request.reviewer_data

        version.revisions = pull_request.revisions
        version.common_ancestor_id = pull_request.common_ancestor_id
        version.pull_request = pull_request
        Session().add(version)
        Session().flush()

        return version

    def _generate_update_diffs(self, pull_request, pull_request_version):

        diff_context = (
            self.DIFF_CONTEXT +
            CommentsModel.needed_extra_diff_context())
        hide_whitespace_changes = False
        source_repo = pull_request_version.source_repo
        source_ref_id = pull_request_version.source_ref_parts.commit_id
        target_ref_id = pull_request_version.target_ref_parts.commit_id
        old_diff = self._get_diff_from_pr_or_version(
            source_repo, source_ref_id, target_ref_id,
            hide_whitespace_changes=hide_whitespace_changes, diff_context=diff_context)

        source_repo = pull_request.source_repo
        source_ref_id = pull_request.source_ref_parts.commit_id
        target_ref_id = pull_request.target_ref_parts.commit_id

        new_diff = self._get_diff_from_pr_or_version(
            source_repo, source_ref_id, target_ref_id,
            hide_whitespace_changes=hide_whitespace_changes, diff_context=diff_context)

        old_diff_data = diffs.DiffProcessor(old_diff)
        old_diff_data.prepare()
        new_diff_data = diffs.DiffProcessor(new_diff)
        new_diff_data.prepare()

        return old_diff_data, new_diff_data

    def _link_comments_to_version(self, pull_request_version):
        """
        Link all unlinked comments of this pull request to the given version.

        :param pull_request_version: The `PullRequestVersion` to which
            the comments shall be linked.

        """
        pull_request = pull_request_version.pull_request
        comments = ChangesetComment.query()\
            .filter(
                # TODO: johbo: Should we query for the repo at all here?
                # Pending decision on how comments of PRs are to be related
                # to either the source repo, the target repo or no repo at all.
                ChangesetComment.repo_id == pull_request.target_repo.repo_id,
                ChangesetComment.pull_request == pull_request,
                ChangesetComment.pull_request_version == None)\
            .order_by(ChangesetComment.comment_id.asc())

        # TODO: johbo: Find out why this breaks if it is done in a bulk
        # operation.
        for comment in comments:
            comment.pull_request_version_id = (
                pull_request_version.pull_request_version_id)
            Session().add(comment)

    def _calculate_commit_id_changes(self, old_ids, new_ids):
        added = [x for x in new_ids if x not in old_ids]
        common = [x for x in new_ids if x in old_ids]
        removed = [x for x in old_ids if x not in new_ids]
        total = new_ids
        return ChangeTuple(added, common, removed, total)

    def _calculate_file_changes(self, old_diff_data, new_diff_data):

        old_files = OrderedDict()
        for diff_data in old_diff_data.parsed_diff:
            old_files[diff_data['filename']] = md5_safe(diff_data['raw_diff'])

        added_files = []
        modified_files = []
        removed_files = []
        for diff_data in new_diff_data.parsed_diff:
            new_filename = diff_data['filename']
            new_hash = md5_safe(diff_data['raw_diff'])

            old_hash = old_files.get(new_filename)
            if not old_hash:
                # file is not present in old diff, we have to figure out from parsed diff
                # operation ADD/REMOVE
                operations_dict = diff_data['stats']['ops']
                if diffs.DEL_FILENODE in operations_dict:
                    removed_files.append(new_filename)
                else:
                    added_files.append(new_filename)
            else:
                if new_hash != old_hash:
                    modified_files.append(new_filename)
                # now remove a file from old, since we have seen it already
                del old_files[new_filename]

        # removed files is when there are present in old, but not in NEW,
        # since we remove old files that are present in new diff, left-overs
        # if any should be the removed files
        removed_files.extend(old_files.keys())

        return FileChangeTuple(added_files, modified_files, removed_files)

    def _render_update_message(self, ancestor_commit_id, changes, file_changes):
        """
        render the message using DEFAULT_COMMENTS_RENDERER (RST renderer),
        so it's always looking the same disregarding on which default
        renderer system is using.

        :param ancestor_commit_id: ancestor raw_id
        :param changes: changes named tuple
        :param file_changes: file changes named tuple

        """
        new_status = ChangesetStatus.get_status_lbl(
            ChangesetStatus.STATUS_UNDER_REVIEW)

        changed_files = (
            file_changes.added + file_changes.modified + file_changes.removed)

        params = {
            'under_review_label': new_status,
            'added_commits': changes.added,
            'removed_commits': changes.removed,
            'changed_files': changed_files,
            'added_files': file_changes.added,
            'modified_files': file_changes.modified,
            'removed_files': file_changes.removed,
            'ancestor_commit_id': ancestor_commit_id
        }
        renderer = RstTemplateRenderer()
        return renderer.render('pull_request_update.mako', **params)

    def edit(self, pull_request, title, description, description_renderer, user):
        pull_request = self.__get_pull_request(pull_request)
        old_data = pull_request.get_api_data(with_merge_state=False)
        if pull_request.is_closed():
            raise ValueError('This pull request is closed')
        if title:
            pull_request.title = title
        pull_request.description = description
        pull_request.updated_on = datetime.datetime.now()
        pull_request.description_renderer = description_renderer
        Session().add(pull_request)
        self._log_audit_action(
            'repo.pull_request.edit', {'old_data': old_data},
            user, pull_request)

    def update_reviewers(self, pull_request, reviewer_data, user):
        """
        Update the reviewers in the pull request

