This release adds many improvements to Ecto with a handful of bug fixes. We recommend reading the full list of enhancements below. Here are some of the highlights.
Ecto now supports specifying fields sources. This is useful when you use a database with non-conventional names, such as uppercase letters or names with hyphens:
field :name, :string, source: :NAME
field :user_name, :string, source: :"user-name"
The source option can be specified programatically by given a function to @field_source_mapper that receives the field name as an atom and returns the source name as an atom.
On the migrations side, execute/2 function was added, which allows developers to describe up and down commands inside the change callback:
def change do
execute "CREATE EXTENSION citext",
"DROP EXTENSION citext"
end
The ecto.migrate and ecto.rollback tasks have also been enhanced with the --log-sql option which is helpful when debugging errors or during deployments to keep track of your production database changes.
Ecto now will also warn at compile time of invalid relationships, such as a belongs_to that points to a schema that does not exist.
The query syntax also seen some improvements: map updates are supported in subqueries, the new Ecto.Query.select_merge/3 makes it easier to write composable select clauses in queries and the type/2 macro has been extended to support casting of fragments and fields in schemaless queries.
Finally, the UPSERT support added on Ecto v2.1 is getting more improvements: the {:constraint, constraint} is now supported as conflict target and the :returning option was added to Ecto.Repo.insert/2, mirroring the behaviour of insert_all.
- [Ecto.Schema] Avoid false positive warnings
- [Ecto] Fix warnings on Elixir v1.7
- [Ecto.Query] Allow field interpolation in update
- [Ecto.Query] Properly expand the right side of
inin joins - [Ecto.Repo] Warn when using preloads with
Ecto.Repo.stream/2 - [Ecto.Repo.Preloader] Only sort by the relationship key and not the whole struct
- [Ecto.Schema] Raise proper error message when any is used with non-virtual field
- [Ecto.Schema] Do not track many-to-many relationships as compile time dependencies
- [Ecto] Add .formatter.exs file
- [Ecto.Changeset] Add
:trimoption forvalidate_required - [Ecto.Type] Support "hh:mm" format in
:timefield
- [Ecto.Changeset] Fix
unsafe_validate_unique/3for partially matching composite keys - [Ecto.Query] Do not generate default ON on cross joins
- [Ecto.Adapters.MySQL] Raise clearer error when using where in on_conflict with MySQL
- [Ecto.Adapters.SQL] Document after_connect callbacks
- [Ecto.Repo] Improve docs for assoc with on_conflict
- [Ecto.Schema] Promote using :jsonb for Postgres for :embeds_many
- [Ecto.Adapters.MySQL] Do not crash ecto.create/drop on log: false with mysql
- [Ecto.Adepters.Postgres] Prepend the schema "public" when referring to the schema_migrations table on ecto.dump
- [Ecto.Type] Do not raise when casting integer to NaiveDateTime
- [Ecto.Repo] Allow
ssl,timeout,pool_timeoutandpool_sizeto be given as URL parameter
- [Ecto.Adapters.MySQL] Fix out of order parameters when issuing an update_all with join in MySQL
- [Ecto.Query] Fix a bug when a parameterized query is given as argument to join
- [Ecto.Query] Ensure a list of fields given to
select_merge: [...]appends to the list of fields previously given in the select - [Ecto.Query] Ensure fields loaded via
select_mergeare available during preloading - [Ecto.Query] Show better error message on ambiguity between query and assoc on
:preload - [Ecto.Query] Allow virtual fields to be updated in subqueries
- [Ecto.Repo] Mark schemas returned from subqueries as loaded
- [Ecto.Repo] Do not surface embeds if repo operation fails
- [Ecto.Schema] Raise if updating with struct on on_replace: :update
- [Ecto.Adapters.Postgres] Properly interpolate parameters when
inis used in query
- [Ecto.Changeset] Support
:prefixonunsafe_validate_unique - [Ecto.Migration] Support for
ON DELETE RESTRICTandON UPDATE RESTRICTin migrations - [Ecto.Query] Fix params counter when using
inin some places in query - [Ecto.Schema] Properly support
on_replace: :updateonbelongs_to/has_one - [Ecto.Type] Allow casting Date from NaiveDateTime ISO
- [Ecto.Repo] Do not surface embeds if repo operation fails
- [Mix.Tasks.Ecto.Migrate] Only keep logger backends when running migrations
- [Ecto.Query] Ensure recursive macros in
selectare correctly expanded
- [Ecto.Repo] Ensure preloads work when the first associated entry is nil
- [Ecto.Changeset] Allow validation of codepoint count in validate_length
- [Ecto.Query] Support
prefixin subqueries
