Solver.poisson support pre-integrated vfuncs

[magnesium2400]

• Solver.poisson always multiplies the input vfunc h by self.mass to generate an integrated map
• This creates difficulties when solving for different maps, where some are integrated, and some are not: a new Solver object has to be created and its mass set to the identity
• It is easier to create a flag where the user can specify if the map is already integrated or not. This is what I have done here.
• This also simplifies the code in diffgeo.compute_geodesic_f etc.
• The default behaviour is preserved
• The figures in Test_TriaMesh_Geodesics.ipynb still look good when vf = fem.poisson(Bi * divx) is replaced by vf = fem.poisson(divx, integrate=False)

[codecov]

Codecov Report

❌ Patch coverage is 88.88889% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 50.01%. Comparing base (dcacc4a) to head (0cd9031).
⚠️ Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
lapy/diffgeo.py	83.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #115      +/-   ##
==========================================
- Coverage   50.04%   50.01%   -0.04%     
==========================================
  Files          15       15              
  Lines        3153     3151       -2     
  Branches      414      415       +1     
==========================================
- Hits         1578     1576       -2     
  Misses       1426     1426              
  Partials      149      149              

Flag	Coverage Δ
unittests	50.01% <88.88%> (-0.04%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 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.

[Copilot]

Pull request overview

This PR adds an option to Solver.poisson to skip mass-matrix integration when the RHS is already pre-integrated, and updates differential-geometry helpers to use that option instead of mutating Solver.mass.

Changes:

• Add integrate: bool = True to Solver.poisson to optionally bypass B*h integration.
• Update diffgeo.compute_geodesic_f, tria_compute_geodesic_f, and tria_compute_rotated_f to call poisson(..., integrate=False) instead of replacing fem.mass with the identity.
• Document the new flag in the poisson docstring.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 5 comments.

File	Description
lapy/solver.py	Introduces integrate flag and conditional mass-matrix application when building the RHS.
lapy/diffgeo.py	Switches geodesic/rotated-function routines to use integrate=False rather than altering solver state.

---

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

[Copilot]

The new parameter doc header uses integrate: bool but the rest of the docstring follows NumPy style (name : type). Also, integrate=False changes the equation being solved and skips mass-matrix scaling for the Neumann term as well (since b = h - nvec is no longer multiplied by B). Please adjust the docstring (including the earlier text that implies A x = B h) and clarify the expected ntup convention when integrate=False.

[Copilot]

The new integrate flag introduces a second execution path for RHS construction. Please add unit tests covering integrate=False (including at least one case with Dirichlet/Neumann terms if supported) and asserting it matches the previous workaround (e.g., setting mass to identity or pre-multiplying h appropriately), so behavior stays stable across refactors.

[Copilot]

There’s now commented-out dead code (# fem.mass = sparse.eye(...)). Since integrate=False is the intended mechanism, consider removing the commented assignment and leaving a short explanatory comment (e.g., that compute_divergence already returns an integrated divergence so Poisson should not re-apply B).

Suggested change

	# fem.mass = sparse.eye(fem.stiffness.shape[0], dtype=fem.stiffness.dtype)
	# compute_divergence already returns an integrated divergence, so Poisson
	# should solve with integrate=False and must not re-apply the mass matrix.

[Copilot]

The comment here still describes the old workaround (“pass identity instead of B here”) and keeps the identity assignment commented out. Please update the comment to reflect the new approach (integrate=False) and remove the commented-out fem.mass = sparse.eye(...) line to avoid confusing future readers.

[Copilot]

There’s leftover commented-out code (# fem.mass = sparse.eye(...)) from the pre-integrate workaround. Since this function now calls poisson(..., integrate=False), please remove the commented assignment and (if needed) replace it with a brief note explaining why integration is disabled here.

Suggested change

	# fem.mass = sparse.eye(fem.stiffness.shape[0], dtype=vfunc.dtype)
	# The divergence has already been assembled, so disable integration in the
	# Poisson solve instead of using the old mass-matrix workaround.