        :param pull_request: the pr to update
        :param reviewer_data: list of tuples
            [(user, ['reason1', 'reason2'], mandatory_flag, role, [rules])]
        :param user: current use who triggers this action
        """

        pull_request = self.__get_pull_request(pull_request)
        if pull_request.is_closed():
            raise ValueError('This pull request is closed')

        reviewers = {}
        for user_id, reasons, mandatory, role, rules in reviewer_data:
            if isinstance(user_id, (int, compat.string_types)):
                user_id = self._get_user(user_id).user_id
            reviewers[user_id] = {
                'reasons': reasons, 'mandatory': mandatory, 'role': role}

        reviewers_ids = set(reviewers.keys())
        current_reviewers = PullRequestReviewers.get_pull_request_reviewers(
            pull_request.pull_request_id, role=PullRequestReviewers.ROLE_REVIEWER)

        current_reviewers_ids = set([x.user.user_id for x in current_reviewers])

        ids_to_add = reviewers_ids.difference(current_reviewers_ids)
        ids_to_remove = current_reviewers_ids.difference(reviewers_ids)

        log.debug("Adding %s reviewers", ids_to_add)
        log.debug("Removing %s reviewers", ids_to_remove)
        changed = False
        added_audit_reviewers = []
        removed_audit_reviewers = []

        for uid in ids_to_add:
            changed = True
            _usr = self._get_user(uid)
            reviewer = PullRequestReviewers()
            reviewer.user = _usr
            reviewer.pull_request = pull_request
            reviewer.reasons = reviewers[uid]['reasons']
            # NOTE(marcink): mandatory shouldn't be changed now
            # reviewer.mandatory = reviewers[uid]['reasons']
            # NOTE(marcink): role should be hardcoded, so we won't edit it.
            reviewer.role = PullRequestReviewers.ROLE_REVIEWER
            Session().add(reviewer)
            added_audit_reviewers.append(reviewer.get_dict())

        for uid in ids_to_remove:
            changed = True
            # NOTE(marcink): we fetch "ALL" reviewers objects using .all().
            # This is an edge case that handles previous state of having the same reviewer twice.
            # this CAN happen due to the lack of DB checks
            reviewers = PullRequestReviewers.query()\
                .filter(PullRequestReviewers.user_id == uid,
                        PullRequestReviewers.role == PullRequestReviewers.ROLE_REVIEWER,
                        PullRequestReviewers.pull_request == pull_request)\
                .all()

            for obj in reviewers:
                added_audit_reviewers.append(obj.get_dict())
                Session().delete(obj)

        if changed:
            Session().expire_all()
            pull_request.updated_on = datetime.datetime.now()
            Session().add(pull_request)

        # finally store audit logs
        for user_data in added_audit_reviewers:
            self._log_audit_action(
                'repo.pull_request.reviewer.add', {'data': user_data},
                user, pull_request)
        for user_data in removed_audit_reviewers:
            self._log_audit_action(
                'repo.pull_request.reviewer.delete', {'old_data': user_data},
                user, pull_request)

        self.notify_reviewers(pull_request, ids_to_add, user)
        return ids_to_add, ids_to_remove

    def update_observers(self, pull_request, observer_data, user):
        """
        Update the observers in the pull request

        :param pull_request: the pr to update
        :param observer_data: list of tuples
            [(user, ['reason1', 'reason2'], mandatory_flag, role, [rules])]
        :param user: current use who triggers this action
        """
        pull_request = self.__get_pull_request(pull_request)
        if pull_request.is_closed():
            raise ValueError('This pull request is closed')

        observers = {}
        for user_id, reasons, mandatory, role, rules in observer_data:
            if isinstance(user_id, (int, compat.string_types)):
                user_id = self._get_user(user_id).user_id
            observers[user_id] = {
                'reasons': reasons, 'observers': mandatory, 'role': role}

        observers_ids = set(observers.keys())
        current_observers = PullRequestReviewers.get_pull_request_reviewers(
            pull_request.pull_request_id, role=PullRequestReviewers.ROLE_OBSERVER)

        current_observers_ids = set([x.user.user_id for x in current_observers])

        ids_to_add = observers_ids.difference(current_observers_ids)
        ids_to_remove = current_observers_ids.difference(observers_ids)

        log.debug("Adding %s observer", ids_to_add)
        log.debug("Removing %s observer", ids_to_remove)
        changed = False
        added_audit_observers = []
        removed_audit_observers = []

        for uid in ids_to_add:
            changed = True
            _usr = self._get_user(uid)
            observer = PullRequestReviewers()
            observer.user = _usr
            observer.pull_request = pull_request
            observer.reasons = observers[uid]['reasons']
            # NOTE(marcink): mandatory shouldn't be changed now
            # observer.mandatory = observer[uid]['reasons']

