Skip to content

Release Notes

0.34.1

Added

  • add_many() to ManyToManyField for adding multiple relations at once.
  • remove_many() to ManyToManyField for removing multiple relations at once.
  • Add arefresh_metadata which doesn't use sometimes a second loop but is async from the start.

Changed

  • Don't autoimport models which are excluded from admin.
  • Deprecate edgy.settings, code should use edgy.monkay.settings instead. This is for Edgy internals, the end user can still use from edgy.conf import settings.

Fixed

  • Lifespan eventloop was not the same as the eventloop in shell (ipython, ptpython).
  • Improve automatic imports.
  • Unnecessary database initializations during create_all.
  • Fix __module__ attribute for autogenerated through models.
  • Fix tenant models not included in migrations.
  • Proper handle db_schema in refresh_metadata.

0.34.0

Changed

  • For ManyToManyField make NEW_M2M_NAMING the default. No need to set it explicit anymore.
  • Use sayer for cli.
  • Add .jinja suffix to admin templates (makes it easier to recognize jinja templates for IDEs and humans).
  • Move 404.html.jinja to admin, as it is solely used by admin_serve now.
  • schemas: Use for non-schema databases (include now fallbacks). This simplifies the codebase.

Fixed

  • The create_admin_app created lilya instance doesn't intercept 404 errors anymore.
  • 404.html.jinja uses now also the base_definitions.html.jinja so we can exchange libraries at one place.
  • schemas: Misc fixes. We could end up with a wrong context or a closed database.

Removed

  • The deprecated activate_schema_path method.

Warning

  • The omitting of through_tablename=edgy.NEW_M2M_NAMING for ManyToManyField only works for edgy >=0.34.0. To stay backward compatible you still have to pass the arguments.

0.33.3

Fixed

  • Handling non-starlette like lifespan when executing shell.

Changed

  • Replace ASGIHelper by monkay asgi helpers.
  • Replace generate_asgi_events by monkay asgi helpers.

0.33.2

Fixed

  • Rollback from a regression in the core.terminal colors causing issues with the shell

0.33.1

Fixed

  • Crash because pydantic requires email-validator.
  • Incorrectly assigned table Alias to .table of model instances.

0.33.0

Added

  • Allow multiple admins.
  • Add asgi multiplexing middleware (edgy.contrib.lilya.middleware.EdgyMiddleware).
  • Support for isnull operator for querysets on a field.
  • Support for isempty operator for querysets on a field.
  • Support NumberedPaginator as alias for Paginator (n fact Paginator is an alias of NumberedPaginator).

Changed

  • JSONField uses now JSONB by default for postgresql (and only for postgresql).
  • Slightly changed InspectDB internal interface.
  • When using a relation, all queryset operations are by default executed on the target model except for ManyToManyFields with embed_through="" (default). In this case the reference is the through model. E.g. 
    class User(edgy.Model):
  name = edgy.CharField(max_length=255)
    ...
class Product(edgy.Model):
    name = edgy.CharField(max_length=255)
    users = edgy.ManyToMany("User", related_name="products", through_tablename=edgy.NEW_M2M_NAMING)
    users2 = edgy.ManyToMany("User", embed_through="", related_name="products", through_tablename=edgy.NEW_M2M_NAMING)
    users3 = edgy.ManyToMany("User", embed_through="embedded", related_name="products", through_tablename=edgy.NEW_M2M_NAMING)

product: Product = ...
await product.users.order_by("name")
# use old way
await product.users2.order_by("user__name")
# sort for product name
await product.users3.order_by("embedded__product__name")

Fixed

  • Status code when page in admin was not found.
  • Fix issues with json-editor ajax and enums by simply inlining all schemas and disabling ajax.
  • For the json schema view: don't use ref_template, it isn't well supported and leads to problems with enums.
  • Fix crash when None is passed to CompositeField.
  • Fix values_list when using relation queryset.
  • Fix values_list when using only string.
  • Fix performance when using relation queryset.
  • Fix order_by when using relation queryset.
  • Fix crash when Noneis passed to IPAddressField.

Breaking

  • JSONField uses now postgresql JSONB by default for postgresql (and only for postgresql). This may can cause an extra migration (because variants are used). If this is unwanted, add: no_jsonb=True to JSONField.
  • By fixing the column retrieval of order_by, ... we may break userspace. Fix it by removing the now superfluous prefix. embed_through="" is not affected as we have here the old interface.

0.32.5

Fixed

  • Admin with foreign keys as primary keys.
  • Revisioning with auto_now as primary key.
  • Fix crash with auto_now_add as primary key.
  • Fix "$defs" not matching def names. This fixes problems with enums which appear multiple times and reduces requests.
  • Fix using ForeignKeys in embedded models/CompositeFields when the embedded model has no registry set (e.g. abstract).
  • Improve admin UI.
  • Add some extra filters and option to make Primarykeys read-only in marshall.

Changed

  • Remove by default read-only fields from admin create/update.
  • Remove by default autoincrement fields from admin create.
  • Make PrimaryKeys read-only for update (they will be removed because they are read-only until we have a better solution).
  • FileStruct name have a minimal length of 1.

0.32.4

Fixed

  • Fix incorrect documentation for admin.
  • Fix admin_serve behind a reverse proxy.

0.32.3

Fixed

  • Add proper error for python>=3.12, when marshall_config is added as instance annotation.

0.32.2

Added

  • Add model search in admin.

