# File lib/arjdbc/jdbc/adapter.rb, line 375 def primary_keys(table) @connection.primary_keys(table) end
class ActiveRecord::ConnectionAdapters::JdbcAdapter
Built on top of `ActiveRecord::ConnectionAdapters::AbstractAdapter` which provides the abstract interface for database-specific functionality, this class serves 2 purposes in AR-JDBC :
-
as a base class for sub-classes
-
usable standalone (or with a mixed in adapter spec module)
Historically this class is mostly been used standalone and that's still a valid use-case esp. since (with it's `arjdbc.jdbc.RubyJdbcConnectionClass`) JDBC provides a unified interface for all databases in Java it tries to do it's best implementing all `ActiveRecord` functionality on top of that. This might no be perfect that's why it checks for a `config` module (or tries to resolve one from the JDBC driver's meta-data) and if the database has “extended” AR-JDBC support mixes in the given module for each adapter instance. This is sufficient for most database specific specs we support, but for compatibility with native (MRI) adapters it's perfectly fine to sub-class the adapter and override some of its API methods.
Constants
- ADAPTER_NAME
Attributes
Public Class Methods
@deprecated re-implemented - no longer used @return [Hash] the AREL visitor to use If there's a `self.arel2_visitors(config)` method on the adapter spec than it is preferred and will be used instead of this one.
# File lib/arjdbc/jdbc/adapter.rb, line 165 def self.arel2_visitors(config) { 'jdbc' => ::Arel::Visitors::ToSql } end
@deprecated re-implemented - no longer used @see arel2_visitors
# File lib/arjdbc/jdbc/adapter.rb, line 171 def self.configure_arel2_visitors(config) visitors = ::Arel::Visitors::VISITORS klass = config[:adapter_spec] klass = self unless klass.respond_to?(:arel2_visitors) visitor = nil klass.arel2_visitors(config).each do |name, arel| visitors[name] = ( visitor = arel ) end if visitor && config[:adapter] =~ /^(jdbc|jndi)$/ visitors[ config[:adapter] ] = visitor end visitor end
ArJdbc::Abstract::Core::new
# File lib/arjdbc/jdbc/adapter.rb, line 48 def self.new(connection, logger = nil, pool = nil) adapter = super Jdbc::JndiConnectionPoolCallbacks.prepare(adapter, adapter.instance_variable_get(:@connection)) adapter end
Initializes the (JDBC connection) adapter instance. The passed configuration Hash's keys are symbolized, thus changes to the original `config` keys won't be reflected in the adapter. If the adapter's sub-class or the spec module that this instance will extend in responds to `configure_connection` than it will be called. @param connection an (optional) connection instance @param logger the `ActiveRecord::Base.logger` to use (or nil) @param config the database configuration @note `initialize(logger, config)` with 2 arguments is supported as well
ArJdbc::Abstract::Core::new
# File lib/arjdbc/jdbc/adapter.rb, line 63 def initialize(connection, logger = nil, config = nil) @config = config.respond_to?(:symbolize_keys) ? config.symbolize_keys : config # FIXME: Rails 5 defaults to prepared statements on and we do not seem # to work yet. So default to off unless it is requested until that is # fixed. @config[:prepared_statements] = false if !@config[:prepared_statements] # NOTE: JDBC 4.0 drivers support checking if connection isValid # thus no need to @config[:connection_alive_sql] ||= 'SELECT 1' # # NOTE: setup to retry 5-times previously - maybe do not set at all ? @config[:retry_count] ||= 1 @config[:adapter_spec] = adapter_spec(@config) unless @config.key?(:adapter_spec) spec = @config[:adapter_spec] super(connection, logger, @config) # kind of like `extend ArJdbc::MyDB if self.class == JdbcAdapter` : klass = @config[:adapter_class] extend spec if spec && ( ! klass || klass == JdbcAdapter) end
Protected Class Methods
@return whether the given SQL string is an 'INSERT' query
# File lib/arjdbc/jdbc/adapter.rb, line 596 def self.insert?(sql) JdbcConnection::insert?(sql) end
Allows changing the prepared statements setting for this connection. @see #prepared_statements?
