= New Features

* Database#run and #<< now accept SQL::PlaceholderLiteralString
  objects, allowing you to more easily run arbitrary DDL queries with
  placeholders:

    DB.run Sequel.lit("CREATE TABLE ? (? integer)", :table, :column)

* You can now provide options for check constraints by calling the
  constraint/add_constraint methods with a hash as the first argument.
  On PostgreSQL, you can now use the :not_valid option for check
  constraints, so they are enforced for inserts and updates, but
  not for existing rows.

    DB.create_table(:table) do
      ...
      constraint({:name=>:constraint_name, :not_valid=>true}) do
        column_name > 10
      end
    end
    
* Dataset#stream has been added to the mysql2 adapter, and will have
  the dataset stream results if used with mysql2 0.3.12+.  This
  allows you to process large datasets without keeping the entire
  dataset in memory.

    DB[:large_table].stream.each{|r| ...}

* Database#error_info has been added to the postgres adapter.  It
  is supported on PostgreSQL 9.3+ if pg-0.16.0+ is used as the
  underlying driver, and it gives you a hash of metadata related
  to the exception:

    DB[:table_name].insert(1) rescue DB.error_info($!)
    # => {:schema=>"public", :table=>"table_name", :column=>nil,
          :constraint=>"constraint_name", :type=>nil}
 
* The :deferrable option is now supported when adding exclusion
  constraints on PostgreSQL, to allow setting up deferred exclusion
  constraints.

* The :inherits option is now supported in Database#create_table on
  PostgreSQL, for table inheritance:

    DB.create_table(:t1, :inherits=>:t0){}
    # CREATE TABLE t1 () INHERITS (t0)

* Dataset#replace and #multi_replace are now supported on SQLite,
  just as they have been previously on MySQL.

* In the jdbc adapter, Java::JavaUtil::HashMap objects are now
  converted to ruby Hash objects.  This is to make it easier to
  handle the PostgreSQL hstore type when using the jdbc/postgres
  adapter.

* The odbc adapter now supports a :drvconnect option that accepts
  an ODBC connection string that is passed to ruby-odbc verbatim.

= Other Improvements

* The prepared_statements plugin no longer breaks the
  instance_filters and update_primary_key plugins.

* Dropping indexes for tables in a specific schema is now supported
  on PostgreSQL.  Sequel now explicitly specifies the same schema
  as the table when dropping such indexes.

* Calling Model#add_association methods with a primary key value
  now raises a Sequel::NoMatchingRow if there is no object in the
  associated table with that primary key.  Previously, this
  situation was not handled and resulted in a NoMethodError being
  raised later.

* When an invalid virtual row block function call is detected, an
  error is now properly raised.  Previously, the error was not
  raised until the SQL was produced for the query.

= Backwards Compatibility

* The :driver option to the odbc adapter is deprecated and will be
  removed in a future version.  It is thought to be broken, and
  users wanting to use DSN-less connections should use the new
  :drvconnect option.

* The Postgres::ArrayOp#text_op private method has been removed.