1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
|
# frozen_string_literal: true
module Gitlab
module BackgroundMigration
# Migrates author and committer names and emails from
# merge_request_diff_commits to two columns that point to
# merge_request_diff_commit_users.
#
# rubocop: disable Metrics/ClassLength
class MigrateMergeRequestDiffCommitUsers
# The number of user rows in merge_request_diff_commit_users to get in a
# single query.
USER_ROWS_PER_QUERY = 1_000
# The number of rows in merge_request_diff_commits to get in a single
# query.
COMMIT_ROWS_PER_QUERY = 10_000
# The number of rows in merge_request_diff_commits to update in a single
# query.
#
# Tests in staging revealed that increasing the number of updates per
# query translates to a longer total runtime for a migration. For example,
# given the same range of rows to migrate, 1000 updates per query required
# a total of roughly 15 seconds. On the other hand, 5000 updates per query
# required a total of roughly 25 seconds. For this reason, we use a value
# of 1000 rows per update.
UPDATES_PER_QUERY = 1_000
# rubocop: disable Style/Documentation
class MergeRequestDiffCommit < ActiveRecord::Base
include FromUnion
extend ::SuppressCompositePrimaryKeyWarning
self.table_name = 'merge_request_diff_commits'
# Yields each row to migrate in the given range.
#
# This method uses keyset pagination to ensure we don't retrieve
# potentially tens of thousands (or even hundreds of thousands) of rows
# in a single query. Such queries could time out, or increase the amount
# of memory needed to process the data.
#
# We can't use `EachBatch` and similar approaches, as
# merge_request_diff_commits doesn't have a single monotonically
# increasing primary key.
def self.each_row_to_migrate(start_id, stop_id, &block)
order = Pagination::Keyset::Order.build(
%w[merge_request_diff_id relative_order].map do |col|
Pagination::Keyset::ColumnOrderDefinition.new(
attribute_name: col,
order_expression: self.arel_table[col.to_sym].asc,
nullable: :not_nullable,
distinct: false
)
end
)
scope = MergeRequestDiffCommit
.where(merge_request_diff_id: start_id...stop_id)
.order(order)
Pagination::Keyset::Iterator
.new(scope: scope, use_union_optimization: true)
.each_batch(of: COMMIT_ROWS_PER_QUERY) { |rows| rows.each(&block) }
end
end
# rubocop: enable Style/Documentation
# rubocop: disable Style/Documentation
class MergeRequestDiffCommitUser < ActiveRecord::Base
self.table_name = 'merge_request_diff_commit_users'
def self.union(queries)
from("(#{queries.join("\nUNION ALL\n")}) #{table_name}")
end
end
# rubocop: enable Style/Documentation
def perform(start_id, stop_id)
# This Hash maps user names + emails to their corresponding rows in
# merge_request_diff_commit_users.
user_mapping = {}
user_details, diff_rows_to_update = get_data_to_update(start_id, stop_id)
get_user_rows_in_batches(user_details, user_mapping)
create_missing_users(user_details, user_mapping)
update_commit_rows(diff_rows_to_update, user_mapping)
Database::BackgroundMigrationJob.mark_all_as_succeeded(
'MigrateMergeRequestDiffCommitUsers',
[start_id, stop_id]
)
end
# Returns the data we'll use to determine what merge_request_diff_commits
# rows to update, and what data to use for populating their
# commit_author_id and committer_id columns.
def get_data_to_update(start_id, stop_id)
# This Set is used to retrieve users that already exist in
# merge_request_diff_commit_users.
users = Set.new
# This Hash maps the primary key of every row in
# merge_request_diff_commits to the (trimmed) author and committer
# details to use for updating the row.
to_update = {}
MergeRequestDiffCommit.each_row_to_migrate(start_id, stop_id) do |row|
author = [prepare(row.author_name), prepare(row.author_email)]
committer = [prepare(row.committer_name), prepare(row.committer_email)]
to_update[[row.merge_request_diff_id, row.relative_order]] =
[author, committer]
users << author if author[0] || author[1]
users << committer if committer[0] || committer[1]
end
[users, to_update]
end
# Gets any existing rows in merge_request_diff_commit_users in batches.
#
# This method may end up having to retrieve lots of rows. To reduce the
# overhead, we batch queries into a UNION query. We limit the number of
# queries per UNION so we don't end up sending a single query containing
# too many SELECT statements.
def get_user_rows_in_batches(users, user_mapping)
users.each_slice(USER_ROWS_PER_QUERY) do |pairs|
queries = pairs.map do |(name, email)|
MergeRequestDiffCommitUser.where(name: name, email: email).to_sql
end
MergeRequestDiffCommitUser.union(queries).each do |row|
user_mapping[[row.name.to_s, row.email.to_s]] = row
end
end
end
# Creates any users for which no row exists in
# merge_request_diff_commit_users.
#
# Not all users queried may exist yet, so we need to create any missing
# ones; making sure we handle concurrent creations of the same user
def create_missing_users(users, mapping)
create = []
users.each do |(name, email)|
create << { name: name, email: email } unless mapping[[name, email]]
end
return if create.empty?