def prepared_statements
=(statements)
@prepared_statements = statements
end
# File lib/arjdbc/jdbc/adapter.rb, line 542 def self.prepared_statements?(config) config.key?(:prepared_statements) ? type_cast_config_to_boolean(config.fetch(:prepared_statements)) : false # off by default - NOTE: on AR 4.x it's on by default !? end
@return whether the given SQL string is a 'SELECT' like query (returning a result set)
# File lib/arjdbc/jdbc/adapter.rb, line 591 def self.select?(sql) JdbcConnection::select?(sql) end
# File lib/arjdbc/jdbc/adapter.rb, line 554 def self.suble_binds=(flag); @@suble_binds = flag; end
# File lib/arjdbc/jdbc/adapter.rb, line 553 def self.suble_binds?; @@suble_binds; end
@private
# File lib/arjdbc/jdbc/adapter.rb, line 615 def self.type_cast_config_to_boolean(config) config == 'false' ? false : (config == 'true' ? true : config) end
@private
# File lib/arjdbc/jdbc/adapter.rb, line 608 def self.type_cast_config_to_integer(config) config =~ /\A\d+\z/ ? config.to_i : config end
@return whether the given SQL string is an 'UPDATE' (or 'DELETE') query
# File lib/arjdbc/jdbc/adapter.rb, line 601 def self.update?(sql) ! select?(sql) && ! insert?(sql) end
Public Instance Methods
@note Used by Java API to convert dates from (custom) SELECTs (might get refactored). @private
# File lib/arjdbc/jdbc/adapter.rb, line 623 def _string_to_date(value); jdbc_column_class.string_to_date(value) end
@note Used by Java API to convert times from (custom) SELECTs (might get refactored). @private
# File lib/arjdbc/jdbc/adapter.rb, line 627 def _string_to_time(value); jdbc_column_class.string_to_dummy_time(value) end
@note Used by Java API to convert times from (custom) SELECTs (might get refactored). @private
# File lib/arjdbc/jdbc/adapter.rb, line 631 def _string_to_timestamp(value); jdbc_column_class.string_to_time(value) end
@return [String] the 'JDBC' adapter name.
# File lib/arjdbc/jdbc/adapter.rb, line 140 def adapter_name ADAPTER_NAME end
Locate the specialized (database specific) adapter specification module if one exists based on provided configuration data. This module will than extend an instance of the adapter (unless an `:adapter_class` provided).
This method is called during {#initialize} unless an explicit `config` is set. @param config the configuration to check for `:adapter_spec` @return [Module] the database specific module
# File lib/arjdbc/jdbc/adapter.rb, line 110 def adapter_spec(config) dialect = (config[:dialect] || config[:driver]).to_s ::ArJdbc.modules.each do |constant| # e.g. ArJdbc::MySQL if constant.respond_to?(:adapter_matcher) spec = constant.adapter_matcher(dialect, config) return spec if spec end end if (config[:jndi] || config[:data_source]) && ! config[:dialect] begin data_source = config[:data_source] || Java::JavaxNaming::InitialContext.new.lookup(config[:jndi]) connection = data_source.getConnection config[:dialect] = connection.getMetaData.getDatabaseProductName rescue Java::JavaSql::SQLException => e warn "failed to set database :dialect from connection meda-data (#{e})" else return adapter_spec(config) # re-try matching a spec with set config[:dialect] ensure connection.close if connection # return to the pool end end nil end
# File lib/arjdbc/jdbc/adapter.rb, line 264 def columns(table_name, name = nil) @connection.columns(table_name.to_s) end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 360 def data_source_exists?(name) table_exists?(name) end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 355 def data_sources tables end
Returns the underlying database name. @override
# File lib/arjdbc/jdbc/adapter.rb, line 224 def database_name @connection.database_name end
Similar to {#exec_query} except it returns “raw” results in an array where each rows is a hash with keys as columns (just like Rails used to do up until 3.0) instead of wrapping them in a {#ActiveRecord::Result}. @param sql the query string (or AREL object) @param name logging marker for the executed SQL statement log entry @param binds the bind parameters @yield [v1, v2] depending on the row values returned from the query In case a block is given it will yield each row from the result set instead of returning mapped query results in an array. @return [Array] unless a block is given
# File lib/arjdbc/jdbc/adapter.rb, line 283 def exec_query_raw(sql, name = 'SQL', binds = [], &block) if sql.respond_to?(:to_sql) sql = to_sql(sql, binds); to_sql = true end if prepared_statements? log(sql, name, binds) { @connection.execute_query_raw(sql, binds, &block) } else sql = suble_binds(sql, binds) unless to_sql # deprecated behavior log(sql, name) { @connection.execute_query_raw(sql, &block) } end end
Executes the SQL statement in the context of this connection. The return value from this method depends on the SQL type (whether it's a SELECT, INSERT etc.). For INSERTs a generated id might get returned while for UPDATE statements the affected row count. Please note that this method returns “raw” results (in an array) for statements that return a result set, while {#exec_query} is expected to return a `ActiveRecord::Result` (since AR 3.1). @note This method does not use prepared statements. @note The method does not emulate various “native” `execute` results on MRI. @see exec_query
@see exec_insert @see exec_update
# File lib/arjdbc/jdbc/adapter.rb, line 313 def execute(sql, name = nil, binds = nil) sql = suble_binds to_sql(sql, binds), binds if binds if name == :skip_logging _execute(sql, name) else log(sql, name) { _execute(sql, name) } end end
Kind of `execute(sql) rescue nil` but logging failures at debug level only.
