🐛 Fix support for Annotated fields with Pydantic 2.12+

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕

[stickm4n]

After pydantic update to support Python 3.14, different declaration syntax were affected by the changed.
I check and on pydantic side there is not changes on the resulting FieldInfo structure, at least at first glance.
In the tests I illustrated the 3 possible declaration syntax for annotations defined by pydantic and supported so far by sqlmodel. Two of them do not work any more with pydantic>=2.12!

[stickm4n]

solves #1602

[wguilen]

Any idea on when a new release with this fix will be available?

[stickm4n]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon!
I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

[wguilen]

Any idea on when a new release with this fix will be available?

Not really, hopefully soon! I would still recommend to use pydantic<2.12 if python 3.14 is not required... there have been big changes recently that could have included some malfunctions like this

That's alright, I'll follow your suggestion. Thanks for the quick reply!

[svlandeg]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

[stickm4n]

Hi @stickm4n, thanks for the PR! I'm going to push some changes to your branch directly, to include also the tests from #1602 and to get the PR in a good state to merge. I'll put it in draft while I work on it.

Perfect, feel free to do it!
It will be nice for this to me merged soon.

[svlandeg]

Thanks for the work on this, @stickm4n!

This now looks good to merge to me. I've added in the tests from #1602 just for good measure, and refactored the code a bit using Yurii's suggestions. I think it's much easier to follow/maintain like this, without the continue & break statements.

Will yet run this by Tiangolo for a final review.
Note that another alternative to this PR is #1607 - if we decide to merge this, we can close the other one.

[KSmanis]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

[stickm4n]

FYI PR #1607 works with __future__.annotations as well, whereas this one does not.

Can you provide more details on this?

[KSmanis]

Sure, I tested the following model with both PRs, with and without future annotations in Python 3.13:

classCreateTimeMixin:
create_time: Annotated[AwareDatetime, Field(sa_type=DATETIME_TIMEZONE)]

This PR failed to pick up on the annotated metadata when future annotations were enabled.

[sr-verde]

@svlandeg, @tiangolo: Is there any chance we get this or #1607 merged soon?

[zhcn000000]

merge it soon may be helpful for pydantic 2.12+

[tiangolo]

Thank you @stickm4n! And thanks everyone for the discussion here.

I went with #1607 because that approach, moving the metadata to its own annotation will allow us (in the future) to stop sub-classing FieldInfo, which the Pydantic team have asked for.

I also added a test with stringified annotations (from __future__ import annotations) and that PR fixes/passes that test. ✔️

The fix will be available in the next few hours in SQLModel 0.0.32 🚀

Given that, I'll now close this one. Thanks! ☕