Allow type_hooks for substituted objects of the same parent type #219
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We've had issues with dacite not supporting
type_hooks
ondatetime
during testing, whenfreezegun
is monkey patching thedatetime
object to beFakeDatetime
, see here.At first it was assumed that the dataclass attribute was monkey patched and therefore the
type_hooks
list doesn't have a match for the attribute's type. However this is not the case, as the interpreter would have already interpreted the type annotation before the monkey patch takes place, taking a reference to the original object.It is instead the
type_hooks
' key (which for clarity is not a string but instead the data type itself) that is patched. Thus on structuring data into a dataclass, thetype_hooks
dict contains aFakeDatetime
key to anisoparse
hook.Last bit of context needed before going to the proposed fix: the way type hooks are matched with dataclass's attributes's types in dacite today is by
in
, which is like iterating over the keys and doing a value comparison (==
, notis
). This does however not allow us to have objects to be replaced following the Liskov Principle, whereby we can substitute an object of one type by another so long as they share the same parent (simplified for brevity, sufficient for this context).Coming to the proposal: If we change the way we compare type_hooks with "is a ..." logic instead of "==" logic, we can substitute objects that should provide the same behaviour.
Getting back to the original problem, freezegun. Since Python allows us to redefine the behaviour of objects via meta classes, we can create objects that aren't actual drop-ins but will still declare themselves as a subclass of another type. This is exactly what freezegun is doing here. This object is not a subclass of
datetime.datetime
by inheritance, instead by explicitly overriding the comparison made byissubclass
. With the proposed change, this means we can monkey patchdatetime.datetime
toFakeDateTime
and still have a matching type hook in thetype_hooks
dict. I imagine this will be helpful for a lot more cases where monkey patching is happening with objects defined in a similar way.