gitlab-org--gitlab-foss/app/models/merge_request_diff.rb

584 lines
17 KiB
Ruby
Raw Normal View History

# frozen_string_literal: true
class MergeRequestDiff < ApplicationRecord
include Sortable
include Importable
Use latest_merge_request_diff association Compared to the merge_request_diff association: 1. It's simpler to query. The query uses a foreign key to the merge_request_diffs table, so no ordering is necessary. 2. It's faster for preloading. The merge_request_diff association has to load every diff for the MRs in the set, then discard all but the most recent for each. This association means that Rails can just query for N diffs from N MRs. 3. It's more complicated to update. This is a bidirectional foreign key, so we need to update two tables when adding a diff record. This also means we need to handle this as a special case when importing a GitLab project. There is some juggling with this association in the merge request model: * `MergeRequest#latest_merge_request_diff` is _always_ the latest diff. * `MergeRequest#merge_request_diff` reuses `MergeRequest#latest_merge_request_diff` unless: * Arguments are passed. These are typically to force-reload the association. * It doesn't exist. That means we might be trying to implicitly create a diff. This only seems to happen in specs. * The association is already loaded. This is important for the reasons explained in the comment, which I'll reiterate here: if we a) load a non-latest diff, then b) get its `merge_request`, then c) get that MR's `merge_request_diff`, we should get the diff we loaded in c), even though that's not the latest diff. Basically, `MergeRequest#merge_request_diff` is the latest diff in most cases, but not quite all.
2017-11-15 12:22:18 -05:00
include ManualInverseAssociation
include IgnorableColumn
include EachBatch
include Gitlab::Utils::StrongMemoize
include ObjectStorage::BackgroundMove
# Don't display more than 100 commits at once
COMMITS_SAFE_SIZE = 100
# Applies to closed or merged MRs when determining whether to migrate their
# diffs to external storage
EXTERNAL_DIFF_CUTOFF = 7.days.freeze
belongs_to :merge_request
Use latest_merge_request_diff association Compared to the merge_request_diff association: 1. It's simpler to query. The query uses a foreign key to the merge_request_diffs table, so no ordering is necessary. 2. It's faster for preloading. The merge_request_diff association has to load every diff for the MRs in the set, then discard all but the most recent for each. This association means that Rails can just query for N diffs from N MRs. 3. It's more complicated to update. This is a bidirectional foreign key, so we need to update two tables when adding a diff record. This also means we need to handle this as a special case when importing a GitLab project. There is some juggling with this association in the merge request model: * `MergeRequest#latest_merge_request_diff` is _always_ the latest diff. * `MergeRequest#merge_request_diff` reuses `MergeRequest#latest_merge_request_diff` unless: * Arguments are passed. These are typically to force-reload the association. * It doesn't exist. That means we might be trying to implicitly create a diff. This only seems to happen in specs. * The association is already loaded. This is important for the reasons explained in the comment, which I'll reiterate here: if we a) load a non-latest diff, then b) get its `merge_request`, then c) get that MR's `merge_request_diff`, we should get the diff we loaded in c), even though that's not the latest diff. Basically, `MergeRequest#merge_request_diff` is the latest diff in most cases, but not quite all.
2017-11-15 12:22:18 -05:00
manual_inverse_association :merge_request, :merge_request_diff
has_many :merge_request_diff_files,
-> { order(:merge_request_diff_id, :relative_order) },
inverse_of: :merge_request_diff
has_many :merge_request_diff_commits, -> { order(:merge_request_diff_id, :relative_order) }