MergeRequestDiffCommitUser
.insert_all(create, returning: %w[id name email])
.each do |row|
mapping[[row['name'], row['email']]] = MergeRequestDiffCommitUser
.new(id: row['id'], name: row['name'], email: row['email'])
end
# It's possible for (name, email) pairs to be inserted concurrently,
# resulting in the above insert not returning anything. Here we get any
# remaining users that were created concurrently.
get_user_rows_in_batches(
users.reject { |pair| mapping.key?(pair) },
mapping
)
end
# Updates rows in merge_request_diff_commits with their new
# commit_author_id and committer_id values.
def update_commit_rows(to_update, user_mapping)
to_update.each_slice(UPDATES_PER_QUERY) do |slice|
updates = {}
slice.each do |(diff_id, order), (author, committer)|
author_id = user_mapping[author]&.id
committer_id = user_mapping[committer]&.id
updates[[diff_id, order]] = [author_id, committer_id]
end
bulk_update_commit_rows(updates)
end
end
# Bulk updates rows in the merge_request_diff_commits table with their new
# author and/or committer ID values.
#
# Updates are batched together to reduce the overhead of having to produce
# a single UPDATE for every row, as we may end up having to update
# thousands of rows at once.
#
# The query produced by this method is along the lines of the following:
#
# UPDATE merge_request_diff_commits
# SET commit_author_id =
# CASE
# WHEN (merge_request_diff_id, relative_order) = (x, y) THEN X
# WHEN ...
# END,
# committer_id =
# CASE
# WHEN (merge_request_diff_id, relative_order) = (x, y) THEN Y
# WHEN ...
# END
# WHERE (merge_request_diff_id, relative_order) IN ( (x, y), ... )
#
# The `mapping` argument is a Hash in the following format:
#
# { [merge_request_diff_id, relative_order] => [author_id, committer_id] }
#
# rubocop: disable Metrics/AbcSize
def bulk_update_commit_rows(mapping)
author_case = Arel::Nodes::Case.new
committer_case = Arel::Nodes::Case.new
primary_values = []
mapping.each do |diff_id_and_order, (author_id, committer_id)|
primary_value = Arel::Nodes::Grouping.new(diff_id_and_order)
primary_values << primary_value
if author_id
author_case.when(primary_key.eq(primary_value)).then(author_id)
end
if committer_id
committer_case.when(primary_key.eq(primary_value)).then(committer_id)
end
end
if author_case.conditions.empty? && committer_case.conditions.empty?
return
end
fields = []
# Statements such as `SET x = CASE END` are not valid SQL statements, so
# we omit setting an ID field if there are no values to populate it
# with.
if author_case.conditions.any?
fields << [arel_table[:commit_author_id], author_case]
end
if committer_case.conditions.any?
fields << [arel_table[:committer_id], committer_case]
end
query = Arel::UpdateManager.new
.table(arel_table)
.where(primary_key.in(primary_values))
.set(fields)
.to_sql
MergeRequestDiffCommit.connection.execute(query)
end
# rubocop: enable Metrics/AbcSize
def primary_key
Arel::Nodes::Grouping.new(
[arel_table[:merge_request_diff_id], arel_table[:relative_order]]
)
end
def arel_table
MergeRequestDiffCommit.arel_table
end
# Prepares a value to be inserted into a column in the table
# `merge_request_diff_commit_users`. Values in this table are limited to
# 512 characters.
#
# We treat empty strings as NULL values, as there's no point in (for
# example) storing a row where both the name and Email are an empty
# string. In addition, if we treated them differently we could end up with
# two rows: one where field X is NULL, and one where field X is an empty
# string. This is redundant, so we avoid storing such data.
def prepare(value)
value.present? ? value[0..511] : nil
end
end
# rubocop: enable Metrics/ClassLength
end
end
|