class Sequel::Postgres::Database
Database class for PostgreSQL databases used with Sequel and the pg, postgres, or postgres-pr driver.
Constants
- DatasetClass
- INFINITE_DATETIME_VALUES
- INFINITE_TIMESTAMP_STRINGS
Attributes
Whether infinite timestamps/dates should be converted on retrieval. By default, no conversion is done, so an error is raised if you attempt to retrieve an infinite timestamp/date. You can set this to :nil to convert to nil, :string to leave as a string, or :float to convert to an infinite float.
Public Instance Methods
Convert given argument so that it can be used directly by pg. Currently, pg doesn't handle fractional seconds in Time/DateTime or blobs with “0”, and it won't ever handle Sequel::SQLTime values correctly. Only public for use by the adapter, shouldn't be used by external code.
# File lib/sequel/adapters/postgres.rb, line 216 def bound_variable_arg(arg, conn) case arg when Sequel::SQL::Blob {:value=>arg, :type=>17, :format=>1} when Sequel::SQLTime literal(arg) when DateTime, Time literal(arg) else arg end end
Connects to the database. In addition to the standard database options, using the :encoding or :charset option changes the client encoding for the connection, :connect_timeout is a connection timeout in seconds, :sslmode sets whether postgres's sslmode, and :notice_receiver handles server notices in a proc. :connect_timeout, :driver_options, :sslmode, and :notice_receiver are only supported if the pg driver is used.
# File lib/sequel/adapters/postgres.rb, line 236 def connect(server) opts = server_opts(server) if USES_PG connection_params = { :host => opts[:host], :port => opts[:port] || 5432, :dbname => opts[:database], :user => opts[:user], :password => opts[:password], :connect_timeout => opts[:connect_timeout] || 20, :sslmode => opts[:sslmode] }.delete_if { |key, value| blank_object?(value) } connection_params.merge!(opts[:driver_options]) if opts[:driver_options] conn = Adapter.connect(connection_params) conn.instance_variable_set(:@prepared_statements, {}) if receiver = opts[:notice_receiver] conn.set_notice_receiver(&receiver) end else conn = Adapter.connect( (opts[:host] unless blank_object?(opts[:host])), opts[:port] || 5432, nil, '', opts[:database], opts[:user], opts[:password] ) end conn.instance_variable_set(:@db, self) if encoding = opts[:encoding] || opts[:charset] if conn.respond_to?(:set_client_encoding) conn.set_client_encoding(encoding) else conn.async_exec("set client_encoding to '#{encoding}'") end end connection_configuration_sqls.each{|sql| conn.execute(sql)} conn end
Set whether to allow infinite timestamps/dates. Make sure the conversion proc for date reflects that setting.
# File lib/sequel/adapters/postgres.rb, line 283 def convert_infinite_timestamps=(v) @convert_infinite_timestamps = case v when Symbol v when 'nil' :nil when 'string' :string when 'float' :float when String typecast_value_boolean(v) else false end pr = old_pr = @use_iso_date_format ? TYPE_TRANSLATOR.method(:date) : Sequel.method(:string_to_date) if v pr = lambda do |val| case val when *INFINITE_TIMESTAMP_STRINGS infinite_timestamp_value(val) else old_pr.call(val) end end end conversion_procs[1082] = pr end
copy_into
uses PostgreSQL's +COPY FROM STDIN+ SQL statement to do very fast inserts into a table
using input preformatting in either CSV or PostgreSQL text format. This
method is only supported if pg 0.14.0+ is the underlying ruby driver. This
method should only be called if you want results returned to the client.
If you are using +COPY FROM+ with a filename, you should just use
run
instead of this method.
The following options are respected:
- :columns
-
The columns to insert into, with the same order as the columns in the input data. If this isn't given, uses all columns in the table.
- :data
-
The data to copy to PostgreSQL, which should already be in CSV or PostgreSQL text format. This can be either a string, or any object that responds to each and yields string.
- :format
-
The format to use. text is the default, so this should be :csv or :binary.
- :options
-
An options SQL string to use, which should contain comma separated options.
- :server
-
The server on which to run the query.
If a block is provided and :data option is not, this will yield to the block repeatedly. The block should return a string, or nil to signal that it is finished.
# File lib/sequel/adapters/postgres.rb, line 415 def copy_into(table, opts=OPTS) data = opts[:data] data = Array(data) if data.is_a?(String) if block_given? && data raise Error, "Cannot provide both a :data option and a block to copy_into" elsif !block_given? && !data raise Error, "Must provide either a :data option or a block to copy_into" end synchronize(opts[:server]) do |conn| conn.execute(copy_into_sql(table, opts)) begin if block_given? while buf = yield conn.put_copy_data(buf) end else data.each{|buff| conn.put_copy_data(buff)} end rescue Exception => e conn.put_copy_end("ruby exception occurred while copying data into PostgreSQL") ensure conn.put_copy_end unless e while res = conn.get_result raise e if e check_database_errors{res.check} end end end end
copy_table
uses PostgreSQL's +COPY TO STDOUT+ SQL statement to return formatted results directly
to the caller. This method is only supported if pg is the underlying ruby
driver. This method should only be called if you want results returned to
the client. If you are using +COPY TO+ with a filename, you should just
use run
instead of this method.
