FXC-1617: switch tensorial mode solver to tidy3d-extras #3218

momchil-flex · 2026-01-29T11:48:08Z

The tensorial mode solver in tidy3d-extras has been made more accurate. Notably, for nominally lossless, fully tensorial media, the mode solver should return very close to lossless modes.

To avoid confusion, we are removing the legacy local tensorial mode solver. Local runs are still fully supported (and more accurate!) they just require installing tidy3d-extras.

Note

Medium Risk
Changes affect the mode-solver execution path and error behavior for tensorial/angled/aniso cases, and rely on optional tidy3d-extras availability; diagonal/local and server paths should remain unchanged but need regression coverage.

Overview
Local mode solving now selects the best available solver at runtime: ModeSolver/ModeSimulation.run_local() will use tidy3d-extras’ mode solver when present, otherwise fall back to the base solver.

The legacy in-core fully tensorial eigenmode implementation is removed; attempting a tensorial solve (non-zero ModeSpec.angle_theta or fully anisotropic media) without tidy3d-extras now raises a clear NotImplementedError instructing users to install tidy3d[extras] or run via the server.

Tests and docs are updated accordingly: new coverage asserts the error/fallback behavior and skips tensorial-local tests when extras aren’t available, and plugin tests that directly exercised compute_modes/angled tensorial paths are removed.

^{Written by Cursor Bugbot for commit 3cb8539. This will update automatically on new commits. Configure here.}

Copilot

Pull request overview

This PR transitions the fully tensorial mode solver from local implementation to requiring either tidy3d-extras or server execution. The change aims to consolidate on a more accurate tensorial mode solver implementation available in tidy3d-extras while maintaining backward compatibility for diagonal (non-tensorial) mode solving.

Changes:

Replaces the local tensorial mode solver implementation with a NotImplementedError that provides clear instructions to users
Adds dynamic solver selection via _get_solver_func() that prefers tidy3d-extras when available
Includes comprehensive test coverage for both with/without tidy3d-extras scenarios

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
tidy3d/components/mode/solver.py	Removed ~130 lines of tensorial solver implementation; replaced with error message directing users to tidy3d-extras or server
tidy3d/components/mode/mode_solver.py	Added `_get_solver_func()` to dynamically select between extras and base solver; updated solver calls to use this function
tests/test_components/test_mode.py	Added comprehensive tests for tensorial solver error handling, tidy3d-extras integration, and diagonal solver functionality
CHANGELOG.md	Documented the breaking change with clear user guidance

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

CHANGELOG.md

tests/test_components/test_mode.py

tidy3d/components/mode/mode_solver.py

cursor

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}

tests/test_components/test_mode.py

- Add explanatory comment to except clause in _get_solver_func - Fix import style in test (use 'from' import) - Disable local subpixel in test to prevent tidy3d-extras bypass

github-actions · 2026-02-02T14:02:41Z

Diff Coverage

Diff: origin/develop...HEAD, staged and unstaged changes

tidy3d/components/mode/mode_solver.py (75.0%): Missing lines 126,131-132,134
tidy3d/components/mode/solver.py (100%)

Summary

Total: 18 lines
Missing: 4 lines
Coverage: 77%

tidy3d/components/mode/mode_solver.py

Lines 122-138

  122     ImportError
  123         If the local solver could not be imported (e.g., scipy not installed).
  124     """
  125     if not LOCAL_SOLVER_IMPORTED:
! 126         raise ImportError(IMPORT_ERROR_MSG)
  127 
  128     # Try to get tidy3d-extras solver (handles all cases including tensorial)
  129     try:
  130         _check_tidy3d_extras_available(quiet=True)
! 131         if tidy3d_extras["mod"] is not None:
! 132             from tidy3d_extras.mode import EigSolver as ExtrasEigSolver
  133 
! 134             return ExtrasEigSolver.compute_modes
  135     except (Tidy3dImportError, ImportError, AttributeError):
  136         # tidy3d-extras is optional; if unavailable or incompatible, fall back to base solver.
  137         pass

caseyflex · 2026-02-05T22:48:26Z

tidy3d/components/mode/mode_solver.py


-        if not LOCAL_SOLVER_IMPORTED:
-            raise ImportError(IMPORT_ERROR_MSG)
+        solver_func = _get_solver_func()


Why not use the @supports_local_subpixel decorator and related logic like in _solve_all_freqs?