# File lib/arjdbc/jdbc/adapter.rb, line 333 def execute_quietly(sql, name = 'SQL') log(sql, name) do begin _execute(sql) rescue => e logger.debug("#{e.class}: #{e.message}: #{sql}") end end end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 380 def foreign_keys(table_name) @connection.foreign_keys(table_name) end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 365 def indexes(table_name, name = nil, schema_name = nil) @connection.indexes(table_name, name, schema_name) end
@override Will return true even when native adapter classes passed in e.g. `jdbc_adapter.is_a? ConnectionAdapter::PostgresqlAdapter`
This is only necessary (for built-in adapters) when `config` is forced to `nil` and the `:adapter_spec` module is used to extend the `JdbcAdapter`, otherwise we replace the class constants for built-in adapters (MySQL, PostgreSQL
and SQLite3).
# File lib/arjdbc/jdbc/adapter.rb, line 152 def is_a?(klass) # This is to fake out current_adapter? conditional logic in AR tests if klass.is_a?(Class) && klass.name =~ /#{adapter_name}Adapter$/i true else super end end
Returns the (JDBC) `ActiveRecord` column class for this adapter. This is used by (database specific) spec modules to override the class. @see ActiveRecord::ConnectionAdapters::JdbcColumn
# File lib/arjdbc/jdbc/adapter.rb, line 98 def jdbc_column_class ::ActiveRecord::ConnectionAdapters::JdbcColumn end
Returns the (JDBC) connection class to be used for this adapter. This is used by (database specific) spec modules to override the class used assuming some of the available methods have been re-defined. @see ActiveRecord::ConnectionAdapters::JdbcConnection
# File lib/arjdbc/jdbc/adapter.rb, line 90 def jdbc_connection_class(spec) connection_class = spec.jdbc_connection_class if spec && spec.respond_to?(:jdbc_connection_class) connection_class ? connection_class : ::ActiveRecord::ConnectionAdapters::JdbcConnection end
Allows for modification of the detected native types. @param types the resolved native database types @see native_database_types
# File lib/arjdbc/jdbc/adapter.rb, line 205 def modify_types(types) types end
DB specific types are detected but adapter specs (or extenders) are expected to hand tune these types for concrete databases. @return [Hash] the native database types @override
# File lib/arjdbc/jdbc/adapter.rb, line 189 def native_database_types @native_database_types ||= begin types = @connection.native_database_types modify_types(types) types end end
@private
# File lib/arjdbc/jdbc/adapter.rb, line 229 def native_sql_to_type(type) if /^(.*?)\(([0-9]+)\)/ =~ type tname, limit = $1, $2.to_i ntypes = native_database_types if ntypes[:primary_key] == type return :primary_key, nil else ntypes.each do |name, val| if name == :primary_key next end if val[:name].downcase == tname.downcase && ( val[:limit].nil? || val[:limit].to_i == limit ) return name, limit end end end elsif /^(.*?)/ =~ type tname = $1 ntypes = native_database_types if ntypes[:primary_key] == type return :primary_key, nil else ntypes.each do |name, val| if val[:name].downcase == tname.downcase && val[:limit].nil? return name, nil end end end else return :string, 255 end return nil, nil end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 370 def pk_and_sequence_for(table) ( key = primary_key(table) ) ? [ key, nil ] : nil end
@override
@private @override
# File lib/arjdbc/jdbc/adapter.rb, line 297 def select_rows(sql, name = nil, binds = []) exec_query_raw(sql, name, binds).map!(&:values) end
Abstract adapter default implementation does nothing silently. @override
# File lib/arjdbc/jdbc/adapter.rb, line 211 def structure_dump raise NotImplementedError, "structure_dump not supported" end
Does our database (+ its JDBC driver) support foreign-keys? @since 1.3.18 @override
# File lib/arjdbc/jdbc/adapter.rb, line 387 def supports_foreign_keys? @connection.supports_foreign_keys? end
JDBC adapters support migration. @return [true] @override
# File lib/arjdbc/jdbc/adapter.rb, line 218 def supports_migrations? true end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 269 def supports_views? @connection.supports_views? end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 349 def table_exists?(name) return false unless name @connection.table_exists?(name) # schema_name = nil end
@override
# File lib/arjdbc/jdbc/adapter.rb, line 344 def tables(name = nil) @connection.tables end
@private
# File lib/arjdbc/jdbc/adapter.rb, line 408 def to_sql(arel, binds = nil) # NOTE: can not handle `visitor.accept(arel.ast)` right arel.respond_to?(:to_sql) ? arel.send(:to_sql) : arel end
@param record the record e.g. `User.find(1)` @param column the model's column e.g. `User.columns_hash` @param value the lob value - string or (IO or Java) stream
# File lib/arjdbc/jdbc/adapter.rb, line 399 def update_lob_value(record, column, value) @connection.update_lob_value(record, column, value) end
@override introduced in AR 4.2
# File lib/arjdbc/jdbc/adapter.rb, line 198 def valid_type?(type) ! native_database_types[type].nil? end
@deprecated Rather use {#update_lob_value} instead.
