- Improvement: Optimize out inspect calls when keys are not found (#84)
- KV source bug fixes (#75)
- Doc improvements
- Bug Fix: Depend on
ecto_sql. Fixes Ecto 2.x support (#67)
- Improvement: Ecto 3 support (which is just improving the tests)
- Improvement: Better error messages if you pass a non-ecto schema to
- Bug Fix: Fix dialyzer spec for run_batch function
- Bug Fix: Fix loading of nested through associations (#65)
- Bug Fix: Poor supervisor structure has been improved, which fixes large memory usage issues
Enhancement: Improved error handling when sources fail to load This provides two additional configurable methods of error handling (these options can be passed to
:return_nil_on_error- This is the previous default. Errors are logged, and values return
:raise_on_error- this will raise a
Dataloader.GetErrorwhen one of the
getmethods are called. This is now the default behaviour for handling errors
:tuples- this changes the
get_many/4methods to return
:errortuples instead of just the value. This frees up the caller to handle errors any way they see fit
Enhancement: Improved caching characteristics on the KV source
Enhancement: More flexible cardinality mapping for Ecto source
Enhancement: Uniq the batched KV and Ecto values
Bug Fix: When using the Ecto source it properly coerces all inputs for known fields.
Note: when upgrading it may be necessary to clear out your
_build directory so
that the compiler picks up the protocol change.
- Errors are raised instead of silently returning nil. See
:return_nil_on_erroroption to keep the previous behavior.
Dataloader.get/4will raise an error if you pass in an
item_keythat was not previously loaded. See https://github.com/absinthe-graphql/dataloader/issues/52 for discussion.
- Enhancement: Custom batch functions for Ecto
- Bug Fix: Ecto source properly caches results now.
- Initial release