The table argument supports the following types:
- String
-
Uses the first argument directly as literal SQL. If you are using a version of PostgreSQL before 9.0, you will probably want to use a string if you are using any options at all, as the syntax Sequel uses for options is only compatible with PostgreSQL 9.0+.
- Dataset
-
Uses a query instead of a table name when copying.
- other
-
Uses a table name (usually a symbol) when copying.
The following options are respected:
- :format
-
The format to use. text is the default, so this should be :csv or :binary.
- :options
-
An options SQL string to use, which should contain comma separated options.
- :server
-
The server on which to run the query.
If a block is provided, the method continually yields to the block, one yield per row. If a block is not provided, a single string is returned with all of the data.
# File lib/sequel/adapters/postgres.rb, line 375 def copy_table(table, opts=OPTS) synchronize(opts[:server]) do |conn| conn.execute(copy_table_sql(table, opts)) begin if block_given? while buf = conn.get_copy_data yield buf end nil else b = String.new b << buf while buf = conn.get_copy_data b end ensure raise DatabaseDisconnectError, "disconnecting as a partial COPY may leave the connection in an unusable state" if buf end end end
Disconnect given connection
# File lib/sequel/adapters/postgres.rb, line 314 def disconnect_connection(conn) conn.finish rescue PGError, IOError nil end
Return a hash of information about the related PGError (or Sequel::DatabaseError that wraps a PGError), with the following entries:
- :schema
-
The schema name related to the error
- :table
-
The table name related to the error
- :column
-
the column name related to the error
- :constraint
-
The constraint name related to the error
- :type
-
The datatype name related to the error
This requires a PostgreSQL 9.3+ server and 9.3+ client library, and ruby-pg 0.16.0+ to be supported.
# File lib/sequel/adapters/postgres.rb, line 332 def error_info(e) e = e.wrapped_exception if e.is_a?(DatabaseError) r = e.result h = {} h[:schema] = r.error_field(::PG::PG_DIAG_SCHEMA_NAME) h[:table] = r.error_field(::PG::PG_DIAG_TABLE_NAME) h[:column] = r.error_field(::PG::PG_DIAG_COLUMN_NAME) h[:constraint] = r.error_field(::PG::PG_DIAG_CONSTRAINT_NAME) h[:type] = r.error_field(::PG::PG_DIAG_DATATYPE_NAME) h end
Execute the given SQL with the given args on an available connection.
# File lib/sequel/adapters/postgres.rb, line 346 def execute(sql, opts=OPTS, &block) synchronize(opts[:server]){|conn| check_database_errors{_execute(conn, sql, opts, &block)}} end
Listens on the given channel (or multiple channels if channel is an array), waiting for notifications. After a notification is received, or the timeout has passed, stops listening to the channel. Options:
- :after_listen
-
An object that responds to
call
that is called with the underlying connection after the LISTEN statement is sent, but before the connection starts waiting for notifications. - :loop
-
Whether to continually wait for notifications, instead of just waiting for a single notification. If this option is given, a block must be provided. If this object responds to
call
, it is called with the underlying connection after each notification is received (after the block is called). If a :timeout option is used, and a callable object is given, the object will also be called if the timeout expires. If :loop is used and you want to stop listening, you can either break from inside the block given to listen, or you can throw :stop from inside the :loop object's call method or the block. - :server
-
The server on which to listen, if the sharding support is being used.
- :timeout
-
How long to wait for a notification, in seconds (can provide a float value for fractional seconds). If this object responds to
call
, it will be called and should return the number of seconds to wait. If the loop option is also specified, the object will be called on each iteration to obtain a new timeout value. If not given or nil, waits indefinitely.
This method is only supported if pg is used as the underlying ruby driver. It returns the channel the notification was sent to (as a string), unless :loop was used, in which case it returns nil. If a block is given, it is yielded 3 arguments:
-
the channel the notification was sent to (as a string)
-
the backend pid of the notifier (as an integer),
-
and the payload of the notification (as a string or nil).
# File lib/sequel/adapters/postgres.rb, line 470 def listen(channels, opts=OPTS, &block) check_database_errors do synchronize(opts[:server]) do |conn| begin channels = Array(channels) channels.each do |channel| sql = "LISTEN ".dup dataset.send(:identifier_append, sql, channel) conn.execute(sql) end opts[:after_listen].call(conn) if opts[:after_listen] timeout = opts[:timeout] if timeout timeout_block = timeout.respond_to?(:call) ? timeout : proc{timeout} end if l = opts[:loop] raise Error, 'calling #listen with :loop requires a block' unless block loop_call = l.respond_to?(:call) catch(:stop) do loop do t = timeout_block ? [timeout_block.call] : [] conn.wait_for_notify(*t, &block) l.call(conn) if loop_call end end nil else t = timeout_block ? [timeout_block.call] : [] conn.wait_for_notify(*t, &block) end ensure conn.execute("UNLISTEN *") end end end end
If #convert_infinite_timestamps is true and the value is infinite, return an appropriate value based on the #convert_infinite_timestamps setting.
