Additional fields and type inferring for union #1834

AlbertvanHouten · 2025-08-20T13:46:47Z

Summary

This pull request introduces robust support for Python Union types in the experimental Datumaro type registry and dataset schema inference. It enables seamless conversion between multiple candidate types (including both typing.Union and modern A | B syntax), with fallback logic and comprehensive test coverage. The changes also improve image type conversion and schema inference for datasets, making the system more flexible and reliable.

Type registry and conversion improvements

Added full support for Union types in the type registry: both typing.Union and Python 3.10+ A | B syntax are now handled, with fallback to subsequent types if the first conversion fails. This includes updated logic in from_polars_data and new tests for ordering, error handling, and fallback behavior. [1] [2] [3]
Added comprehensive tests for type registry conversions, including basic types, union types, error cases, ordering, and converter functionality for numpy and torch tensors.

Dataset and schema inference enhancements

Improved schema inference in Dataset to resolve string annotations to actual type objects, supporting cases where from __future__ import annotations is used, and added correct handling for Union types to preserve the original annotation. [1] [2]
Updated type variable definitions and method signatures in dataset.py for clarity and correctness, and removed unnecessary imports. [1] [2] [3]

API and import improvements

Updated the experimental module’s public API to expose new converters, dataset classes, fields, schema types, and registry functions.

Test coverage

Added targeted tests for union type handling in dataset samples, ensuring both modern and legacy union syntax are supported.

These changes significantly improve the flexibility and reliability of type conversion and schema inference in Datumaro’s experimental pipeline.

How to test

Checklist

I have added unit tests to cover my changes.
I have added integration tests to cover my changes.
I have added the description of my changes into CHANGELOG.
I have updated the documentation accordingly

License

I submit my code changes under the same MIT License that covers the project.
Feel free to contact the maintainers if that's a concern.
I have updated the license header for each file (see an example below).

# Copyright (C) 2025 Intel Corporation
#
# SPDX-License-Identifier: MIT

src/datumaro/experimental/dataset.py

Copilot

Pull Request Overview

This pull request introduces comprehensive support for Python Union types in Datumaro's experimental type registry and dataset schema inference. It enables seamless conversion between multiple candidate types using both modern (A | B) and legacy (typing.Union) syntax, with fallback logic when conversions fail.

Key changes include:

Union type support in the type registry with fallback behavior
Enhanced schema inference for string annotations and Union types
New image type conversion capabilities
Expanded API with additional fields and converters

Reviewed Changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 13 comments.

Show a summary per file

File	Description
`tests/unit/experimental/test_type_registry.py`	New comprehensive test suite for type registry functionality
`tests/unit/experimental/test_dataset.py`	Added Union type handling tests for dataset samples
`src/datumaro/experimental/type_registry.py`	Core Union type support and image conversion functionality
`src/datumaro/experimental/legacy.py`	Minor type annotation fix
`src/datumaro/experimental/fields.py`	New annotation field classes and helper functions
`src/datumaro/experimental/dataset.py`	Enhanced schema inference and type resolution
`src/datumaro/experimental/converters.py`	New ImageTypeConverter for image format transformations
`src/datumaro/experimental/__init__.py`	Updated public API exports

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}

src/datumaro/experimental/legacy.py

src/datumaro/experimental/fields.py

src/datumaro/experimental/type_registry.py

src/datumaro/experimental/fields.py

gdlg

Thanks Albert! First round of review, I still need to review the part on the type union.

src/datumaro/experimental/converters.py

src/datumaro/experimental/dataset.py

gdlg · 2025-08-21T09:18:50Z

src/datumaro/experimental/dataset.py

-DType = TypeVar("DType", bound=Sample, default=Sample)
-DTargetType = TypeVar("DTargetType", bound=Sample, default=Sample)
+DType = TypeVar("DType")
+DTargetType = TypeVar("DTargetType")


Add a comment in the code explaining why removing them. Mention the issue with Dataset(schema) too.

src/datumaro/experimental/dataset.py

src/datumaro/experimental/fields.py

src/datumaro/experimental/legacy.py

src/datumaro/experimental/type_registry.py

Co-authored-by: Copilot <[email protected]>

…up that appears to be unused

…ert/otx-integration

gdlg · 2025-08-22T10:14:35Z

src/datumaro/experimental/fields.py

+            if isinstance(value, np.ndarray):
+                value_list = value.tolist()
+            elif isinstance(value, list):
+                value_list = value


