refactor(tests): auto-convert array-api-strict modules

[njzjz]

Summary:

• add a dpmodel-to-array-api-strict wrapper registry with lazy module registration
• replace repeated array_api_strict descriptor/fitting/utils setattr conversions with shared conversion logic
• convert residual data lists during append/assignment for DPA2/DPA3 strict wrappers

Tests:

• ruff check .
• ruff format --check .
• pytest source/tests/consistent/descriptor/test_dpa3.py -k array_api_strict -q
• pytest source/tests/consistent/descriptor/test_dpa2.py -k array_api_strict -q
• pytest source/tests/consistent/descriptor -k array_api_strict -q
• pytest source/tests/consistent/fitting -k array_api_strict -q

Summary by CodeRabbit

• New Features

  • Introduced centralized conversion mechanism for standardizing array handling across model objects with registry-based mappings and lazy module imports.
  • Added support for five additional descriptor types.
  • Enhanced list-value conversion with optional preservation of original list structure.

• Refactor

  • Simplified descriptor and fitting network classes by consolidating array conversion logic into a unified decorator-based approach.
  • Converted dynamically generated network classes to explicit class definitions for improved clarity.

[github-advanced-security]

CodeQL found more than 20 potential problems in the proposed changes. Check the Files changed tab for more details.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR refactors the array_api_strict testing module to centralize object conversion logic. Instead of individual __setattr__ overrides in descriptor, fitting, and utility classes, a registry-based decorator system (@array_api_strict_module) and conversion functions automatically wrap dpmodel objects into strict array-api types during attribute assignment.

Changes

Central conversion framework

Layer / File(s)	Summary
Array conversion function updates and module state source/tests/array_api_strict/common.py	to_array_api_strict_array return type is clarified and docstring updated; module-level registry, lazy registration tracking, and ArrayAPIList (a list subclass that auto-converts items) are added.
Conversion detection and lookup utilities source/tests/array_api_strict/common.py	Public register_dpmodel_mapping API, internal predicates for object classification, lazy registration loader _ensure_registrations, and try_convert_module for registry lookups are implemented.
Central conversion entry points and attribute setters source/tests/array_api_strict/common.py	List conversion helper _try_convert_list, core entry point convert_array_api_strict_value, attribute setter array_api_strict_setattr (with special handling for data-list attributes), and the array_api_strict_module decorator that wires classes and auto-registers dpmodel mappings are added.

Descriptor module simplification and centralization

Layer / File(s)	Summary
Descriptor namespace expansion and DPA1 base refactoring source/tests/array_api_strict/descriptor/__init__.py, source/tests/array_api_strict/descriptor/dpa1.py	Five new descriptor classes (DPA2, DPA3, SeAttenV2, SeT, SeTTebd) are added to namespace exports; DPA1 module classes are decorated with @array_api_strict_module and all custom __setattr__ implementations are removed, becoming thin pass subclasses.
DPA2 and DPA3 descriptor simplification source/tests/array_api_strict/descriptor/dpa2.py, source/tests/array_api_strict/descriptor/dpa3.py	DPA2 and DPA3 descriptors are simplified from coercion/deserialization implementations to minimal @array_api_strict_module-decorated pass subclasses inheriting from their DP counterparts.
SE descriptor variant refactoring source/tests/array_api_strict/descriptor/se_atten_v2.py, source/tests/array_api_strict/descriptor/se_e2_a.py, source/tests/array_api_strict/descriptor/se_e2_r.py, source/tests/array_api_strict/descriptor/se_t.py, source/tests/array_api_strict/descriptor/se_t_tebd.py	SE variants (SeAttenV2, SeA, SeR, SeT, SeTTebd) remove setattr overrides and adopt @array_api_strict_module decoration; SeAttenV2 adds dpmodel mapping registration for serialize/deserialize round-trips.
Hybrid and specialized descriptor refactoring source/tests/array_api_strict/descriptor/hybrid.py, source/tests/array_api_strict/descriptor/repflows.py, source/tests/array_api_strict/descriptor/repformers.py	Hybrid, repflows, and repformers modules remove custom setattr implementations and apply @array_api_strict_module decoration; RepflowLayer and RepformerLayer add _array_api_strict_data_list_attrs declarations for residual list attributes.

Utility modules simplification

Layer / File(s)	Summary
Exclude mask utility simplification source/tests/array_api_strict/utils/exclude_mask.py	AtomExcludeMask and PairExcludeMask remove setattr attribute conversion and are decorated with @array_api_strict_module.
Network utility refactoring with explicit classes and registration source/tests/array_api_strict/utils/network.py	NativeNet, EmbeddingNet, FittingNet are refactored from factory-produced module-level assignments to explicit @array_api_strict_module-decorated classes; Identity class is added; NativeLayer retains specific attribute coercion (w, b, idt); dpmodel mappings are registered for all network types via serialize/deserialize round-trips.
Type embedding utility simplification source/tests/array_api_strict/utils/type_embed.py	TypeEmbedNet removes setattr deserialization logic and is decorated with @array_api_strict_module.