validates :base_commit_sha, :head_commit_sha, :start_commit_sha, sha: true
state_machine :state, initial: :empty do
event :clean do
transition any => :without_files
end
state :collected
2016-03-03 12:38:44 -05:00
state :overflow
# Diff files have been deleted by the system
state :without_files
2016-03-03 12:38:44 -05:00
# Deprecated states: these are no longer used but these values may still occur
# in the database.
state :timeout
state :overflow_commits_safe_size
state :overflow_diff_files_limit
state :overflow_diff_lines_limit
end
scope :with_files, -> { without_states(:without_files, :empty) }
scope :viewable, -> { without_state(:empty) }
scope :by_commit_sha, ->(sha) do
joins(:merge_request_diff_commits).where(merge_request_diff_commits: { sha: sha }).reorder(nil)
end
scope :by_project_id, -> (project_id) do
joins(:merge_request).where(merge_requests: { target_project_id: project_id })
end
scope :recent, -> { order(id: :desc).limit(100) }
scope :files_in_database, -> { where(stored_externally: [false, nil]) }
scope :not_latest_diffs, -> do
merge_requests = MergeRequest.arel_table
mr_diffs = arel_table
join_condition = merge_requests[:id].eq(mr_diffs[:merge_request_id])
.and(mr_diffs[:id].not_eq(merge_requests[:latest_merge_request_diff_id]))
arel_join = mr_diffs.join(merge_requests).on(join_condition)
joins(arel_join.join_sources)
end
scope :old_merged_diffs, -> (before) do
merge_requests = MergeRequest.arel_table
mr_metrics = MergeRequest::Metrics.arel_table
mr_diffs = arel_table
mr_join = mr_diffs
.join(merge_requests)
.on(mr_diffs[:merge_request_id].eq(merge_requests[:id]))
metrics_join_condition = mr_diffs[:merge_request_id]
.eq(mr_metrics[:merge_request_id])
.and(mr_metrics[:merged_at].not_eq(nil))
metrics_join = mr_diffs.join(mr_metrics).on(metrics_join_condition)
condition = MergeRequest.arel_table[:state].eq(:merged)
.and(MergeRequest::Metrics.arel_table[:merged_at].lteq(before))
.and(MergeRequest::Metrics.arel_table[:merged_at].not_eq(nil))
joins(metrics_join.join_sources, mr_join.join_sources).where(condition)
end
scope :old_closed_diffs, -> (before) do
condition = MergeRequest.arel_table[:state].eq(:closed)
.and(MergeRequest::Metrics.arel_table[:latest_closed_at].lteq(before))
joins(merge_request: :metrics).where(condition)
end
def self.ids_for_external_storage_migration(limit:)
# No point doing any work unless the feature is enabled
return [] unless Gitlab.config.external_diffs.enabled
case Gitlab.config.external_diffs.when
when 'always'
files_in_database.limit(limit).pluck(:id)
when 'outdated'
# Outdated is too complex to be a single SQL query, so split into three
before = EXTERNAL_DIFF_CUTOFF.ago
ids = files_in_database
.old_merged_diffs(before)
.limit(limit)
.pluck(:id)
return ids if ids.size >= limit
ids += files_in_database
.old_closed_diffs(before)
.limit(limit - ids.size)
.pluck(:id)
return ids if ids.size >= limit
ids + files_in_database
.not_latest_diffs
.limit(limit - ids.size)
.pluck(:id)
else
[]
end
end
mount_uploader :external_diff, ExternalDiffUploader
# All diff information is collected from repository after object is created.
# It allows you to override variables like head_commit_sha before getting diff.
after_create :save_git_content, unless: :importing?
after_save :update_external_diff_store, if: -> { !importing? && saved_change_to_external_diff? }
2017-04-06 18:13:28 -04:00
def self.find_by_diff_refs(diff_refs)
2017-04-06 18:24:51 -04:00
find_by(start_commit_sha: diff_refs.start_sha, head_commit_sha: diff_refs.head_sha, base_commit_sha: diff_refs.base_sha)
2017-04-06 18:13:28 -04:00
end
def viewable?
collected? || without_files? || overflow?
end
# Collect information about commits and diff from repository
# and save it to the database as serialized data
def save_git_content
MergeRequest
.where('id = ? AND COALESCE(latest_merge_request_diff_id, 0) < ?', self.merge_request_id, self.id)
.update_all(latest_merge_request_diff_id: self.id)
ensure_commit_shas
save_commits
save_diffs
# Another set of `after_save` hooks will be called here when we update the record
save
# We need to reset so that dirty tracking is reset when running the original set
# of `after_save` hooks that come after this `after_create` hook. Otherwise, the
# hooks that run when an attribute was changed are run twice.
reset
keep_around_commits
end
def ensure_commit_shas
self.start_commit_sha ||= merge_request.target_branch_sha
self.head_commit_sha ||= merge_request.source_branch_sha
self.base_commit_sha ||= find_base_sha
end
# Override head_commit_sha to keep compatibility with merge request diff
# created before version 8.4 that does not store head_commit_sha in separate db field.
def head_commit_sha
if persisted? && super.nil?