Changed

  • Split out base head from admin into an extra template. This improves the customizability and allows to self-host.

Fixed

  • Circular import errors with marshalls.
  • Missing typing export for PGArrayField.
  • Proper escape json values.

0.32.1

Added

  • Customization hooks for admin.
  • Expose marshalls in a global edgy.marshalls namespace.
  • Add authentication for admin_serve.

Changed

  • Don't import mixins from sub-modules in edgy.core.db.models.mixins (to prevent import-cycles).
  • Move Metaclass from BaseMarshall to Marshall.
  • admin_prefix_url is automatically inferred.
  • Require session for embedding admin.

Fixed

  • Allow marshalls with the name Marshall.
  • Fix repr on marshalls.
  • Fix recent_models.

0.32.0

Added

  • Allow partial Marshalls.
  • Add admin_serve for modifying db.
  • FileFields have now a json-schema mode. You can pass a FileStruct like dict.

Changed

  • _loaded_or_deleted is now _db_loaded_or_deleted. The former name is now deprecated.
  • _loaded_or_deleted is splitted in _db_loaded and _db_deleted.
  • Migrations show now tracebacks when manual migration of objects fails.
  • Bump python requirement to python 3.10 to align with (optional) dependencies.

Fixed

  • Recursive saving.
  • Various json schema fixes.
  • Inherit register_default.

Breaking

  • Python 3.9 is no longer supported.
  • Remove long-deprecated edgy_setattr alias for object.__setattr__.

0.31.3

Fixed

  • Loose minimum version of rich causing conflicts.

0.31.2

Added

  • Add the method get_server_default.
  • Add get_columns_nullable, get_server_default and customize_default_for_server_default to factory overwritable methods.

Changed

  • server_default is not extracted anymore by ColumnDefinitionModel. Use get_server_default instead.
  • Lazy evaluate the environment variable for the settings import. This relaxes the restraints on the import order. You can e.g. import edgy settings before adjusting EDGY_SETTINGS_MODULE as long as you don't access the settings object.
  • Don't call evaluate_settings_once_ready during init of Registry anymore.

Removed

  • Remove the abstract method customize_default_for_server_default from BaseFieldType.

Fixed

  • Fix potential too early import of settings because of the auto computation of server_defaults.

Breaking

When creating a custom field providing get_columns and relying on the extraction of server defaults from field, you may need to update the code to use get_server_default.

0.31.1

Fixed

  • Fix and test QuerySet caching.
  • Fix potential hang in get_metadata_of_all_schemes with force_rollback active.

0.31.0

Added

  • Add the ChoiceField alternative CharChoiceField.
  • model_instance parameter for _save, _update signals.

Changed

  • Cleanup the callback interface:
  • The instances are only passed via ContextVars.
  • force_insert is by is_update replaced.
  • Delete uses now *_INSTANCE ContextVars.
  • Split delete in delete (only used for direct calls) and raw_delete for better customizations.
  • Split save in save (only used for direct calls) and real_save for better customizations.
  • Remove internal only parameter remove_referenced_call from delete (but not from raw_delete).
  • Change the remove_referenced_call to also accept strings, which are the fields from which the deletion originates from.
  • Virtual cascade and model based deletions doesn't trigger delete signals anymore (except if __deletion_with_signals__=True is set).
  • QuerySet.create passes now the QuerySet instance as CURRENT_INSTANCE.
  • QuerySet.create passes now the QuerySet instance as signal parameter instance.

Fixed

  • QuerySet.create passed the model instance as CURRENT_INSTANCE.
  • Virtual cascade and model based deletions doesn't trigger delete signals anymore.
  • Fix spurious iterate bug in bulk_get_or_create, triggered in edgy-guardian.
  • Fix wrong signal sender class for proxy models.

Breaking

  • For authors of custom fields: the interface of the async callback hooks changed:
  • The instances are only passed via ContextVars.
  • force_insert is by is_update replaced.
  • Delete callbacks use now *_INSTANCE ContextVars.
  • When you overwrote the delete method of a model, you probably want to overwrite now raw_delete.
  • When you overwrote the save method of a model, you probably want to overwrite now real_save.
  • delete losses its internal only parameter remove_referenced_call.
  • When using the pre_save/post_save signal, you might want to use the model_instance parameter instead of the instance parameter. We pass down QuerySets now in create.

0.30.1

Changed

  • Split from Page and Paginator BasePaginator and BasePage.
  • Add fields next_page, previous_page, current_page to Page.
  • Add field current_cursor to CursorPage.
  • Make specifying all of the page attributes mandatory.

Fixed

  • Fix and improve pagination examples.

0.30.0

Added

  • Add optimized pagination for QuerySet.

Changed

  • Internal cleanup how the current row is passed in QuerySet.
  • Optimize reverse method of QuerySet. Now it reuses a potential available cache.
  • build_pknames and build_pkcolumns return their result instead of setting it automatically.
  • Introduce _edgy_namespace for edgy related attributes.
  • Use _edgy_namespace for managers.

Fixed

  • FieldFile did use potential heavily blocking sync API in async code paths.
  • reverse when no order_by was set.
  • Serialization fixes for newer pydantic versions (>=2.11).
  • Private attributes handling.

0.29.0

Added

  • Convert for most fields defaults to server_default to ease migrations. There are some exceptions.
  • Add the setting allow_auto_compute_server_defaults which allows to disable the automatic generation of server defaults.
  • pre_migrate and post_migrate signals.