            # NOTE(marcink): role should be hardcoded, so we won't edit it.
            observer.role = PullRequestReviewers.ROLE_OBSERVER
            Session().add(observer)
            added_audit_observers.append(observer.get_dict())

        for uid in ids_to_remove:
            changed = True
            # NOTE(marcink): we fetch "ALL" reviewers objects using .all().
            # This is an edge case that handles previous state of having the same reviewer twice.
            # this CAN happen due to the lack of DB checks
            observers = PullRequestReviewers.query()\
                .filter(PullRequestReviewers.user_id == uid,
                        PullRequestReviewers.role == PullRequestReviewers.ROLE_OBSERVER,
                        PullRequestReviewers.pull_request == pull_request)\
                .all()

            for obj in observers:
                added_audit_observers.append(obj.get_dict())
                Session().delete(obj)

        if changed:
            Session().expire_all()
            pull_request.updated_on = datetime.datetime.now()
            Session().add(pull_request)

        # finally store audit logs
        for user_data in added_audit_observers:
            self._log_audit_action(
                'repo.pull_request.observer.add', {'data': user_data},
                user, pull_request)
        for user_data in removed_audit_observers:
            self._log_audit_action(
                'repo.pull_request.observer.delete', {'old_data': user_data},
                user, pull_request)

        self.notify_observers(pull_request, ids_to_add, user)
        return ids_to_add, ids_to_remove

    def get_url(self, pull_request, request=None, permalink=False):
        if not request:
            request = get_current_request()

        if permalink:
            return request.route_url(
                'pull_requests_global',
                pull_request_id=pull_request.pull_request_id,)
        else:
            return request.route_url('pullrequest_show',
                repo_name=safe_str(pull_request.target_repo.repo_name),
                pull_request_id=pull_request.pull_request_id,)

    def get_shadow_clone_url(self, pull_request, request=None):
        """
        Returns qualified url pointing to the shadow repository. If this pull
        request is closed there is no shadow repository and ``None`` will be
        returned.
        """
        if pull_request.is_closed():
            return None
        else:
            pr_url = urllib.unquote(self.get_url(pull_request, request=request))
            return safe_unicode('{pr_url}/repository'.format(pr_url=pr_url))

    def _notify_reviewers(self, pull_request, user_ids, role, user):
        # notification to reviewers/observers
        if not user_ids:
            return

        log.debug('Notify following %s users about pull-request %s', role, user_ids)

        pull_request_obj = pull_request
        # get the current participants of this pull request
        recipients = user_ids
        notification_type = EmailNotificationModel.TYPE_PULL_REQUEST

        pr_source_repo = pull_request_obj.source_repo
        pr_target_repo = pull_request_obj.target_repo

        pr_url = h.route_url('pullrequest_show',
                             repo_name=pr_target_repo.repo_name,
                             pull_request_id=pull_request_obj.pull_request_id,)

        # set some variables for email notification
        pr_target_repo_url = h.route_url(
            'repo_summary', repo_name=pr_target_repo.repo_name)

        pr_source_repo_url = h.route_url(
            'repo_summary', repo_name=pr_source_repo.repo_name)

        # pull request specifics
        pull_request_commits = [
            (x.raw_id, x.message)
            for x in map(pr_source_repo.get_commit, pull_request.revisions)]

        current_rhodecode_user = user
        kwargs = {
            'user': current_rhodecode_user,
            'pull_request_author': pull_request.author,
            'pull_request': pull_request_obj,
            'pull_request_commits': pull_request_commits,

            'pull_request_target_repo': pr_target_repo,
            'pull_request_target_repo_url': pr_target_repo_url,

            'pull_request_source_repo': pr_source_repo,
            'pull_request_source_repo_url': pr_source_repo_url,

            'pull_request_url': pr_url,
            'thread_ids': [pr_url],
            'user_role': role
        }

        # create notification objects, and emails
        NotificationModel().create(
            created_by=current_rhodecode_user,
            notification_subject='',  # Filled in based on the notification_type
            notification_body='',  # Filled in based on the notification_type
            notification_type=notification_type,
            recipients=recipients,
            email_kwargs=kwargs,
        )

    def notify_reviewers(self, pull_request, reviewers_ids, user):
        return self._notify_reviewers(pull_request, reviewers_ids,
                                      PullRequestReviewers.ROLE_REVIEWER, user)

    def notify_observers(self, pull_request, observers_ids, user):
        return self._notify_reviewers(pull_request, observers_ids,
                                      PullRequestReviewers.ROLE_OBSERVER, user)

    def notify_users(self, pull_request, updating_user, ancestor_commit_id,
                     commit_changes, file_changes):

        updating_user_id = updating_user.user_id
        reviewers = set([x.user.user_id for x in pull_request.get_pull_request_reviewers()])
        # NOTE(marcink): send notification to all other users except to
        # person who updated the PR
        recipients = reviewers.difference(set([updating_user_id]))

        log.debug('Notify following recipients about pull-request update %s', recipients)

        pull_request_obj = pull_request

        # send email about the update
        changed_files = (
                file_changes.added + file_changes.modified + file_changes.removed)

        pr_source_repo = pull_request_obj.source_repo
        pr_target_repo = pull_request_obj.target_repo

        pr_url = h.route_url('pullrequest_show',
                             repo_name=pr_target_repo.repo_name,
                             pull_request_id=pull_request_obj.pull_request_id,)