# File lib/sequel/adapters/postgres.rb, line 511 def to_application_timestamp(value) if convert_infinite_timestamps case value when *INFINITE_TIMESTAMP_STRINGS infinite_timestamp_value(value) else super end else super end end
Private Instance Methods
Execute the given SQL string or prepared statement on the connection object.
# File lib/sequel/adapters/postgres.rb, line 527 def _execute(conn, sql, opts, &block) if sql.is_a?(Symbol) execute_prepared_statement(conn, sql, opts, &block) else conn.execute(sql, opts[:arguments], &block) end end
Execute the prepared statement name with the given arguments on the connection.
# File lib/sequel/adapters/postgres.rb, line 536 def _execute_prepared_statement(conn, ps_name, args, opts) conn.exec_prepared(ps_name, args) end
Add the primary_keys and primary_key_sequences instance variables, so we can get the correct return values for inserted rows.
# File lib/sequel/adapters/postgres.rb, line 542 def adapter_initialize @use_iso_date_format = typecast_value_boolean(@opts.fetch(:use_iso_date_format, Postgres.use_iso_date_format)) initialize_postgres_adapter conversion_procs[1082] = TYPE_TRANSLATOR.method(:date) if @use_iso_date_format self.convert_infinite_timestamps = @opts[:convert_infinite_timestamps] end
Convert exceptions raised from the block into DatabaseErrors.
# File lib/sequel/adapters/postgres.rb, line 550 def check_database_errors begin yield rescue => e raise_error(e, :classes=>CONVERTED_EXCEPTIONS) end end
Set the DateStyle to ISO if configured, for faster date parsing.
# File lib/sequel/adapters/postgres.rb, line 559 def connection_configuration_sqls sqls = super sqls << "SET DateStyle = 'ISO'" if @use_iso_date_format sqls end
# File lib/sequel/adapters/postgres.rb, line 565 def database_error_classes [PGError] end
# File lib/sequel/adapters/postgres.rb, line 575 def database_exception_sqlstate(exception, opts) if exception.respond_to?(:result) && (result = exception.result) result.error_field(::PGresult::PG_DIAG_SQLSTATE) end end
# File lib/sequel/adapters/postgres.rb, line 569 def disconnect_error?(exception, opts) super || Adapter::DISCONNECT_ERROR_CLASSES.any?{|klass| exception.is_a?(klass)} || exception.message =~ Adapter::DISCONNECT_ERROR_RE end
Execute the prepared statement with the given name on an available connection, using the given args. If the connection has not prepared a statement with the given name yet, prepare it. If the connection has prepared a statement with the same name and different SQL, deallocate that statement first and then prepare this statement. If a block is given, yield the result, otherwise, return the number of rows changed.
# File lib/sequel/adapters/postgres.rb, line 588 def execute_prepared_statement(conn, name, opts=OPTS, &block) ps = prepared_statement(name) sql = ps.prepared_sql ps_name = name.to_s if args = opts[:arguments] args = args.map{|arg| bound_variable_arg(arg, conn)} end unless conn.prepared_statements[ps_name] == sql conn.execute("DEALLOCATE #{ps_name}") if conn.prepared_statements.include?(ps_name) conn.check_disconnect_errors{log_connection_yield("PREPARE #{ps_name} AS #{sql}", conn){conn.prepare(ps_name, sql)}} conn.prepared_statements[ps_name] = sql end log_sql = "EXECUTE #{ps_name}" if ps.log_sql log_sql += " (" log_sql << sql log_sql << ")" end q = conn.check_disconnect_errors{log_connection_yield(log_sql, conn, args){_execute_prepared_statement(conn, ps_name, args, opts)}} begin block_given? ? yield(q) : q.cmd_tuples ensure q.clear if q && q.respond_to?(:clear) end end
Return an appropriate value for the given infinite timestamp string.
# File lib/sequel/adapters/postgres.rb, line 619 def infinite_timestamp_value(value) case convert_infinite_timestamps when :nil nil when :string value else value == 'infinity' ? PLUS_INFINITY : MINUS_INFINITY end end
Don't log, since logging is done by the underlying connection.
# File lib/sequel/adapters/postgres.rb, line 631 def log_connection_execute(conn, sql) conn.execute(sql) end
If the value is an infinite value (either an infinite float or a string returned by by PostgreSQL for an infinite timestamp), return it without converting it if #convert_infinite_timestamps is set.
# File lib/sequel/adapters/postgres.rb, line 638 def typecast_value_date(value) if convert_infinite_timestamps case value when *INFINITE_DATETIME_VALUES value else super end else super end end
If the value is an infinite value (either an infinite float or a string returned by by PostgreSQL for an infinite timestamp), return it without converting it if #convert_infinite_timestamps is set.
# File lib/sequel/adapters/postgres.rb, line 654 def typecast_value_datetime(value) if convert_infinite_timestamps case value when *INFINITE_DATETIME_VALUES value else super end else super end end