Changed

  • get_default_value is now also overwritable by factories.
  • The undocumented default of BooleanField of False is removed.
  • The signal receiver functions receive more parameters.

Fixed

  • JSONField default is deepcopied to prevent accidental modifications of the default. There is no need anymore to provide a lambda.
  • pre_update and post_update is not called anymore when calling save and the update path is executed.
  • apply_default_force_nullable_fields could run into transaction problems.
  • bulk_update and bulk_create optimizations and small fixes.
  • pre_update and post_update are now received when using a QuerySet.
  • Allow prefixing column names in CompositeField via prefix_column_name parameter.

Breaking

  • The undocumented default of BooleanField of False is removed.
  • Fields compute now server_defaults for a set default. This can be turned off via the allow_auto_compute_server_defaults setting.

0.28.2

Fixed

  • Missing message parameter passing in migrate command after refactory.

0.28.1

Changed

  • hash_names has now two prefix parameters: inner_prefix, outer_prefix.

Fixed

  • hash_names with unique_together incorrectly fed the uc prefix into the hasher. This makes the names unidentificatable.
  • BooleanField logic for automic migrations works by using server_default.

0.28.0

Added

  • null-field or nf parameter for makemigrations/revision.
  • Add FORCE_FIELDS_NULLABLE ContextVar.
  • Add CURRENT_FIELD_CONTEXT ContextVar.
  • Add automigrations (migrations capsuled in Registry).

Changed

  • The default migration templates allow now to use complex defaults for migrations.
  • Fields must use get_columns_nullable instead of ColumnDefinitionModel null. for determining if the columns should be nullable.
  • For model save and update the values can be callables which are evaluated.
  • Streamline ContentTypeField in using a parameter-less default function. Use CURRENT_FIELD_CONTEXT ContextVar field for referencing the owner.
  • Fail when specifying a server_default for ForeignKey, ManyToMany, FileField, ImageField. This is not possible.
  • Internals: _insert and _update have now a different signature.

Fixed

  • ForeignKeys aren't required to be saved models when passed.
  • Cli command revision takes now also the arg argument.
  • Revisioning works now with relative revisions with - (e.g. -2).
  • Downgrades are now possible with unique_together. Build a constraint name from the fields.
  • is_partial was incorrectly set to always False for the model.save update path.
  • FileFields could have make problems when migrating.
  • Differing ruff versions and rules between hatch fmt and pre-commit.

0.27.4

Changed

0.27.3

Added

Fixed

  • BooleanField typing. Thanks to Izcarmt95.

0.27.2

Changed

  • Edgy now allows inheritance of unique_together and indexes from abstract classes.

Fixed

  • Typo in BaseContentTypeField from BaseContentTypeFieldField.

0.27.1

Added

  • SubFactory in the model factory helpers allowing to import ModelFactory objects from other parts of the codebase and apply them directly as dependencies.
  • New template to edgy init migrations for sequencial migration file names.
  • Sequences like calls for ModelFactory.
  • context as new parameter to the Marshalls.

Changed

  • The faker argument was switched out to context. The context parameter however behaves like a faker instance.

Fixed

  • exclude_autoincrement argument applies now also for the sub-factories. Previously we would have need to provide the parameters.

0.27.0

Changed

  • Added a new table naming schema for through models of ManyToMany fields. This requires however setting an explicit table naming schema via the through_tablename parameter.
  • When through_tablename is a string, it is formatted with the field injected as field in the context.

Fixed

  • Multiple ManyFields from the same model to the same target model. Note: this is unrelated to related_name, setting the the related_name would not help.

BREAKING

For now ManyToMany fields must explicitly have a through_tablename parameter in which the table naming schema for the through model is selected. For existing ManyToMany fields you should pass edgy.OLD_M2M_NAMING" so migrations doesn't drop the tables, for new ManyToMany fields edgy.NEW_M2M_NAMING" which prevents clashes of through model names. You may can also rename the through model table via migration and use the edgy.NEW_M2M_NAMING" or simply use a non-empty string for the through_tablename parameter.

0.26.1

Added

  • Add PGArrayField.
  • Add build_and_save(...) build variant to ModelFactory.
  • Add build(save=True) to ModelFactory.

Changed

  • Reorganize fields. Make more lazy.
  • Make models, querysets imports in __init__.py lazy so circular dependency errors arise no more.
  • Change in ModelFactory the default of exclude_autoincrement from False to True. There is a strong demand in having save ready models.

0.26.0

Added

  • __drop_extra_kwargs__ parameter for Models which drops extra kwargs instead passing down. Required for mashalling StrictModel.
  • reference_select parameter in QuerySet.
  • extra_select parameter in QuerySet for creating extra select clauses which can be referenced.
  • ref_foreign_key_fields in meta.

Changed

  • Model allows now like ReflectModel arbitary arguments. Required for reference_select.
  • Use a partly copied model as ProxyModel. This ensures that the logic of ReflectField still works.

Fixed

  • Cleanup of event-loops and threads in run_sync.
  • Fix running shell with auto reflected models.
  • Fix DatabaseNotConnectedWarning in shell.
  • Usage of StrictModels in Mashalls.
  • Different autoincrement field in Marshalls than id.
  • Fix Marshall required check.
  • Fix ReflectModel with exclude_secrets.
  • Fix non-edgy-fields were stripped instead kept.
  • Allow ComputedFields being part of a serialization.

