Version 0.36.1

Fixed

• Fix copying model without a registry and adding it to one when its meta.registry equals None.
• Fix CompositeField handling of Mappings.

Version 0.36.0

Added

• Support for unconstraint numeric mode of postgresql for DecimalField.
• Add the bulk_update_or_create operation to querysets.
• Add the update_embed_parent operation to querysets.
• Add the ignore_conflicts parameter to bulk_create.
• Add relationship signals (pre/post_relation_add and pre/post_relation_remove).
• Add bulk signals (pre/post_bulk).
• Add SkipOperation exception for signals.
• Add injected_filters parameter for pre_delete to dynamically inject protection rules.
• Allow lazy references for ForeignKey and ManyToMany. This is useful for libraries.
• Add dirty tracking.

Changed

• DecimalField supports now also None for decimal_places and max_digits.
  It emits an UserWarning and falls back to None for each of both parameters not provided.
• bulk_get_or_create mirrors now get_or_create and returns a list with (instance, created) tuples. The returned instances are not always complete.
• CombinedQuerySet is moved from edgy.core.db.queryset.mixins.combined to edgy.core.db.queryset.combined. It is not a mixin.
• The bulk operations are moved to a mixin.
• The typings changed for QuerySet: it is now a Generic.
• The typings changed for QuerySetType: EdgyEmbedTarget and EdgyModel (the queryset model) are switched in the Generic definition.
• Bulk operations are now keywords only (except the first objs parameter).
• Dedupe bulk_create, bulk_get_or_create, and bulk_update_or_create inputs.
• Refactor QueryExecutor so it can update itself.
• real_save, raw_delete and delete are now keyword only to prevent bad overwrites.
• Remove the skip_post_delete_hooks parameter from delete as it is for internal use and shouldn't be exposed.
• (Internal detail only) bypass the overwritten __getattr__, __getattribute__ and __setattr__ for speed improvements.

Fixed

• Relations did not use the tenancy/used schema properly when querying.
• Bulk operations with reflected fields did not always work properly.
• run_concurrently now properly cleans up not executed coroutines in case of an error.
• SuspiciousFileOperation inherits now correctly from EdgyException.
• Return row count for update.
• Fix get_created_time, get_modified_time and get_accessed_time on filesystem storage. Previously it used a non-existing setting and didn't follow the specs.
• Comparing reflected models.
• Directly issuing queries on through models of ManyToMany.

Version 0.35.11

Fixed

• Improved typings for QuerySet.
• bulk_get_or_create supports now officially also other types than lists and is exposed in QuerySetType.
• Update .gitignore.

Version 0.35.10

Added

• Add explicit SQLAlchemy compatibility mode via SQLAlchemyModelMixin for opted-in models.
  This enables SQLAlchemy Core expressions such as select(Model.id).where(Model.id == value).
• Add foreign key scalar alias support in compatibility mode (for example Model.owner_id).
• Compatibility mode can be declared once on an abstract base model and inherited by concrete models.

Fixed

• Preserve existing Edgy query behavior for non-opted-in models and relationship handling.

Version 0.35.9

Fixed

• Missing table_prefix when ManyToMany is declared.

Version 0.35.8

Added

• where, select, select or none
  and insert as idiomatic alternatives to filter, get, get_or_none and create.
• table_prefix is now allowed in the Meta of models. This allows to prefix tables and inherit any prefix to related models.
  This is useful for multi-tenancy and sharding.

Fixed

• Contrib parsing of UUID was not working correctly.
• CSS of Edit of a model.

Version 0.35.7

Changed

• Fix internal typing search for SQLAlchemy.

Version 0.35.6

Changed

• Fix queryset bulk calls with empty iterable.
• Rename to esmerald references to ravyn.
• Update import references.

Version 0.35.5

Fixed

• Resolved duplicate row issues when combining Q expressions with | (global OR).
  Edgy now correctly collapses duplicates at hydration level for non-embedded querysets.
• .count() now returns the correct number of logical rows when OR or join-fan-out is present
  (uses COUNT(DISTINCT pk) automatically when required).
• Fixed incorrect behaviour where OR clauses polluted .exists() uniqueness checks.
• Ensured select_related() does not introduce duplicate rows under OR or complex joins.
• Fixed values() and values_list() returning duplicated logical rows when OR joins were involved.
• first() and last() now operate on deduped results in OR scenarios.
• Prevented ManyToMany and embedded relation querysets from being deduped incorrectly
  (avoid breaking album.tracks.all() and similar accessors).
• Stabilised internal OR/local_OR clause handling to ensure consistent join inference.

Version 0.35.4

Added

• Support for queryset_class in the managers. This allows to override the default Queryset with a custom one.
  This is useful when we need to have nested querysets that relies on an initial manager to be custom.

Example

from edgy import Manager, QuerySet

class CustomQuerySet(QuerySet):

  def filter(...):
    ...


class CustomManager(Manager):
  queryset_class = CustomQuerySet