debian-mirror-gitlab/lib/gitlab/database.rb
2019-07-07 11:18:12 +05:30

276 lines
7.5 KiB
Ruby

# frozen_string_literal: true
module Gitlab
module Database
# The max value of INTEGER type is the same between MySQL and PostgreSQL:
# https://www.postgresql.org/docs/9.2/static/datatype-numeric.html
# http://dev.mysql.com/doc/refman/5.7/en/integer-types.html
MAX_INT_VALUE = 2147483647
# The max value between MySQL's TIMESTAMP and PostgreSQL's timestampz:
# https://www.postgresql.org/docs/9.1/static/datatype-datetime.html
# https://dev.mysql.com/doc/refman/5.7/en/datetime.html
MAX_TIMESTAMP_VALUE = Time.at((1 << 31) - 1).freeze
def self.config
ActiveRecord::Base.configurations[Rails.env]
end
def self.username
config['username'] || ENV['USER']
end
def self.database_name
config['database']
end
def self.adapter_name
config['adapter']
end
def self.human_adapter_name
postgresql? ? 'PostgreSQL' : 'MySQL'
end
def self.mysql?
adapter_name.casecmp('mysql2').zero?
end
def self.postgresql?
adapter_name.casecmp('postgresql').zero?
end
def self.read_only?
false
end
def self.read_write?
!self.read_only?
end
# Check whether the underlying database is in read-only mode
def self.db_read_only?
if postgresql?
pg_is_in_recovery =
ActiveRecord::Base.connection.execute('SELECT pg_is_in_recovery()')
.first.fetch('pg_is_in_recovery')
Gitlab::Utils.to_boolean(pg_is_in_recovery)
else
false
end
end
def self.db_read_write?
!self.db_read_only?
end
def self.version
@version ||= database_version.match(/\A(?:PostgreSQL |)([^\s]+).*\z/)[1]
end
def self.postgresql_9_or_less?
postgresql? && version.to_f < 10
end
def self.join_lateral_supported?
postgresql? && version.to_f >= 9.3
end
def self.replication_slots_supported?
postgresql? && version.to_f >= 9.4
end
def self.postgresql_minimum_supported_version?
postgresql? && version.to_f >= 9.6
end
# map some of the function names that changed between PostgreSQL 9 and 10
# https://wiki.postgresql.org/wiki/New_in_postgres_10
def self.pg_wal_lsn_diff
Gitlab::Database.postgresql_9_or_less? ? 'pg_xlog_location_diff' : 'pg_wal_lsn_diff'
end
def self.pg_current_wal_insert_lsn
Gitlab::Database.postgresql_9_or_less? ? 'pg_current_xlog_insert_location' : 'pg_current_wal_insert_lsn'
end
def self.pg_last_wal_receive_lsn
Gitlab::Database.postgresql_9_or_less? ? 'pg_last_xlog_receive_location' : 'pg_last_wal_receive_lsn'
end
def self.pg_last_wal_replay_lsn
Gitlab::Database.postgresql_9_or_less? ? 'pg_last_xlog_replay_location' : 'pg_last_wal_replay_lsn'
end
def self.pg_last_xact_replay_timestamp
'pg_last_xact_replay_timestamp'
end
def self.nulls_last_order(field, direction = 'ASC')
order = "#{field} #{direction}"
if postgresql?
order = "#{order} NULLS LAST"
else
# `field IS NULL` will be `0` for non-NULL columns and `1` for NULL
# columns. In the (default) ascending order, `0` comes first.
order = "#{field} IS NULL, #{order}" if direction == 'ASC'
end
order
end
def self.nulls_first_order(field, direction = 'ASC')
order = "#{field} #{direction}"
if postgresql?
order = "#{order} NULLS FIRST"
else
# `field IS NULL` will be `0` for non-NULL columns and `1` for NULL
# columns. In the (default) ascending order, `0` comes first.
order = "#{field} IS NULL, #{order}" if direction == 'DESC'
end
order
end
def self.random
postgresql? ? "RANDOM()" : "RAND()"
end
def self.true_value
if postgresql?
"'t'"
else
1
end
end
def self.false_value
if postgresql?
"'f'"
else
0
end
end
def self.with_connection_pool(pool_size)
pool = create_connection_pool(pool_size)
begin
yield(pool)
ensure
pool.disconnect!
end
end
# Bulk inserts a number of rows into a table, optionally returning their
# IDs.
#
# table - The name of the table to insert the rows into.
# rows - An Array of Hash instances, each mapping the columns to their
# values.
# return_ids - When set to true the return value will be an Array of IDs of
# the inserted rows, this only works on PostgreSQL.
# disable_quote - A key or an Array of keys to exclude from quoting (You
# become responsible for protection from SQL injection for
# these keys!)
def self.bulk_insert(table, rows, return_ids: false, disable_quote: [])
return if rows.empty?
keys = rows.first.keys
columns = keys.map { |key| connection.quote_column_name(key) }
return_ids = false if mysql?
disable_quote = Array(disable_quote).to_set
tuples = rows.map do |row|
keys.map do |k|
disable_quote.include?(k) ? row[k] : connection.quote(row[k])
end
end
sql = <<-EOF
INSERT INTO #{table} (#{columns.join(', ')})
VALUES #{tuples.map { |tuple| "(#{tuple.join(', ')})" }.join(', ')}
EOF
if return_ids
sql = "#{sql}RETURNING id"
end
result = connection.execute(sql)
if return_ids
result.values.map { |tuple| tuple[0].to_i }
else
[]
end
end
def self.sanitize_timestamp(timestamp)
MAX_TIMESTAMP_VALUE > timestamp ? timestamp : MAX_TIMESTAMP_VALUE.dup
end
# pool_size - The size of the DB pool.
# host - An optional host name to use instead of the default one.
def self.create_connection_pool(pool_size, host = nil)
# See activerecord-4.2.7.1/lib/active_record/connection_adapters/connection_specification.rb
env = Rails.env
original_config = ActiveRecord::Base.configurations
env_config = original_config[env].merge('pool' => pool_size)
env_config['host'] = host if host
config = original_config.merge(env => env_config)
spec =
ActiveRecord::
ConnectionAdapters::
ConnectionSpecification::Resolver.new(config).spec(env.to_sym)
ActiveRecord::ConnectionAdapters::ConnectionPool.new(spec)
end
def self.connection
ActiveRecord::Base.connection
end
def self.cached_column_exists?(table_name, column_name)
connection.schema_cache.columns_hash(table_name).has_key?(column_name.to_s)
end
def self.cached_table_exists?(table_name)
connection.schema_cache.data_source_exists?(table_name)
end
private_class_method :connection
def self.database_version
row = connection.execute("SELECT VERSION()").first
if postgresql?
row['version']
else
row.first
end
end
private_class_method :database_version
def self.add_post_migrate_path_to_rails(force: false)
return if ENV['SKIP_POST_DEPLOYMENT_MIGRATIONS'] && !force
Rails.application.config.paths['db'].each do |db_path|
path = Rails.root.join(db_path, 'post_migrate').to_s
unless Rails.application.config.paths['db/migrate'].include? path
Rails.application.config.paths['db/migrate'] << path
# Rails memoizes migrations at certain points where it won't read the above
# path just yet. As such we must also update the following list of paths.
ActiveRecord::Migrator.migrations_paths << path
end
end
end
end
end