New Features¶ ↑

fdbsql and jdbc/fdbsql adapters have been added, for connecting to FoundationDB SQL Layer.
A split_values plugin has been added, for moving non-column entries from the values hash into a separate hash. This allows you to select additional columns (e.g. computed columns) when retrieving model instances, and be able to save those instances without removing the additional columns.
A Sequel::Model.cache_associations accessor has been added, which can be set to false to not cache any association metadata. This can fix issues in a development environment that uses code reloading.
The active_model plugin now supports activemodel 4.2.0beta1.
More PostgreSQL array types are handled automatically by the pg_array extension, such as xml[] and uuid[].
Creating foreign tables is now supported on PostgreSQL via the :foreign and :options create_table options.
The :nolog Database option is now supported in the informix adapter, where it disables the use of transactions.
PlaceholderLiteralizer#with_dataset has been added, allowing you to create another PlaceholderLiteralizer with a modified dataset, useful if you want to change the row_proc or any non-SQL dataset options.

Other Improvements¶ ↑

The tactical_eager_loading plugin once again works correctly with limited associations.
A bug in older versions of MySQL is now worked around when schema dumping a table with multiple timestamp columns.
On PostgreSQL, create_view(:view_name, dataset, :materialized=>true) is now reversible.
Postgres::{JSON,JSONB}Op#to_record and to_recordset no longer take an optional argument. This was supported in PostgreSQL 9.4beta1, but removed before PostgreSQL 9.4beta2.
Dataset#insert now returns the last inserted id in the informix adapter.
Sequel no longer raises an exception in AssociationReflection#reciprocal if the associated class has an association that does not have a valid associated class.
Sequel now raises an exception if a primary key is necessary to use an association, but the model does not have a primary key.