- [Ecto.Adapters] Properly alias table names starting with numbers
- [Ecto.Adapters.Postgres] Make sure type/2 with integer/id types are tagged as bigint in Posgres
- [Ecto.Changeset] Properly validate
has_many/many_to_manyassociations onvalidate_length/3when entries are being deleted/replaced - [Ecto.Migration] Set foreign key type to the same type as the primary key in the repository configuration
- [Ecto.Repo] Do not attempt to preload
has_one/belongs_toassociations already loaded as nil - [Ecto.Schema] Do not attempt to validate associations if the associated module is in context
- [Ecto.Schema] Bring back
__schema__(:types)for backwards compatibility
- [Ecto.Changeset] Do not raise when an empty association is given to a field that has an association already stored in the struct
- [Ecto.Repo] Do not lookup nil values in
get_by
- [Ecto.Adapters] Accept IO data from adapters to reduce memory usage when performing large queries
- [Ecto.Adapters.SQL] Also add
Ecto.Repo.to_sql/2to Ecto.Repo based on SQL adapters - [Ecto.Adapters.Postgres] Use the "postgres" database for create/drop database commands
- [Ecto.Adapters.MySQL] Use TCP connections instead of MySQL command client to create & drop database
- [Ecto.Changeset] Support
action: :ignorein changeset which is useful when casting associations and embeds and one or more children need to be rejected/ignored under certain circumstances - [Ecto.Changeset] Add
:repo_optsfield toEcto.Changesetwhich are given as options to to the repository whenever an operation is performed - [Ecto.Changeset] Add
unsafe_validate_unique/3which validates uniqueness for faster feedback cycles but without data-integrity guarantees - [Ecto.Changeset] Add
apply_action/2 - [Ecto.Changeset] Add prefix constraint name checking to constraint validations
- [Ecto.Changeset] Allow assocs and embeds in
change/2andput_change/3- this gives a more generic API for developers to work that does not require explicit knowledge of the field type - [Ecto.Migration] Add reversible
execute/2to migrations - [Ecto.Migration] Add
:migration_timestampsand:migration_primary_keyto control the migration defaults from the repository - [Ecto.Migrator] Allow migration/rollback to log SQL commands via the
--log-sqlflag - [Ecto.LogEntry] Add
:caller_pidto the Ecto.LogEntry struct - [Ecto.Query] Allow map updates in subqueries
- [Ecto.Query] Support fragment, aggregates and field access in
type/2in select - [Ecto.Query] Add
select_merge/3as a composable API for selects - [Ecto.Repo] Implement
:returningoption on insert - [Ecto.Repo] Add ON CONSTRAINT support to
:conflict_targetoninsertandinsert_all - [Ecto.Repo] Raise
MultiplePrimaryKeyErrorwhen primary key is not unique on DB side - [Ecto.Schema] Validate schemas after compilation - this helps developers catch early mistakes such as foreign key mismatches early on
- [Ecto.Schema] Support the
:sourceoption in thefield/3macro which configures the column/field name in the data storage - [Ecto.Schema] Support
@field_source_mapperinEcto.Schemaas a mechanism to programatically set the:sourceoption - [Ecto.Type] Allow adapters to pass Date, Time, NaiveDateTime and DateTime on load if desired
- [Ecto.UUID] Allow casting binary UUIDs
- [mix ecto.drop] Add
--force - [mix ecto.load] Add
--forceand prompt user to confirm before continuing in production
- [Ecto.Changeset] Remove the field from changes if it does not pass
validate_required - [Ecto.Changeset] Raise if changeset struct does not match relation
- [Ecto.Query] Consistently raise if
nilis interpolated on the right side ofin - [Ecto.Query] Properly expand macros in
select - [Ecto.Query] Support
or_havingin keyword query - [Ecto.Query] Properly count the parameters when using interpolation inside a
selectinside asubquery* [Ecto.Query] Properlyselectexpressions in subquery (for example,{p.field}doesn't make sense as a subquery return but it was silently accepted in the past) * [Ecto.Repo] Set struct prefix oninsert,delete, andupdatewhen the prefix is given as an option - [Ecto.UUID] Validate UUID version on casting
- [mix ecto.*] Make sure
Loggeris rebootted when running ecto tasks - [mix ecto.*] No longer mark tasks as recursive and instead collect all repositories upfront. This fixes a bug where migration and rollback commands for a given repository could be executed multiple times from an umbrella project
- [Ecto.DateTime]
Ecto.DateTimeas well asEcto.DateandEcto.Timeare deprecated in favor of:naive_datetime,:dateand:timerespectively - [Ecto.Repo] Using
{:system, env}to configure the repository URL is deprecated in favor of a custom init/2 callback
- See the CHANGELOG.md in the v2.1 branch