        # set some variables for email notification
        pr_target_repo_url = h.route_url(
            'repo_summary', repo_name=pr_target_repo.repo_name)

        pr_source_repo_url = h.route_url(
            'repo_summary', repo_name=pr_source_repo.repo_name)

        email_kwargs = {
            'date': datetime.datetime.now(),
            'updating_user': updating_user,

            'pull_request': pull_request_obj,

            'pull_request_target_repo': pr_target_repo,
            'pull_request_target_repo_url': pr_target_repo_url,

            'pull_request_source_repo': pr_source_repo,
            'pull_request_source_repo_url': pr_source_repo_url,

            'pull_request_url': pr_url,

            'ancestor_commit_id': ancestor_commit_id,
            'added_commits': commit_changes.added,
            'removed_commits': commit_changes.removed,
            'changed_files': changed_files,
            'added_files': file_changes.added,
            'modified_files': file_changes.modified,
            'removed_files': file_changes.removed,
            'thread_ids': [pr_url],
        }

        # create notification objects, and emails
        NotificationModel().create(
            created_by=updating_user,
            notification_subject='',  # Filled in based on the notification_type
            notification_body='',  # Filled in based on the notification_type
            notification_type=EmailNotificationModel.TYPE_PULL_REQUEST_UPDATE,
            recipients=recipients,
            email_kwargs=email_kwargs,
        )

    def delete(self, pull_request, user=None):
        if not user:
            user = getattr(get_current_rhodecode_user(), 'username', None)

        pull_request = self.__get_pull_request(pull_request)
        old_data = pull_request.get_api_data(with_merge_state=False)
        self._cleanup_merge_workspace(pull_request)
        self._log_audit_action(
            'repo.pull_request.delete', {'old_data': old_data},
            user, pull_request)
        Session().delete(pull_request)

    def close_pull_request(self, pull_request, user):
        pull_request = self.__get_pull_request(pull_request)
        self._cleanup_merge_workspace(pull_request)
        pull_request.status = PullRequest.STATUS_CLOSED
        pull_request.updated_on = datetime.datetime.now()
        Session().add(pull_request)
        self.trigger_pull_request_hook(pull_request, pull_request.author, 'close')

        pr_data = pull_request.get_api_data(with_merge_state=False)
        self._log_audit_action(
            'repo.pull_request.close', {'data': pr_data}, user, pull_request)

    def close_pull_request_with_comment(
            self, pull_request, user, repo, message=None, auth_user=None):

        pull_request_review_status = pull_request.calculated_review_status()

        if pull_request_review_status == ChangesetStatus.STATUS_APPROVED:
            # approved only if we have voting consent
            status = ChangesetStatus.STATUS_APPROVED
        else:
            status = ChangesetStatus.STATUS_REJECTED
        status_lbl = ChangesetStatus.get_status_lbl(status)

        default_message = (
            'Closing with status change {transition_icon} {status}.'
        ).format(transition_icon='>', status=status_lbl)
        text = message or default_message

        # create a comment, and link it to new status
        comment = CommentsModel().create(
            text=text,
            repo=repo.repo_id,
            user=user.user_id,
            pull_request=pull_request.pull_request_id,
            status_change=status_lbl,
            status_change_type=status,
            closing_pr=True,
            auth_user=auth_user,
        )

        # calculate old status before we change it
        old_calculated_status = pull_request.calculated_review_status()
        ChangesetStatusModel().set_status(
            repo.repo_id,
            status,
            user.user_id,
            comment=comment,
            pull_request=pull_request.pull_request_id
        )

        Session().flush()

        self.trigger_pull_request_hook(pull_request, user, 'comment',
                                       data={'comment': comment})

        # we now calculate the status of pull request again, and based on that
        # calculation trigger status change. This might happen in cases
        # that non-reviewer admin closes a pr, which means his vote doesn't
        # change the status, while if he's a reviewer this might change it.
        calculated_status = pull_request.calculated_review_status()
        if old_calculated_status != calculated_status:
            self.trigger_pull_request_hook(pull_request, user, 'review_status_change',
                                           data={'status': calculated_status})

        # finally close the PR
        PullRequestModel().close_pull_request(pull_request.pull_request_id, user)

        return comment, status

    def merge_status(self, pull_request, translator=None, force_shadow_repo_refresh=False):
        _ = translator or get_current_request().translate

        if not self._is_merge_enabled(pull_request):
            return None, False, _('Server-side pull request merging is disabled.')