0.25.1

Added

  • Add exclude_autoincrement parameter/class attribute to ModelFactory.
  • Add build_values method to ModelFactory. It can be used to extract the values without a model.
  • Make Registry initialization compatible with sync contexts via with_async_env(loop=None) method.
  • run_sync has now an optional loop parameter.

Changed

  • The Relation mappings doesn't set the autoincrement id anymore.
  • Relationship fields of auto generated ModelFactories of ForeignKey, ... are now excluded when not parametrized. This way large trees are avoided.
  • Reduce the amount of generated related models to max 10 by default.
  • Make to_factory_field and to_list_factory_field a method instead a classmethod. Otherwise they are quite limited.
  • to_list_factory_field honors the min and max parameter specified by parameters. It defaults however to the provided min and max parameters.
  • RefForeignKey has now an extra subclass of BaseField. This way the exclusion of works reliable.
  • run_sync reuses idling loops.
  • run_sync uses the loop set by the Registry contextmanager with_async_env.

Fixed

  • ModelFactory was prone to run in infinite recursions.
  • The RefForeignKey field was not working correctly.

0.25.0

Added

  • Add testing.factory.ModelFactory.
  • ManyToManyField create_through_model method allows now the keyword only argument replace_related_field.
  • add_to_registry and models have now an additional keyword-only argument on_conflict for controlling what happens when a same named model already exists. For models this can be passed : class Foo(edgy.Model, on_conflict="keep"): ....
  • Passing a tuple or list of types to replace_related_field is now allowed.
  • Add through_registry to ManyToMany.
  • Add no_copy to models MetaInfo.
  • Add ModelCollisionError exception.
  • Add keyword only hook function real_add_to_registry. It can be used to customize the add_to_registry behaviour.
  • Add __no_load_trigger_attrs__ to edgy base model to prevent some attrs from causing a deferred load.

Changed

  • create_edgy_model has now __type_kwargs__ which contains a dict of keyword arguments provided to __new__ of type.
  • RelatedField uses now no_copy.
  • add_to_registry returns the type which was actually added to registry instead of None.
  • Through models use now no_copy when autogenerated. This way they don't land in copied registries but are autogenerated again.
  • Instead of silent replacing models with the same __name__ now an error is raised.
  • skip_registry has now also an allowed literal value: "allow_search". It enables the search of the registry but doesn't register the model.
  • Move testclient to testing but keep a forward reference.
  • Change the default for ManyToMany embed_through from "" to False which affected traversing ManyToMany.
  • Better protect secrets from leaking. Prevent load when accessing a secret field or column.

Fixed

  • Copying registries and models is working now.
  • Fix deleting (clearing cache) of BaseForeignKey target.
  • Creating two models with the same name did lead to silent replacements.
  • Invalidating caused schema errors.
  • ManyToMany and ForeignKey fields didn't worked when referencing tenant models.
  • ManyToMany fields didn't worked when specified on tenant models.
  • Fix transaction method to work on instance and class.
  • Fix missing file conversion in File. Move from ContentFile.
  • Fix mypy crashing after the cache was build (cause ChoiceField annotation).
  • Fix handling unknown fields via the generic_field.
  • Sanify default for embed_through which affected traversing ManyToMany. It defaulted to the backward compatible "" which requires the user to traverse the m2m model first.
  • Prevent fully initialized models from triggering a deferred load.
  • Prevent accessing excluded secrets from triggering a deferred load.

BREAKING

  • Instead of silent replacing models with the same __name__ now an error is raised.
  • The return value of add_to_registry changed. If you customize the function you need to return now the actual model added to the registry.
  • The default for ManyToMany embed_through changed from "" to False which affected traversing ManyToMany. For keeping the old behaviour pass: embed_through="".
  • Accessing field values excluded by exclude_secrets doesn't trigger an implicit load anymore.

0.24.2

Fixed

  • Try harder to avoid circular imports when providing settings with edgy references.

0.24.1

Fixed

  • Comparation of a model with non-models.

0.24.0

Added

  • True multi-database migrations. You may need to rework your migrations in case you want to use it.
  • Generalized hash_to_identifier function.
  • get_name function on metadata_by_url dict.
  • Differing databases can be passed via database attribute on models.
  • create method on relations (reverse side of ForeignKeys and both sides of ManyToMany).

Changed

  • Breaking: empty names are not allowed anymore for extra. This includes names consisting of spaces.

Fixed

  • ForeignKey remote check failed for objects with different database but same registry.

0.23.3

Fixed

  • Version number mismatch.

0.23.1 & 0.23.2

Fixed

  • Docs were broken.
  • Autodetection of the instance was broken when a directory containing dots is in the folder.
  • Fix compatibility with pydantic 2.10.

0.23.0

Added

  • Extension support via Monkay.

Changed

  • Rework edgy to use Monkay.
  • Imports are now lazy.
  • Rework the migrate and shell system to simply use Monkay instance.
  • Replace get_registry_copy by get_migration_prepared_registry.
  • Breaking: migration configuration takes place in settings.
  • Breaking: EdgyExtra and Migrate are replaced by edgy.Instance but are still available.
  • Breaking: EdgyExtra is provided by the also obsolete Migrate.
  • Breaking: model_apps is replaced by preloads but still available during the migration time.
  • Breaking: An automatic registration is assumed. See Connection for examples.
  • Breaking: --app or EDGY_DEFAULT_APP must point to a module which does the self-registration not an app instance anymore.
  • Deprecate edgy.conf.enums.EnvironmentType. Esmeralds EnvironmentType or an own definition should be used instead.
  • Deprecate edgy.conf.module_import.import_string. Use monkay.load instead.