last_commit_sha
else
super
end
end
# This method will rely on repository branch sha
# in case start_commit_sha is nil. Its necesarry for old merge request diff
# created before version 8.4 to work
def safe_start_commit_sha
start_commit_sha || merge_request.target_branch_sha
end
2016-03-03 12:38:44 -05:00
def size
real_size.presence || raw_diffs.size
end
def lines_count
strong_memoize(:lines_count) do
diffs.diff_files.sum(&:line_count)
end
end
def raw_diffs(options = {})
2016-03-03 12:38:44 -05:00
if options[:ignore_whitespace_change]
@diffs_no_whitespace ||= compare.diffs(options)
2016-03-03 12:38:44 -05:00
else
@raw_diffs ||= {}
@raw_diffs[options] ||= load_diffs(options)
2016-03-03 12:38:44 -05:00
end
end
def commits
@commits ||= load_commits
end
def last_commit_sha
commit_shas.first
end
def first_commit
commits.last
end
def base_commit
return unless base_commit_sha
project.commit_by(oid: base_commit_sha)
end
def start_commit
return unless start_commit_sha
project.commit_by(oid: start_commit_sha)
end
def head_commit
return unless head_commit_sha
project.commit_by(oid: head_commit_sha)
end
def commit_shas
merge_request_diff_commits.map(&:sha)
end
def commits_by_shas(shas)
return MergeRequestDiffCommit.none unless shas.present?
merge_request_diff_commits.where(sha: shas)
end
2017-04-10 16:17:47 -04:00
def diff_refs=(new_diff_refs)
2017-04-11 11:51:48 -04:00
self.base_commit_sha = new_diff_refs&.base_sha
self.start_commit_sha = new_diff_refs&.start_sha
self.head_commit_sha = new_diff_refs&.head_sha
2017-04-10 16:17:47 -04:00
end
def diff_refs
return unless start_commit_sha || base_commit_sha
Gitlab::Diff::DiffRefs.new(
base_sha: base_commit_sha,
start_sha: start_commit_sha,
head_sha: head_commit_sha
)
end
# MRs created before 8.4 don't store their true diff refs (start and base),
# but we need to get a commit SHA for the "View file @ ..." link by a file,
2017-05-16 19:58:23 -04:00
# so we use an approximation of the diff refs if we can't get the actual one.
#
# These will not be the actual diff refs if the target branch was merged into
# the source branch after the merge request was created, but it is good enough
# for the specific purpose of linking to a commit.
2017-05-16 19:58:23 -04:00
#
# It is not good enough for highlighting diffs, so we can't simply pass
# these as `diff_refs.`
def fallback_diff_refs
real_refs = diff_refs
return real_refs if real_refs
likely_base_commit_sha = (first_commit&.parent || first_commit)&.sha
Gitlab::Diff::DiffRefs.new(
base_sha: likely_base_commit_sha,
start_sha: safe_start_commit_sha,
head_sha: head_commit_sha
)
end
def diff_refs_by_sha?
base_commit_sha? && head_commit_sha? && start_commit_sha?
end
def diffs(diff_options = nil)
if without_files? && comparison = diff_refs&.compare_in(project)
# It should fetch the repository when diffs are cleaned by the system.
# We don't keep these for storage overload purposes.
# See https://gitlab.com/gitlab-org/gitlab-ce/issues/37639
comparison.diffs(diff_options)
else
diffs_collection(diff_options)
end
end
# Should always return the DB persisted diffs collection
# (e.g. Gitlab::Diff::FileCollection::MergeRequestDiff.
# It's useful when trying to invalidate old caches through
# FileCollection::MergeRequestDiff#clear_cache!
def diffs_collection(diff_options = nil)
Gitlab::Diff::FileCollection::MergeRequestDiff.new(self, diff_options: diff_options)
end
def project
merge_request.target_project
end
def compare
@compare ||=
Gitlab::Git::Compare.new(
repository.raw_repository,
safe_start_commit_sha,
head_commit_sha
)
end
def latest?
