SemiLagrangian: remove the unused swarm_degree / swarm_continuous and _workVar (#704)

[lmoresi]

Fixes #704.

SemiLagrangian accepted swarm_degree / swarm_continuous, stored them, and used them only to size _workVar, which nothing read: the trace-back samples at psi_star's own nodes and every projection overwrote the work variable's symbol before solving. The issue's sweep shows the answer bit-identical across the parameters while the variable spanned 98 to 972 nodes.

This removes the pair, the allocation, its remesh registration, and the docstring and comment that claimed they set the sample points. The projection solver's placeholder source becomes psi_fn, which every use already installs before solving. No caller in the tree passed the parameters (one commented-out line in the SLCN solver), no test or doc names them, so they are dropped rather than deprecated. The docstring now says the departure points are the field's own nodes and points to the integration-point history (PR #703) for denser sampling.

Semi-Lagrangian test files (snapshot, solver smoke, Navier-Stokes SLCN, recursion, old-frame, units, transport plugin): 59 passed.

Review

• Removal only; no numerical path changes. The one behavioural difference is one fewer MeshVariable per SemiLagrangian (and per AdvDiffusionSLCN, which creates two), so the mesh DM has fewer fields; nothing indexes them by position.
• Breaking for a script that passed the keywords: it now raises TypeError. We chose that over a silent deprecated no-op because a no-op is exactly what the report calls misleading.
• The _psi_star_use_multicomponent branch keeps its placeholder row built from psi_fn with the same independent-index selection.

Underworld development team with AI support from Claude Code

🤖 Generated with Claude Code

https://claude.ai/code/session_01MSGAFeA7qYXgkuw9ud8F2G

[Copilot]

🟡 Changes recommended

The updated docstring note should explicitly state that no nodal swarm cache is used (sampling is via uw.function.global_evaluate) to fully eliminate remaining user-facing ambiguity.

Once you've addressed the issues Copilot identified, you can request another Copilot review.

Pull request overview

This PR simplifies the SemiLagrangian history manager by removing the unused swarm_degree / swarm_continuous parameters and the associated unused _workVar allocation, aligning the public API and documentation with the implementation (departure-point sampling at the history variable’s own nodes).

Changes:

• Removed swarm_degree / swarm_continuous from SemiLagrangian.__init__ and deleted the unused _workVar MeshVariable and its remesh registration.
• Updated the initial projection-solver placeholder source to use psi_fn directly (including the multicomponent/tensor branch).
• Adjusted docstrings/comments to stop claiming a separate sampling discretisation is configured via the removed parameters.

File summaries

File	Description
src/underworld3/systems/ddt.py	Drops unused SemiLagrangian sampling kwargs and _workVar, and updates projection placeholder + docs to match the nodal sampling implementation.

Review details

• Files reviewed: 1/1 changed files
• Comments generated: 1
• Review effort level: Lite

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[lmoresi]

Fixed: the class description now says the characteristics are traced from the history variable's own nodes and sampled with uw.function.global_evaluate, and that no swarm is allocated or used.