Fixed

  • Migrations with ManyToMany fields are broken.
  • get_engine_url_and_metadata was broken for some operations (thanks @kokoserver).
  • IPAddressField was not exposed as edgy.IPAddressField.

Removed

  • edgy.conf.functional. It was only used for configuration and is now superseeded by Monkay.

Contributors

Thanks a lot to @kokoserver. He provided a lot of valuable bug reports and PRs.

0.22.0

Added

  • Global constraints via meta.
  • Allow functional indexes.
  • Expose further parameters for UniqueConstraints.
  • no_copy attribute for fields.

Changes

  • Breaking: Factories pass now the kwargs as dict to get_pydantic_type, get_column_type, get_constraints. This allows now modifying the arguments passed down to the field.
  • Breaking: init_fields_mapping doesn't initializes the field stats anymore.
  • Breaking: model rebuilds are executed lazily when calling init_fields_mapping not when assigning fields manually anymore.

Fixed

  • Indexes and unique_together worked only for fields with columns of the same name.
  • MigrateConfig has no get_registry_copy.
  • Migrations have duplicate fks and crash.
  • ContentTypes were not copyable.
  • VirtualCascade was not automatically enabled.
  • Improve lazyness by splitting in two variable sets.

0.21.2

Added

  • PlaceholderField.
  • StrictModel which forbids extra attributes.

Changed

  • Validate on assignment is enabled.
  • IPAddressField uses now pydantic validation and a simplified TypeDecorator.
  • URLField is now validated.

Fixed

  • Pydantic validators are fixed. Field based ones as well as model based ones.
  • __dict__ was wiped out when intializing a model.

0.21.1

Changed

  • Breaking: from_kwargs doesn't require model or table anymore. It is simply ignored.

Fixed

  • Q, and_, or_ support now complex kwargs like querysets.
  • Failure querying when using proxy model table and kwargs.
  • Proxy and main model use now the same tables. This could have been a problem when filtering against table columns of the proxy table in a query from the main table.
  • Queries operate now always on the main model not the proxy model.
  • Stacklevel of performance warning was wrong.

0.21.0

Added

  • Allow multi schema and database migrations.
  • metadata_by_url dictionary.

Changed

  • metadata of registry is now metadata_by_name.
  • hash_tablekey uses a faster hash function.
  • Migrate object provides now a function for getting a suitable registry copy for migrations.
  • Change license to bsd-3.
  • Proper deprecate unset autoincrement when using it with primary key. The behavior is already deprecated in the documentation.

Fixed

  • Migrate executed refresh_metadata everytime causing reflected models to vanish.
  • Fix edgy wrapping esmerald with the asgi helper causing esmeralds cli to disappear.

0.20.0

Added

  • Add DurationField.
  • Allow passing max_digits to FloatField.
  • Add local_or function to QuerySets.

Changed

  • Only the main table of a queryset is queryable via model_class.columns.foo == foo. Select related models have now an unique name for their path. The name can be retrieved via tables_and_models or using f"{hash_tablekey(...)}_{column}".
  • Breaking: Alter tables_and_models to use the prefix as key with '' for the maintable and model.
  • Breaking: Functions passed to filter functions reveive now a second positional parameter tables_and_models.
  • build_where_clause conditionally uses a subquery.
  • Rename QueryType to QuerySetType. The old name stays as an alias.
  • The debug property of QuerySet named sql inserts now the blanks and uses the dialect.

Fixed

  • Triggering load on non-existent field when reflecting.
  • InspectDB mapping was incorrect.
  • Fix query edge cases.
  • Fix using related queries with update/delete.

0.19.1

Fixed

  • Migration issues with ManyToMany fields.
  • Wrong UUIDField type.
  • Passing options from Migrate to the alembic context.

0.19.0

Added

  • New SET_DEFAULT, and PROTECT to on_delete in the ForeignKey.
  • New through_tablename parameter for ManyToMany.

Removed

  • __db_model__ is removed. Replaced by registry = False.

Changed

  • Allow setting registry = False, for disabling retrieving the registry from parents.
  • Removed unecessary warning for ManyToMany.
  • Add warnings for problematic combinations in ForeignKey.
  • Make QuerySet nearly keyword only and deprecate keywords not matching function names.
  • Clone QuerySet via __init__.
  • Make select_related variadic and deprecate former call taking a Sequence.
  • Improved QuerySet caching.

Fixed

  • Multi-column fields honor now column_name. This allows special characters in model names.

0.18.1

Changed

  • Cleanup Model inheritance: Database related operations are put into a mixin. The customized metaclass is moved from EdgyBaseModel to edgy.Model as well as some db related ClassVars.
  • multi_related is now a set containing tuples (from_fk, to_fk). This can be used to identify fields used by ManyToMany fields.
  • Deprecate is_multi.
  • Deprecate parents. There are no users, it was undocumented and uses are limited.

Fixed

  • Non-abstract through-models wouldn't be marked as many to many relations.
  • Issues related with the Edgy shell initialization.

0.18.0

Added

  • ComputedField.
  • Permission template.
  • reverse_clean for ForeignKeys.
  • Expanded filter methods of querysets (can pass now dict and querysets).
  • Properly scoped with_tenant and with_schema.