Use latest_merge_request_diff association Compared to the merge_request_diff association: 1. It's simpler to query. The query uses a foreign key to the merge_request_diffs table, so no ordering is necessary. 2. It's faster for preloading. The merge_request_diff association has to load every diff for the MRs in the set, then discard all but the most recent for each. This association means that Rails can just query for N diffs from N MRs. 3. It's more complicated to update. This is a bidirectional foreign key, so we need to update two tables when adding a diff record. This also means we need to handle this as a special case when importing a GitLab project. There is some juggling with this association in the merge request model: * `MergeRequest#latest_merge_request_diff` is _always_ the latest diff. * `MergeRequest#merge_request_diff` reuses `MergeRequest#latest_merge_request_diff` unless: * Arguments are passed. These are typically to force-reload the association. * It doesn't exist. That means we might be trying to implicitly create a diff. This only seems to happen in specs. * The association is already loaded. This is important for the reasons explained in the comment, which I'll reiterate here: if we a) load a non-latest diff, then b) get its `merge_request`, then c) get that MR's `merge_request_diff`, we should get the diff we loaded in c), even though that's not the latest diff. Basically, `MergeRequest#merge_request_diff` is the latest diff in most cases, but not quite all.
2017-11-15 12:22:18 -05:00
self.id == merge_request.latest_merge_request_diff_id
end
# rubocop: disable CodeReuse/ServiceClass
def compare_with(sha)
# When compare merge request versions we want diff A..B instead of A...B
# so we handle cases when user does squash and rebase of the commits between versions.
# For this reason we set straight to true by default.
CompareService.new(project, head_commit_sha).execute(project, sha, straight: true)
end
# rubocop: enable CodeReuse/ServiceClass
def modified_paths
strong_memoize(:modified_paths) do
merge_request_diff_files.pluck(:new_path, :old_path).flatten.uniq
end
end
# Carrierwave defines `write_uploader` dynamically on this class, so `super`
# does not work. Alias the carrierwave method so we can call it when needed
alias_method :carrierwave_write_uploader, :write_uploader
# The `external_diff`, `external_diff_store`, and `stored_externally`
# columns were introduced in GitLab 11.8, but some background migration specs
# use factories that rely on current code with an old schema. Without these
# `has_attribute?` guards, they fail with a `MissingAttributeError`.
#
# For more details, see: https://gitlab.com/gitlab-org/gitlab-ce/issues/44990
def write_uploader(column, identifier)
carrierwave_write_uploader(column, identifier) if has_attribute?(column)
end
def update_external_diff_store
update_column(:external_diff_store, external_diff.object_store) if
has_attribute?(:external_diff_store)
end
def saved_change_to_external_diff?
super if has_attribute?(:external_diff)
end
def stored_externally
super if has_attribute?(:stored_externally)
end
alias_method :stored_externally?, :stored_externally
# If enabled, yields the external file containing the diff. Otherwise, yields
# nil. This method is not thread-safe, but it *is* re-entrant, which allows
# multiple merge_request_diff_files to load their data efficiently
def opening_external_diff
return yield(nil) unless stored_externally?
return yield(@external_diff_file) if @external_diff_file
external_diff.open do |file|
2019-03-13 09:42:43 -04:00
@external_diff_file = file
2019-03-13 09:42:43 -04:00
yield(@external_diff_file)
ensure
@external_diff_file = nil
end
end
# Transactionally migrate the current merge_request_diff_files entries to
# external storage. If external storage isn't an option for this diff, the
# method is a no-op.
def migrate_files_to_external_storage!
return if stored_externally? || !use_external_diff? || merge_request_diff_files.count == 0
rows = build_merge_request_diff_files(merge_request_diff_files)
transaction do
MergeRequestDiffFile.where(merge_request_diff_id: id).delete_all
create_merge_request_diff_files(rows)
save!
end
merge_request_diff_files.reset
end
private
def encode_in_base64?(diff_text)
(diff_text.encoding == Encoding::BINARY && !diff_text.ascii_only?) ||
diff_text.include?("\0")
end
def build_external_merge_request_diff_files(rows)
tempfile = build_external_diff_tempfile(rows)
self.external_diff = tempfile
self.stored_externally = true
rows
ensure
tempfile&.unlink
end
def create_merge_request_diff_files(rows)
rows = build_external_merge_request_diff_files(rows) if use_external_diff?