        if pull_request.is_closed():
            return None, False, _('This pull request is closed.')

        merge_possible, msg = self._check_repo_requirements(
            target=pull_request.target_repo, source=pull_request.source_repo,
            translator=_)
        if not merge_possible:
            return None, merge_possible, msg

        try:
            merge_response = self._try_merge(
                pull_request, force_shadow_repo_refresh=force_shadow_repo_refresh)
            log.debug("Merge response: %s", merge_response)
            return merge_response, merge_response.possible, merge_response.merge_status_message
        except NotImplementedError:
            return None, False, _('Pull request merging is not supported.')

    def _check_repo_requirements(self, target, source, translator):
        """
        Check if `target` and `source` have compatible requirements.

        Currently this is just checking for largefiles.
        """
        _ = translator
        target_has_largefiles = self._has_largefiles(target)
        source_has_largefiles = self._has_largefiles(source)
        merge_possible = True
        message = u''

        if target_has_largefiles != source_has_largefiles:
            merge_possible = False
            if source_has_largefiles:
                message = _(
                    'Target repository large files support is disabled.')
            else:
                message = _(
                    'Source repository large files support is disabled.')

        return merge_possible, message

    def _has_largefiles(self, repo):
        largefiles_ui = VcsSettingsModel(repo=repo).get_ui_settings(
            'extensions', 'largefiles')
        return largefiles_ui and largefiles_ui[0].active

    def _try_merge(self, pull_request, force_shadow_repo_refresh=False):
        """
        Try to merge the pull request and return the merge status.
        """
        log.debug(
            "Trying out if the pull request %s can be merged. Force_refresh=%s",
            pull_request.pull_request_id, force_shadow_repo_refresh)
        target_vcs = pull_request.target_repo.scm_instance()
        # Refresh the target reference.
        try:
            target_ref = self._refresh_reference(
                pull_request.target_ref_parts, target_vcs)
        except CommitDoesNotExistError:
            merge_state = MergeResponse(
                False, False, None, MergeFailureReason.MISSING_TARGET_REF,
                metadata={'target_ref': pull_request.target_ref_parts})
            return merge_state

        target_locked = pull_request.target_repo.locked
        if target_locked and target_locked[0]:
            locked_by = 'user:{}'.format(target_locked[0])
            log.debug("The target repository is locked by %s.", locked_by)
            merge_state = MergeResponse(
                False, False, None, MergeFailureReason.TARGET_IS_LOCKED,
                metadata={'locked_by': locked_by})
        elif force_shadow_repo_refresh or self._needs_merge_state_refresh(
                pull_request, target_ref):
            log.debug("Refreshing the merge status of the repository.")
            merge_state = self._refresh_merge_state(
                pull_request, target_vcs, target_ref)
        else:
            possible = pull_request.last_merge_status == MergeFailureReason.NONE
            metadata = {
                'unresolved_files': '',
                'target_ref': pull_request.target_ref_parts,
                'source_ref': pull_request.source_ref_parts,
            }
            if pull_request.last_merge_metadata:
                metadata.update(pull_request.last_merge_metadata_parsed)

            if not possible and target_ref.type == 'branch':
                # NOTE(marcink): case for mercurial multiple heads on branch
                heads = target_vcs._heads(target_ref.name)
                if len(heads) != 1:
                    heads = '\n,'.join(target_vcs._heads(target_ref.name))
                    metadata.update({
                        'heads': heads
                    })

            merge_state = MergeResponse(
                possible, False, None, pull_request.last_merge_status, metadata=metadata)

        return merge_state

    def _refresh_reference(self, reference, vcs_repository):
        if reference.type in self.UPDATABLE_REF_TYPES:
            name_or_id = reference.name
        else:
            name_or_id = reference.commit_id

        refreshed_commit = vcs_repository.get_commit(name_or_id)
        refreshed_reference = Reference(
            reference.type, reference.name, refreshed_commit.raw_id)
        return refreshed_reference

    def _needs_merge_state_refresh(self, pull_request, target_reference):
        return not(
            pull_request.revisions and
            pull_request.revisions[0] == pull_request._last_merge_source_rev and
            target_reference.commit_id == pull_request._last_merge_target_rev)

    def _refresh_merge_state(self, pull_request, target_vcs, target_reference):
        workspace_id = self._workspace_id(pull_request)
        source_vcs = pull_request.source_repo.scm_instance()
        repo_id = pull_request.target_repo.repo_id
        use_rebase = self._use_rebase_for_merging(pull_request)
        close_branch = self._close_branch_before_merging(pull_request)
        merge_state = target_vcs.merge(
            repo_id, workspace_id,
            target_reference, source_vcs, pull_request.source_ref_parts,
            dry_run=True, use_rebase=use_rebase,
            close_branch=close_branch)