I think that handling is somewhat more flexible actually, in addition to already being factored out into a subroutine. It lets the user control the behavior: config.use_local_subpixel=None means use extras if available, False means don't use it, True means always use it. I would do the same thing here, and just raise an error if tensorial mode solve is required and tidy3d_extras["use_local_subpixel"] is False.

caseyflex · 2026-02-05T22:57:25Z

tests/test_components/test_mode.py

+
+    # Also disable local subpixel to prevent tidy3d-extras from bypassing our patched code path
+    # when tidy3d-extras is installed and licensed
+    monkeypatch.setitem(tidy3d_extras, "use_local_subpixel", False)


on a related note, not sure about this test logic. Seems we want the error when this setting is None too.

To take advantage of the integration tests, in addition to the mock test for the error message, you could also have a test that checks whether tidy3d_extras is available (I think there are others like that) and has different behavior in the two cases

caseyflex · 2026-02-05T22:57:55Z

tests/test_components/test_mode.py

+    assert result_aniso.modes_raw.n_eff is not None
+
+
+def test_diagonal_mode_solver_still_works():


This test seems redundant

caseyflex · 2026-02-05T23:00:19Z

tests/test_plugins/test_mode_solver.py



-@td.packaging.disable_local_subpixel
-def test_mode_solver_straight_vs_angled():


Interesting, so this test previously had local subpixel disabled, presumably because subpixel would change the results of the test. I guess your new tests cover this use case? We could have one of them in the frontend but only run if tidy3d_extras is present?

momchil-flex force-pushed the momchil/anisotropic_mode branch from d657e01 to a266ac8 Compare January 29, 2026 12:00

momchil-flex marked this pull request as ready for review January 30, 2026 10:59

Copilot AI review requested due to automatic review settings January 30, 2026 10:59

Copilot started reviewing on behalf of momchil-flex January 30, 2026 10:59 View session

Copilot AI reviewed Jan 30, 2026

View reviewed changes

CHANGELOG.md Outdated Show resolved Hide resolved

tests/test_components/test_mode.py Outdated Show resolved Hide resolved

tidy3d/components/mode/mode_solver.py Show resolved Hide resolved

cursor bot reviewed Jan 30, 2026

View reviewed changes

tests/test_components/test_mode.py Show resolved Hide resolved

momchil-flex added 3 commits February 2, 2026 13:50

refactor: switch tensorial mode solver to tidy3d-extras

ea52ef2

fix: address PR review comments for tensorial mode solver

4d8c52c

- Add explanatory comment to except clause in _get_solver_func - Fix import style in test (use 'from' import) - Disable local subpixel in test to prevent tidy3d-extras bypass

Move angled mode solver tests to tidy3d_extras

3cb8539

momchil-flex force-pushed the momchil/anisotropic_mode branch from a266ac8 to 3cb8539 Compare February 2, 2026 13:38

momchil-flex requested a review from caseyflex February 3, 2026 08:59

caseyflex requested changes Feb 5, 2026

View reviewed changes

caseyflex reviewed Feb 5, 2026

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

FXC-1617: switch tensorial mode solver to tidy3d-extras #3218

FXC-1617: switch tensorial mode solver to tidy3d-extras #3218

Uh oh!

momchil-flex commented Jan 29, 2026 •

edited by cursor bot

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

cursor bot left a comment

Uh oh!

Uh oh!

github-actions bot commented Feb 2, 2026

Uh oh!

caseyflex Feb 5, 2026 •

edited

Loading

Uh oh!

caseyflex Feb 5, 2026

Uh oh!

caseyflex Feb 5, 2026

Uh oh!

caseyflex Feb 5, 2026

Uh oh!

caseyflex Feb 5, 2026

Uh oh!

caseyflex Feb 5, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		assert result_aniso.modes_raw.n_eff is not None


		def test_diagonal_mode_solver_still_works():



		@td.packaging.disable_local_subpixel
		def test_mode_solver_straight_vs_angled():

FXC-1617: switch tensorial mode solver to tidy3d-extras #3218

Are you sure you want to change the base?

FXC-1617: switch tensorial mode solver to tidy3d-extras #3218

Uh oh!

Conversation

momchil-flex commented Jan 29, 2026 • edited by cursor bot Loading Uh oh! There was an error while loading. Please reload this page.

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!

cursor bot left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

github-actions bot commented Feb 2, 2026

Diff Coverage

Diff: origin/develop...HEAD, staged and unstaged changes

Summary

tidy3d/components/mode/mode_solver.py

Uh oh!

caseyflex Feb 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

caseyflex Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

caseyflex Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

caseyflex Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

caseyflex Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

caseyflex Feb 5, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

momchil-flex commented Jan 29, 2026 •

edited by cursor bot

Loading

caseyflex Feb 5, 2026 •

edited

Loading