Fitting module simplification

Layer / File(s)	Summary
Fitting namespace export and class simplification source/tests/array_api_strict/fitting/__init__.py, source/tests/array_api_strict/fitting/fitting.py	Five fitting network classes (EnergyFittingNet, PropertyFittingNet, DOSFittingNet, DipoleFittingNet, PolarFittingNet) are re-exported in __init__.py; fitting.py removes the setattr_for_general_fitting helper and all setattr overrides, replacing them with @array_api_strict_module-decorated pass subclasses.

Estimated code review effort

🎯 4 (Complex) | ⏱️ ~60 minutes

Possibly related PRs

• deepmodeling/deepmd-kit#5204: Both PRs establish registry-driven dpmodel-to-wrapper conversion infrastructure and centralized setattr/wrapping flows for automatic strict-type conversion.

Suggested labels

enhancement, Python

Suggested reviewers

• wanghan-iapcm
• iProzd

🚥 Pre-merge checks | ✅ 4 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 21.05% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (4 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'refactor(tests): auto-convert array-api-strict modules' accurately summarizes the main change: centralizing array-api-strict conversion logic through an auto-conversion system.
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

🧹 Nitpick comments (1)

source/tests/array_api_strict/common.py (1)

160-167: 💤 Low value

Minor: double conversion when keep_converting=True.

When converted is an ArrayAPIList, calling converted.append(converted_item) invokes the overridden append that calls convert_array_api_strict_value again on an already-converted item. The conversion is idempotent so this is correct, but wasteful.

♻️ Optional: bypass the override for initial population

 def _try_convert_list(value: list[Any], *, keep_converting: bool = False) -> list[Any]:
     converted = ArrayAPIList() if keep_converting else []
     changed = keep_converting
     for item in value:
         converted_item = convert_array_api_strict_value(item)
-        converted.append(converted_item)
+        list.append(converted, converted_item)
         changed = changed or converted_item is not item
     return converted if changed else value

🤖 Prompt for AI Agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

In `@source/tests/array_api_strict/common.py` around lines 160 - 167, The loop in
_try_convert_list uses converted.append on an ArrayAPIList when
keep_converting=True, triggering ArrayAPIList.append which re-calls
convert_array_api_strict_value and causes a redundant conversion; fix by
collecting converted items into a temporary plain list (e.g., temp = []) inside
the loop (use temp.append(converted_item) and update changed there), and after
the loop, if keep_converting is True build the ArrayAPIList once from that temp
(converted = ArrayAPIList(temp)) so the overridden append is not invoked for
each item; leave the return logic (return converted if changed else value)
unchanged.

🤖 Prompt for all review comments with AI agents

Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.

Nitpick comments:
In `@source/tests/array_api_strict/common.py`:
- Around line 160-167: The loop in _try_convert_list uses converted.append on an
ArrayAPIList when keep_converting=True, triggering ArrayAPIList.append which
re-calls convert_array_api_strict_value and causes a redundant conversion; fix
by collecting converted items into a temporary plain list (e.g., temp = [])
inside the loop (use temp.append(converted_item) and update changed there), and
after the loop, if keep_converting is True build the ArrayAPIList once from that
temp (converted = ArrayAPIList(temp)) so the overridden append is not invoked
for each item; leave the return logic (return converted if changed else value)
unchanged.

---

ℹ️ Review info

⚙️ Run configuration

Configuration used: Repository UI

Review profile: CHILL

Plan: Pro

Run ID: c3e8b7bc-bafd-4e48-8dee-69eb6913cd3f

📥 Commits

Reviewing files that changed from the base of the PR and between 5d94bd6 and c558c14.

📒 Files selected for processing (18)

• source/tests/array_api_strict/common.py
• source/tests/array_api_strict/descriptor/__init__.py
• source/tests/array_api_strict/descriptor/dpa1.py
• source/tests/array_api_strict/descriptor/dpa2.py
• source/tests/array_api_strict/descriptor/dpa3.py
• source/tests/array_api_strict/descriptor/hybrid.py
• source/tests/array_api_strict/descriptor/repflows.py
• source/tests/array_api_strict/descriptor/repformers.py
• source/tests/array_api_strict/descriptor/se_atten_v2.py
• source/tests/array_api_strict/descriptor/se_e2_a.py
• source/tests/array_api_strict/descriptor/se_e2_r.py
• source/tests/array_api_strict/descriptor/se_t.py
• source/tests/array_api_strict/descriptor/se_t_tebd.py
• source/tests/array_api_strict/fitting/__init__.py
• source/tests/array_api_strict/fitting/fitting.py
• source/tests/array_api_strict/utils/exclude_mask.py
• source/tests/array_api_strict/utils/network.py
• source/tests/array_api_strict/utils/type_embed.py

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 82.19%. Comparing base (5d94bd6) to head (c558c14).

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5528      +/-   ##
==========================================
- Coverage   82.19%   82.19%   -0.01%     
==========================================
  Files         891      891              
  Lines      101599   101599              
  Branches     4242     4242              
==========================================
- Hits        83507    83506       -1     
  Misses      16789    16789              
- Partials     1303     1304       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.