Changed

  • Managers use now instance attributes (database, schema).
  • Expose as_select instead of raw_query.
  • model_fields contain now the fields so we can actually use the pydantic magic.
  • BREAKING: deprecate set_tenant and remove manager hack. This way the tenant scope could leak. Use with_tenant instead.

Fixed

  • select_related works across ManyToMany fields.
  • select_related couldn't handle multiple pathes to the same table.
  • select_related would remove valid model instances because of non-existent related objects.
  • Fix identifying clashing column names in joins, so every model gets its right parameters.
  • Dependency tracking for join, so it doesn't depend on the order of select_related.
  • select_related entries work in any order and don't overwrite each other. -only and defer work on select_related.
  • Autogenerated id wasn't added in model_dump.
  • Tenants worked only till the first query.

Breaking changes (upgrade path)

set_tenant("foo")

Becomes now

with with_tenant("foo"):
  ...

activate_schema("foo")
...
deactivate_schema()

Becomes now

with with_schema("foo"):
  ...

0.17.4

Fixed

  • model_dump_json returns right result.
  • show_pk=False can now be used to disable the inclusion of pk fields regardless of __show_pk__.
  • __setattr__ is called after insert/update. We have transform_input already.

0.17.3

Fixed

  • Lazy ManyToMany fields.

0.17.2

Added

  • build_where_clause method in QuerySet for easing the integration in raw SQLAlchemy queries.

Fixed

  • update and delete methods of QuerySet did ignore or clauses.
  • Fixed ManyToMany fields not able to use their owner Model as target.
  • Fixed makemigration failing with foreignkeys when using model_apps.

0.17.1

Added

  • CURRENT_MODEL_INSTANCE ContextVariable which always point to a model instance.

Fixed

  • Under circumstances it was possible for the assigned database attribute to appear as value.
  • Typings of some (class) properties are now correctly detected.
  • FileFields doesn't need an explicit to_file call assignment anymore to work with revisions. This restriction was lifted.
  • Execute database operation of bulk_update in right scope.

0.17.0

Added

  • Support for querying across multiple databases.
  • Support for passing functions as clauses or keyword parameters.
  • Support for autocreated reflection objects by pattern matching.
  • Added some context variables for extract_column_values and transform_input:
  • CURRENT_PHASE: allows retrieving the current context in which it was executed.
  • EXPLICIT_SPECIFIED_VALUES: when set, it returns a set of the keys from the explicitly specified values.

Changed

  • crawl_relationship has a slightly changed interface: it has cross_db_remainder as kwarg for callbacks and in the result. Also it doesn't raise NotImplementedError anymore when detecting a crossdb situation.
  • is_cross_db optionally gets a database as parameter and compares the databases instead of registries.
  • Relax fields parameter requirements for values/values_list.
  • More lazy meta/metadata.
  • database attribute of models is used for queries instead of the main database of the registry.
  • We use more metaclass kwarg arguments.
  • Switch to python >= 3.9.
  • Rename internal _is_init of MetaInfo to _fields_are_initialized.
  • phase argument is shifted to CURRENT_PHASE context_var. If you rely on the correct phase you need to use it instead.
  • extract_column_values provides now also a CURRENT_PHASE environment.
  • is_update argument of get_defaults is now replaced by CURRENT_PHASE too. It is way more accurate.
  • Deprecate force_save kwarg of save in favor of force_insert. This is way more precise.
  • post_save_callback receives now also the force_insert parameter.

Fixed

  • Fix ForeignKey not None but empty hull under some conditions.
  • Fix DecimalField requiring max_digits.

0.16.0

Added

  • ASGI, async context manager support to connect/disconnect multiple dbs at once.
  • create_all/drop_all, create_schema/drop_schema are now capable of initializing dbs in extra.
  • Add transaction helper to Model and QuerySet.
  • Allow copying models properly.
  • Allow None as default.

Changed

  • Unify using to allow setting schema and database via keyword arguments and deprecate both former calls.

Fixed

  • Esmerald typing issues.
  • Migration metadata was not completely initialized which caused problems.
  • Foreign keys None/null handling was inconsistent. Now we have a None for unset foreign keys.

0.15.0

Added

  • FileField and File handling.
  • ImageField stub (more is comming soon)
  • stage method in relations.
  • ModelRefs passed as normal positional arguments are automatically staged.
  • ContentType was added.
  • Proper callback support was added.
  • remove_referenced parameter for ForeignKeys was added.
  • Virtual CASCADE deletion is now used for ForeignKeys without constraint.

Changed

  • Breaking: ModelReferences use now the related name instead of the model name.
  • Breaking: Field Factories pass now keyword arguments as positional dictionary. This way keyword arguments manipulations in validate are possible. This way it can be distinguished between multiple foreign keys to the same model and self-references are possible.
  • model_references are superseeded by post_save_fields in meta.
  • ModelParser mixin is gone. Use the classmethod extract_column_values instead.
  • edgy_settr is not used internally anymore (circular imports).
  • modify_input receives now an argument phase.
  • foreign_key_fields is now a frozenset.
  • Switch away from nest_asyncio.
  • All sqlalchemy operators are now accessable. The setting for them is gone.

Fixed

  • Factory overwrites can now access owner.
  • Foreign keys to reflected models.
  • Handling to string references not in registry loaded yet.
  • Timezone aware Datetime saved in timezone unaware database columns.