Convert to Numpy instead of to a list. You can use the to_numpy function from type_registry.py

gdlg · 2025-08-22T10:16:27Z

src/datumaro/experimental/fields.py

+            return {name: pl.Series(name, [value_list], dtype=pl.List(self.dtype))}
+
+        # Handle single integer value
+        elif isinstance(value, (int, np.integer)):


Suggested change

elif isinstance(value, (int, np.integer)):

else:

In this case, you can assume that we are working with a single label. No need to check the input type.

gdlg · 2025-08-22T10:17:05Z

src/datumaro/experimental/fields.py

+            if target_type == np.ndarray or target_type is np.ndarray:
+                return np.array(data, dtype=np.int64)
+            elif target_type is list:
+                return data


Same here, use from_polars_data from type_registry.py

gdlg

LGTM, thanks Albert!

Additional fields and type inferring for union types.

7135fa0

AlbertvanHouten commented Aug 20, 2025

View reviewed changes

src/datumaro/experimental/dataset.py Show resolved Hide resolved

AlbertvanHouten added 2 commits August 20, 2025 15:49

Remove type checkign for torch/pilimage

6c85e96

Remove comment

5e712f2

AlbertvanHouten requested a review from Copilot August 20, 2025 13:58

Copilot AI reviewed Aug 20, 2025

View reviewed changes

Improve optional typing in fields.py

279f0b2

gdlg reviewed Aug 20, 2025

View reviewed changes

src/datumaro/experimental/converters.py Outdated Show resolved Hide resolved

src/datumaro/experimental/dataset.py Show resolved Hide resolved

src/datumaro/experimental/dataset.py Outdated Show resolved Hide resolved

AlbertvanHouten added 3 commits August 21, 2025 09:28

Remove fields other than those necessary for classification

a56c4d7

Remove imports from fields that were deleted.

e17850c

Fix imports in dataset.py

ac35626

AlbertvanHouten marked this pull request as ready for review August 21, 2025 08:49

Remove bound from DType.

e030ce3

gdlg reviewed Aug 21, 2025

View reviewed changes

AlbertvanHouten and others added 6 commits August 21, 2025 14:30

Address PR comments.

a4789d7

Fix failing tests

8e95a7a

Update src/datumaro/experimental/type_registry.py

768abe7

Co-authored-by: Copilot <[email protected]>

Add converter import back with # type: ignore to prevent import clean…

acfb046

…up that appears to be unused

Merge remote-tracking branch 'origin/albert/otx-integration' into alb…

730f830

…ert/otx-integration

Remove conversion to uint8 and fix media_converter variable assignment

9d82682

gdlg reviewed Aug 22, 2025

View reviewed changes

AlbertvanHouten added 4 commits August 22, 2025 14:01

Simplify to/from polars methods

795c8d8

Remove .tolist()

3a174e2

further simplifying of Labelfield

54d570c

Remove unnecessary if statement for multi label

ec296a3

gdlg approved these changes Aug 22, 2025

View reviewed changes

AlbertvanHouten merged commit 6417998 into develop Aug 22, 2025
15 checks passed

gdlg deleted the albert/otx-integration branch August 28, 2025 09:55

leoll2 mentioned this pull request Nov 20, 2025

Support common annotation types #1787

Closed

6 tasks

Additional fields and type inferring for union #1834

Additional fields and type inferring for union #1834

Uh oh!

Conversation

AlbertvanHouten commented Aug 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Type registry and conversion improvements

Dataset and schema inference enhancements

API and import improvements

Test coverage

How to test

Checklist

License

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull Request Overview

Reviewed Changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gdlg left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gdlg Aug 21, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

gdlg Aug 22, 2025

Choose a reason for hiding this comment

Uh oh!

gdlg Aug 22, 2025

Choose a reason for hiding this comment

Uh oh!

gdlg Aug 22, 2025

Choose a reason for hiding this comment

Uh oh!

gdlg left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

AlbertvanHouten commented Aug 20, 2025 •

edited

Loading