        # Do not store the response if there was an unknown error.
        if merge_state.failure_reason != MergeFailureReason.UNKNOWN:
            pull_request._last_merge_source_rev = \
                pull_request.source_ref_parts.commit_id
            pull_request._last_merge_target_rev = target_reference.commit_id
            pull_request.last_merge_status = merge_state.failure_reason
            pull_request.last_merge_metadata = merge_state.metadata

            pull_request.shadow_merge_ref = merge_state.merge_ref
            Session().add(pull_request)
            Session().commit()

        return merge_state

    def _workspace_id(self, pull_request):
        workspace_id = 'pr-%s' % pull_request.pull_request_id
        return workspace_id

    def generate_repo_data(self, repo, commit_id=None, branch=None,
                           bookmark=None, translator=None):
        from rhodecode.model.repo import RepoModel

        all_refs, selected_ref = \
            self._get_repo_pullrequest_sources(
                repo.scm_instance(), commit_id=commit_id,
                branch=branch, bookmark=bookmark, translator=translator)

        refs_select2 = []
        for element in all_refs:
            children = [{'id': x[0], 'text': x[1]} for x in element[0]]
            refs_select2.append({'text': element[1], 'children': children})

        return {
            'user': {
                'user_id': repo.user.user_id,
                'username': repo.user.username,
                'firstname': repo.user.first_name,
                'lastname': repo.user.last_name,
                'gravatar_link': h.gravatar_url(repo.user.email, 14),
            },
            'name': repo.repo_name,
            'link': RepoModel().get_url(repo),
            'description': h.chop_at_smart(repo.description_safe, '\n'),
            'refs': {
                'all_refs': all_refs,
                'selected_ref': selected_ref,
                'select2_refs': refs_select2
            }
        }

    def generate_pullrequest_title(self, source, source_ref, target):
        return u'{source}#{at_ref} to {target}'.format(
            source=source,
            at_ref=source_ref,
            target=target,
        )

    def _cleanup_merge_workspace(self, pull_request):
        # Merging related cleanup
        repo_id = pull_request.target_repo.repo_id
        target_scm = pull_request.target_repo.scm_instance()
        workspace_id = self._workspace_id(pull_request)

        try:
            target_scm.cleanup_merge_workspace(repo_id, workspace_id)
        except NotImplementedError:
            pass

    def _get_repo_pullrequest_sources(
            self, repo, commit_id=None, branch=None, bookmark=None,
            translator=None):
        """
        Return a structure with repo's interesting commits, suitable for
        the selectors in pullrequest controller

        :param commit_id: a commit that must be in the list somehow
            and selected by default
        :param branch: a branch that must be in the list and selected
            by default - even if closed
        :param bookmark: a bookmark that must be in the list and selected
        """
        _ = translator or get_current_request().translate

        commit_id = safe_str(commit_id) if commit_id else None
        branch = safe_unicode(branch) if branch else None
        bookmark = safe_unicode(bookmark) if bookmark else None

        selected = None

        # order matters: first source that has commit_id in it will be selected
        sources = []
        sources.append(('book', repo.bookmarks.items(), _('Bookmarks'), bookmark))
        sources.append(('branch', repo.branches.items(), _('Branches'), branch))

        if commit_id:
            ref_commit = (h.short_id(commit_id), commit_id)
            sources.append(('rev', [ref_commit], _('Commit IDs'), commit_id))

        sources.append(
            ('branch', repo.branches_closed.items(), _('Closed Branches'), branch),
        )

        groups = []

        for group_key, ref_list, group_name, match in sources:
            group_refs = []
            for ref_name, ref_id in ref_list:
                ref_key = u'{}:{}:{}'.format(group_key, ref_name, ref_id)
                group_refs.append((ref_key, ref_name))

                if not selected:
                    if set([commit_id, match]) & set([ref_id, ref_name]):
                        selected = ref_key

            if group_refs:
                groups.append((group_refs, group_name))

        if not selected:
            ref = commit_id or branch or bookmark
            if ref:
                raise CommitDoesNotExistError(
                    u'No commit refs could be found matching: {}'.format(ref))
            elif repo.DEFAULT_BRANCH_NAME in repo.branches:
                selected = u'branch:{}:{}'.format(
                    safe_unicode(repo.DEFAULT_BRANCH_NAME),
                    safe_unicode(repo.branches[repo.DEFAULT_BRANCH_NAME])
                )
            elif repo.commit_ids:
                # make the user select in this case
                selected = None
            else:
                raise EmptyRepositoryError()
        return groups, selected

    def get_diff(self, source_repo, source_ref_id, target_ref_id,
                 hide_whitespace_changes, diff_context):

        return self._get_diff_from_pr_or_version(
            source_repo, source_ref_id, target_ref_id,
            hide_whitespace_changes=hide_whitespace_changes, diff_context=diff_context)

    def _get_diff_from_pr_or_version(
            self, source_repo, source_ref_id, target_ref_id,
            hide_whitespace_changes, diff_context):

        target_commit = source_repo.get_commit(
            commit_id=safe_str(target_ref_id))
        source_commit = source_repo.get_commit(
            commit_id=safe_str(source_ref_id), maybe_unreachable=True)
        if isinstance(source_repo, Repository):
            vcs_repo = source_repo.scm_instance()
        else:
            vcs_repo = source_repo