0.14.1

Changed

  • Bump dependency of databasez.

Fixed

  • Fix load_recursive.

0.14.0

Added

  • Cache per queryset.
  • Expose raw_query for getting a raw sqlalchemy query.
  • Expose control for used batch size.
  • Add load_recursive for initializing a nested model structure.

Changed

  • Server defaults don't trigger load after saving.
  • Less nested run_sync calls.
  • Alter queryset methods to contain useful arguments, instead of ignoring kwargs.
  • Replace QuerysetProtocol by abstract base type.

Fixed

  • get_or_none was not abidding embed_parent.
  • Correctly apply filters on relations.
  • last and first don't fail without id column anymore.

Removed

  • Unused and broken debug arguments, like raw_query on model classes.
  • Remove run_query of models.

0.13.1

Added

  • Testclient can be configured via environment variables.

Fixed

  • Compatibility with newer databasez releases (>0.9).

0.13.0

Added

  • default_timezone, force_timezone, remove_timezone for DateTimeField.
  • default_timezone, force_timezone for DateField.
  • Add attribute inject_default_on_partial_update.
  • Allow factories overwriting field methods.
  • Deep embedding via embed_parent possible.
  • Support for nearly all async sqlalchemy drivers plus more from databasez.

Changed

  • get_default_values has now an extra keyword argument is_update
  • meta.fields_mapping was renamed to meta.fields.-
  • Factories have now different named configuration variables.
  • Refactoring of fields and models:
  • BaseModelType and BaseFieldType are now added and should be used for typings.
  • get_column works now via extractor.
  • Renamed internals.
  • Splitted model_references and column values extraction.
  • Comparisons of models use now only primary keys.
  • Deprecate on model forwards signals, fields. Use now meta.signals, meta.fields.

Removed

  • execsync (nearly no uses).
  • DateParser mixin (superseeded by sqlalchemy's logic).

Fixed

  • auto_now and auto_now_add now also work for date fields.
  • BinaryField had a wrong type.
  • Metaclasses with keyword arguments are now possible.
  • Relaxed Prefetching, tables can now appear multiple times.
  • select_related now correctly filters columns.

0.12.0

Added

  • Allow multiple ForeignKeys with the same related_name in one model.
  • Support ForeignKeys on multiple columns.
  • Allow selecting columns for ForeignKeys.
  • Allow skipping creating a ForeignKeyConstraint.
  • Add ExcludeField for masking fields in submodels.
  • ManyToMany fields pass unique through to target foreignkey.
  • Add ConditionalRedirect constant for CompositeField.
  • Embeddables via CompositeField. See Embedding
  • Add ForeignKeyFactory, a factory with presets for building foreign keys.
  • Multiple primary keys and names different from "id" are possible now.
  • Add inherit flag for Manager, BaseFields and Models (when used as an embeddable). It is used for controlling the inheritance.
  • Managers are now instance aware. You can customize the instance and they can react. They are also shallow copied for every class and instance.
  • Improved Relations (reverse side of ForeignKeys and forward side of Many2Many). Have now add and remove methods and work like RefForeignKey (you can just specify an Array with assignment targets and they will be added).
  • Allow skip building reverse RelatedFields for ForeignKeys with related_name=False.
  • pkcolumns attribute of models (contains all found primary key columns).
  • Some new methods on BaseField:
  • embed_field: for controlling embedding a field in an CompositeField.
  • get_column_names: helper function for retrieving the column names of a field.
  • Add RelationshipField for traversable fields.

Changed

  • ForeignKeys use now global constraints and indexes.
  • Breaking: clean has now the argument to_query. See Custom Fields.
  • Breaking: ManyToMany doesn't have a RelatedField on owner anymore and uses proxying. See ManyToMany.
  • Breaking: use singular related_name for unique ForeignKeys (or OneToOne). See related_name
  • MetaInfo (meta) is now partly lazy.
  • pk is now a PKField (a variant of the BaseCompositeField).
  • clean and to_columns of BaseField do return empty objects instead of raising NotImplementedError.
  • Major refactory of ForeignKeys, move logic for single ForeignKeys to subclass.
  • Move FieldFactory and ForeignKeyFieldFactory to factories.
  • Remove superfluous BaseOneToOneKeyField. Merged into BaseForeignKeyField.
  • Remove unused attributes of MetaInfo and added some lazy evaluations for fields.

Breaking

  • Prefetch traversal of foreign keys uses now the foreign key name. For the traversal of RelatedFields everything stays the same.

0.11.1

Added

  • create_schema allowing a dynamic and programatically way of creating schemas in the database.

Changed

  • Updated reflection from edgy.ReflectModel to use the internal database session and avoid deadlocks.
  • Internal cleaning for tenancy.

0.11.0

Added

  • Support for Marshall allowing custom serialization of an Edgy model by @tarsil.
  • Support for the new CompositeField enhancing the ability of having multiple primary keys (or composed keys) in an Edgy model by @devkral.
  • Support for the Q queryset clause by @devkral.

Changed

  • Cleaned up FieldFactory internals by @devkral.
  • pyproject.toml definitions by @devkral.
  • Internal Edgy model_dump covering internals of Edgy and corner cases of Pydantic for compatibility reasons by @devkral.

Fixed

  • Typos in is_primary key attribute by @devkral.
  • InspectDB when provided with a schema name was not using it in the registry.