# File lib/arjdbc/jdbc/adapter.rb, line 392 def write_large_object(*args) @connection.write_large_object(*args) end
Protected Instance Methods
Provides backwards-compatibility on ActiveRecord
4.1 for DB adapters that override this and than call super expecting to work. @note This method is available in 4.0 but won't be in 4.1 @private
# File lib/arjdbc/jdbc/adapter.rb, line 495 def add_column_options!(sql, options) sql << " DEFAULT #{quote(options[:default], options[:column])}" if options_include_default?(options) # must explicitly check for :null to allow change_column to work on migrations sql << " NOT NULL" if options[:null] == false sql << " AUTO_INCREMENT" if options[:auto_increment] == true end
`TableDefinition.new native_database_types
, name, temporary, options` and ActiveRecord
4.1 supports optional `as` argument (which defaults to nil) to provide the SQL to use to generate the table: `TableDefinition.new native_database_types
, name, temporary, options, as` @private
# File lib/arjdbc/jdbc/adapter.rb, line 467 def create_table_definition(*args) table_definition(*args) end
Take an id from the result of an INSERT query. @return [Integer, NilClass]
# File lib/arjdbc/jdbc/adapter.rb, line 438 def last_inserted_id(result) if result.is_a?(Hash) || result.is_a?(ActiveRecord::Result) result.first.first[1] # .first = { "id"=>1 } .first = [ "id", 1 ] else result end end
@override so that we do not have to care having 2 arguments on 3.0
# File lib/arjdbc/jdbc/adapter.rb, line 425 def log(sql, name = nil, binds = []) unless binds.blank? binds = binds.map do |column, value| column ? [column.name, value] : [nil, value] end sql = "#{sql} #{binds.inspect}" end super(sql, name || 'SQL') # `log(sql, name)` on AR <= 3.0 end
@private
# File lib/arjdbc/jdbc/adapter.rb, line 483 def new_index_definition(table, name, unique, columns, lengths, orders = nil, where = nil, type = nil, using = nil) IndexDefinition.new(table, name, unique, columns, lengths, orders, where, type, using) end
@note AR-4x arguments expected: `(name, temporary, options)` @private documented bellow
# File lib/arjdbc/jdbc/adapter.rb, line 473 def new_table_definition(table_definition, *args) if ActiveRecord::VERSION::MAJOR > 4 table_definition.new(*args) else table_definition.new native_database_types, *args end end
@return whether `:prepared_statements` are to be used
# File lib/arjdbc/jdbc/adapter.rb, line 531 def prepared_statements? return @prepared_statements unless (@prepared_statements ||= nil).nil? @prepared_statements = self.class.prepared_statements?(config) end
aliasing create_table_definition
as table_definition
:
Private Instance Methods
We need to do it this way, to allow Rails stupid tests to always work even if we define a new `execute` method. Instead of mixing in a new `execute`, an `_execute` should be mixed in. @deprecated it was only introduced due tests @private
# File lib/arjdbc/jdbc/adapter.rb, line 327 def _execute(sql, name = nil) @connection.execute(sql) end
@deprecated No longer used, kept for 1.2 API compatibility.
# File lib/arjdbc/jdbc/adapter.rb, line 568 def extract_sql(arel) arel.respond_to?(:to_sql) ? arel.send(:to_sql) : arel end
Helper to get local/UTC time (based on `ActiveRecord::Base.default_timezone`).
# File lib/arjdbc/jdbc/adapter.rb, line 582 def get_time(value) get = ::ActiveRecord::Base.default_timezone == :utc ? :getutc : :getlocal value.respond_to?(get) ? value.send(get) : value end
Helper useful during {#quote} since AREL might pass in it's literals to be quoted, fixed since AREL 4.0.0.beta1 : git.io/7gyTig
# File lib/arjdbc/jdbc/adapter.rb, line 575 def sql_literal?(value); ::Arel::Nodes::SqlLiteral === value; end
@private Supporting “string-subling” on AR 4.0 would require {#to_sql} to consume binds parameters otherwise it happens twice e.g. for a record insert it is called during {#insert} as well as on {#exec_insert} … but that than leads to other issues with libraries that save the binds array and run a query again since it's the very same instance on 4.0 !
# File lib/arjdbc/jdbc/adapter.rb, line 563 def suble_binds(sql, binds) sql end