        # TODO: johbo: In the context of an update, we cannot reach
        # the old commit anymore with our normal mechanisms. It needs
        # some sort of special support in the vcs layer to avoid this
        # workaround.
        if (source_commit.raw_id == vcs_repo.EMPTY_COMMIT_ID and
                vcs_repo.alias == 'git'):
            source_commit.raw_id = safe_str(source_ref_id)

        log.debug('calculating diff between '
                  'source_ref:%s and target_ref:%s for repo `%s`',
                  target_ref_id, source_ref_id,
                  safe_unicode(vcs_repo.path))

        vcs_diff = vcs_repo.get_diff(
            commit1=target_commit, commit2=source_commit,
            ignore_whitespace=hide_whitespace_changes, context=diff_context)
        return vcs_diff

    def _is_merge_enabled(self, pull_request):
        return self._get_general_setting(
            pull_request, 'rhodecode_pr_merge_enabled')

    def _use_rebase_for_merging(self, pull_request):
        repo_type = pull_request.target_repo.repo_type
        if repo_type == 'hg':
            return self._get_general_setting(
                pull_request, 'rhodecode_hg_use_rebase_for_merging')
        elif repo_type == 'git':
            return self._get_general_setting(
                pull_request, 'rhodecode_git_use_rebase_for_merging')

        return False

    def _user_name_for_merging(self, pull_request, user):
        env_user_name_attr = os.environ.get('RC_MERGE_USER_NAME_ATTR', '')
        if env_user_name_attr and hasattr(user, env_user_name_attr):
            user_name_attr = env_user_name_attr
        else:
            user_name_attr = 'short_contact'

        user_name = getattr(user, user_name_attr)
        return user_name

    def _close_branch_before_merging(self, pull_request):
        repo_type = pull_request.target_repo.repo_type
        if repo_type == 'hg':
            return self._get_general_setting(
                pull_request, 'rhodecode_hg_close_branch_before_merging')
        elif repo_type == 'git':
            return self._get_general_setting(
                pull_request, 'rhodecode_git_close_branch_before_merging')

        return False

    def _get_general_setting(self, pull_request, settings_key, default=False):
        settings_model = VcsSettingsModel(repo=pull_request.target_repo)
        settings = settings_model.get_general_settings()
        return settings.get(settings_key, default)

    def _log_audit_action(self, action, action_data, user, pull_request):
        audit_logger.store(
            action=action,
            action_data=action_data,
            user=user,
            repo=pull_request.target_repo)

    def get_reviewer_functions(self):
        """
        Fetches functions for validation and fetching default reviewers.
        If available we use the EE package, else we fallback to CE
        package functions
        """
        try:
            from rc_reviewers.utils import get_default_reviewers_data
            from rc_reviewers.utils import validate_default_reviewers
            from rc_reviewers.utils import validate_observers
        except ImportError:
            from rhodecode.apps.repository.utils import get_default_reviewers_data
            from rhodecode.apps.repository.utils import validate_default_reviewers
            from rhodecode.apps.repository.utils import validate_observers

        return get_default_reviewers_data, validate_default_reviewers, validate_observers


class MergeCheck(object):
    """
    Perform Merge Checks and returns a check object which stores information
    about merge errors, and merge conditions
    """
    TODO_CHECK = 'todo'
    PERM_CHECK = 'perm'
    REVIEW_CHECK = 'review'
    MERGE_CHECK = 'merge'
    WIP_CHECK = 'wip'

    def __init__(self):
        self.review_status = None
        self.merge_possible = None
        self.merge_msg = ''
        self.merge_response = None
        self.failed = None
        self.errors = []
        self.error_details = OrderedDict()
        self.source_commit = AttributeDict()
        self.target_commit = AttributeDict()
        self.reviewers_count = 0
        self.observers_count = 0

    def __repr__(self):
        return '<MergeCheck(possible:{}, failed:{}, errors:{})>'.format(
            self.merge_possible, self.failed, self.errors)

    def push_error(self, error_type, message, error_key, details):
        self.failed = True
        self.errors.append([error_type, message])
        self.error_details[error_key] = dict(
            details=details,
            error_type=error_type,
            message=message
        )

    @classmethod
    def validate(cls, pull_request, auth_user, translator, fail_early=False,
                 force_shadow_repo_refresh=False):
        _ = translator
        merge_check = cls()

        # title has WIP:
        if pull_request.work_in_progress:
            log.debug("MergeCheck: cannot merge, title has wip: marker.")

            msg = _('WIP marker in title prevents from accidental merge.')
            merge_check.push_error('error', msg, cls.WIP_CHECK, pull_request.title)
            if fail_early:
                return merge_check