0.10.1

Added

  • Support for list and tuples as a type for model_apps.

0.10.0

Added

  • Support for model_apps inside the Migrate object allowing global discovery by application. This will make sure all apps will be properly inspected.
  • Add documentation about the new model_apps

Changed

  • Upgrade internal requirements.

0.9.2

Changed

  • Update internal anyio dependency.

0.9.1

Changed

  • Upgrade internal requirements.

Fixed

  • auto_now and auto_now_add on save() and update() wasn't only updating the field with auto_now.
  • Extraction of the default field for date and datetime.

0.9.0

Added

  • Tenancy for internal proxy_model validation added.
  • Support for ManyToMany to accept strings to the to attribute.
  • Allowing querying inner foreign keys without needing to use select_related or load

Changed

Fixed

  • When generating a many to many through model, the maximum length is enforced to be 63 characters.
  • Allow ManyToMany to also accept a string as a parameter for the to.
  • Object discovery for intellisense in any IDE.

0.8.1

Added

  • Added new experimental activate_schema for tenant models using the using queryset operator.

Fixed

  • Multiple join tables were not generating the complete join statement when using select_related.
  • Fixed metaclass for TenantMixin making sure all the queries are correctly pointing to the right tenant.

0.8.0

Added

  • Support for sync queries. This will enable Edgy to run in blocking frameworks like Flask, bottle or any other by using the newly added run_sync. #60.

Fixed

  • Fixed multi tenancy from contrib. #58.
  • Fixed using where schema name was raising a not found reference for foreign key when querying the tenant. #58.
  • Fixed exclude_secrets when querying models with foreign keys. #59.

0.7.1

Fixed

  • exclude_secrets quick patch for the way it can recursively understand the model fields when filtering the query.

0.7.0

Added

  • Support for or_, and_ and not_ for SQLAlchemy style queries and Edgy syntax sugar queries.
  • Support for secrets field and queryset allowing to query the fields that are not marked as secret directly.

Changed

  • Updated internal support for databasez 0.7.0 and this fixes the URL parsing errors for complex passwords caused by the urlsplit.

0.6.0

Changed

  • inspectdb is now handled by an independent isolated called InspectDB.

Fixed

  • server_default does not raise a FieldValidationError.
  • server_default added as validation for nullable.
  • null was not being applied properly from the newly added server_default.

Warning

This could impact your migrations, so the advise would be to generate a new migration after upgrading to the new version of Edgy to make sure the database reflects the proper nullables/non-nullable fields.

0.5.2

Changed

Fixed

  • Database object docstring.

0.5.1

Changed

  • Base metaclass that is now generating the initial annotations for validations.
  • Update Pydantic version.
  • Update documentation to reflect the new fixed issues.

Fixed

  • Raise ImproperlyConfigured for managers that are not typed as ClassVar avoiding maximum recursion error. #19.

0.5.0

Added

  • Added support for Python 3.12.

Changed

  • Update base requirements.

Fixed

  • Removed copy of docs folder
  • minimum_length to min_length.
  • values_list() to implement the queryset.clone().
  • Fixed .json() to model_dump_json().

0.4.0

Added

  • New Prefetch support allowing to simultaneously load nested data onto models.
  • New Signal support allowing to "listen" to model events upon actions being triggered.

Changed

  • Updated pydantic and alembic

0.3.1

Fix

  • DecimalField definition raising unknown constraint max_digits.
  • DeclarativeModel generating internal mappings names was breaking for class objects.

0.3.0

Added

  • Multi tenancy support by updating the registry and allowing to create the multi schema.
  • Add new using(schema=...) and using_with_db(database=..., schema=...) to querysets.
  • Add support for create_schema and drop_schema via registry.
  • Add support to get_default_schema from the registry.schema.
  • Documentation for tenancy.
  • Improved the documentation for schemas.
  • Added a new parameter extra to registry allowing to pass a Dict like object containing more database connections. This is an alternative to the registries.
  • Improved documentation for registry explaining how to use the extra parameters. and query them.

Changed

  • Update the build for Model and ReflectModel to allow passing the schema.

Fixed

  • Registry metaclass wasn't reflecting 100% the schema being passed into the metadata and therefore, querying the database public schema.

0.2.0

Added

  • RefForeignKey. A model reference where it can create sub objects from the main object directly.
  • Performance enhancements in general.

Changed

  • The way fields are evaluated and caching for the __model_reference__ fields.

Fixed

  • Improved performance of the __proxy_model__ by caching.

0.1.0

This is the initial release of Edgy.

Key features

  • Model inheritance - For those cases where you don't want to repeat yourself while maintaining integrity of the models.
  • Abstract classes - That's right! Sometimes you simply want a model that holds common fields that doesn't need to created as a table in the database.
  • Meta classes - If you are familiar with Django, this is not new to you and Edgy offers this in the same fashion.
  • Managers - Versatility at its core, you can have separate managers for your models to optimise specific queries and querysets at ease.
  • Filters - Filter by any field you want and need.
  • Model operators - Classic operations such as update, get, get_or_none, bulk_create, bulk_update, values, values_list, only, defer and a lot more.
  • Relationships made it easy - Support for OneToOne, ForeignKey and ManyToMany in the same Django style.
  • Constraints - Unique constraints through meta fields.
  • Indexes - Unique indexes through meta fields.
  • Native test client - We all know how hard it can be to setup that client for those tests you need so we give you already one.