# Faster inserts
Gitlab::Database.bulk_insert('merge_request_diff_files', rows)
end
def build_external_diff_tempfile(rows)
Tempfile.open(external_diff.filename) do |file|
rows.each do |row|
data = row.delete(:diff)
row[:external_diff_offset] = file.pos
row[:external_diff_size] = data.bytesize
file.write(data)
end
file
end
end
def build_merge_request_diff_files(diffs)
diffs.map.with_index do |diff, index|
diff_hash = diff.to_hash.merge(
binary: false,
merge_request_diff_id: self.id,
relative_order: index
)
# Compatibility with old diffs created with Psych.
diff_hash.tap do |hash|
diff_text = hash[:diff]
if encode_in_base64?(diff_text)
hash[:binary] = true
hash[:diff] = [diff_text].pack('m0')
end
end
end
end
def use_external_diff?
return false unless has_attribute?(:external_diff)
return false unless Gitlab.config.external_diffs.enabled
case Gitlab.config.external_diffs.when
when 'always'
true
when 'outdated'
outdated_by_merge? || outdated_by_closure? || old_version?
else
false # Disable external diffs if misconfigured
end
end
def outdated_by_merge?
return false unless merge_request&.metrics&.merged_at
merge_request.merged? && merge_request.metrics.merged_at < EXTERNAL_DIFF_CUTOFF.ago
end
def outdated_by_closure?
return false unless merge_request&.metrics&.latest_closed_at
merge_request.closed? && merge_request.metrics.latest_closed_at < EXTERNAL_DIFF_CUTOFF.ago
end
# We can't rely on `merge_request.latest_merge_request_diff_id` because that
# may have been changed in `save_git_content` without being reflected in
# the association's instance. This query is always subject to races, but
# the worst case is that we *don't* make a diff external when we could. The
# background worker will make it external at a later date.
def old_version?
latest_id = MergeRequest
.where(id: merge_request_id)
.limit(1)
.pluck(:latest_merge_request_diff_id)
.first
self.id != latest_id
end
def load_diffs(options)
# Ensure all diff files operate on the same external diff file instance if
# present. This reduces file open/close overhead.
opening_external_diff do
collection = merge_request_diff_files
if paths = options[:paths]
collection = collection.where('old_path IN (?) OR new_path IN (?)', paths, paths)
end
Gitlab::Git::DiffCollection.new(collection.map(&:to_hash), options)
end
end
def load_commits
commits = merge_request_diff_commits.map { |commit| Commit.from_hash(commit.to_hash, project) }
CommitCollection
.new(merge_request.source_project, commits, merge_request.source_branch)
end
def save_diffs
new_attributes = {}
if compare.commits.size.zero?
new_attributes[:state] = :empty
else
diff_collection = compare.diffs(Commit.max_diff_options)
new_attributes[:real_size] = diff_collection.real_size
2016-03-03 12:38:44 -05:00
if diff_collection.any?
new_attributes[:state] = :collected
rows = build_merge_request_diff_files(diff_collection)
create_merge_request_diff_files(rows)
end
# Set our state to 'overflow' to make the #empty? and #collected?
# methods (generated by StateMachine) return false.
#
# This attribution has to come at the end of the method so 'overflow'
# state does not get overridden by 'collected'.
new_attributes[:state] = :overflow if diff_collection.overflow?
end
assign_attributes(new_attributes)
end
def save_commits
MergeRequestDiffCommit.create_bulk(self.id, compare.commits.reverse)
# merge_request_diff_commits.reset is preferred way to reload associated
# objects but it returns cached result for some reason in this case
# we can circumvent that by specifying that we need an uncached reload
commits = self.class.uncached { merge_request_diff_commits.reset }
self.commits_count = commits.size
end
def repository
project.repository
end
def find_base_sha
return unless head_commit_sha && start_commit_sha
project.merge_base_commit(head_commit_sha, start_commit_sha).try(:sha)
end
def keep_around_commits
[repository, merge_request.source_project.repository].uniq.each do |repo|
repo.keep_around(start_commit_sha, head_commit_sha, base_commit_sha)
end
end
end