        # permissions to merge
        user_allowed_to_merge = PullRequestModel().check_user_merge(pull_request, auth_user)
        if not user_allowed_to_merge:
            log.debug("MergeCheck: cannot merge, approval is pending.")

            msg = _('User `{}` not allowed to perform merge.').format(auth_user.username)
            merge_check.push_error('error', msg, cls.PERM_CHECK, auth_user.username)
            if fail_early:
                return merge_check

        # permission to merge into the target branch
        target_commit_id = pull_request.target_ref_parts.commit_id
        if pull_request.target_ref_parts.type == 'branch':
            branch_name = pull_request.target_ref_parts.name
        else:
            # for mercurial we can always figure out the branch from the commit
            # in case of bookmark
            target_commit = pull_request.target_repo.get_commit(target_commit_id)
            branch_name = target_commit.branch

        rule, branch_perm = auth_user.get_rule_and_branch_permission(
            pull_request.target_repo.repo_name, branch_name)
        if branch_perm and branch_perm == 'branch.none':
            msg = _('Target branch `{}` changes rejected by rule {}.').format(
                branch_name, rule)
            merge_check.push_error('error', msg, cls.PERM_CHECK, auth_user.username)
            if fail_early:
                return merge_check

        # review status, must be always present
        review_status = pull_request.calculated_review_status()
        merge_check.review_status = review_status
        merge_check.reviewers_count = pull_request.reviewers_count
        merge_check.observers_count = pull_request.observers_count

        status_approved = review_status == ChangesetStatus.STATUS_APPROVED
        if not status_approved and merge_check.reviewers_count:
            log.debug("MergeCheck: cannot merge, approval is pending.")
            msg = _('Pull request reviewer approval is pending.')

            merge_check.push_error('warning', msg, cls.REVIEW_CHECK, review_status)

            if fail_early:
                return merge_check

        # left over TODOs
        todos = CommentsModel().get_pull_request_unresolved_todos(pull_request)
        if todos:
            log.debug("MergeCheck: cannot merge, {} "
                      "unresolved TODOs left.".format(len(todos)))

            if len(todos) == 1:
                msg = _('Cannot merge, {} TODO still not resolved.').format(
                    len(todos))
            else:
                msg = _('Cannot merge, {} TODOs still not resolved.').format(
                    len(todos))

            merge_check.push_error('warning', msg, cls.TODO_CHECK, todos)

            if fail_early:
                return merge_check

        # merge possible, here is the filesystem simulation + shadow repo
        merge_response, merge_status, msg = PullRequestModel().merge_status(
            pull_request, translator=translator,
            force_shadow_repo_refresh=force_shadow_repo_refresh)

        merge_check.merge_possible = merge_status
        merge_check.merge_msg = msg
        merge_check.merge_response = merge_response

        source_ref_id = pull_request.source_ref_parts.commit_id
        target_ref_id = pull_request.target_ref_parts.commit_id

        try:
            source_commit, target_commit = PullRequestModel().get_flow_commits(pull_request)
            merge_check.source_commit.changed = source_ref_id != source_commit.raw_id
            merge_check.source_commit.ref_spec = pull_request.source_ref_parts
            merge_check.source_commit.current_raw_id = source_commit.raw_id
            merge_check.source_commit.previous_raw_id = source_ref_id

            merge_check.target_commit.changed = target_ref_id != target_commit.raw_id
            merge_check.target_commit.ref_spec = pull_request.target_ref_parts
            merge_check.target_commit.current_raw_id = target_commit.raw_id
            merge_check.target_commit.previous_raw_id = target_ref_id
        except (SourceRefMissing, TargetRefMissing):
            pass

        if not merge_status:
            log.debug("MergeCheck: cannot merge, pull request merge not possible.")
            merge_check.push_error('warning', msg, cls.MERGE_CHECK, None)

            if fail_early:
                return merge_check

        log.debug('MergeCheck: is failed: %s', merge_check.failed)
        return merge_check

    @classmethod
    def get_merge_conditions(cls, pull_request, translator):
        _ = translator
        merge_details = {}

        model = PullRequestModel()
        use_rebase = model._use_rebase_for_merging(pull_request)

        if use_rebase:
            merge_details['merge_strategy'] = dict(
                details={},
                message=_('Merge strategy: rebase')
            )
        else:
            merge_details['merge_strategy'] = dict(
                details={},
                message=_('Merge strategy: explicit merge commit')
            )

        close_branch = model._close_branch_before_merging(pull_request)
        if close_branch:
            repo_type = pull_request.target_repo.repo_type
            close_msg = ''
            if repo_type == 'hg':
                close_msg = _('Source branch will be closed before the merge.')
            elif repo_type == 'git':
                close_msg = _('Source branch will be deleted after the merge.')

            merge_details['close_branch'] = dict(
                details={},
                message=close_msg
            )

        return merge_details


ChangeTuple = collections.namedtuple(
    'ChangeTuple', ['added', 'common', 'removed', 'total'])

FileChangeTuple = collections.namedtuple(
    'FileChangeTuple', ['added', 'modified', 'removed'])