[BradyPlanden]

Description

Stacked on #5782, whose Memory tests job failed test_cycle_count_memory_scaling with a 20-to-5 cycle memory ratio of 1.403 against a 1.4 gate.

Profiling with tracemalloc on Python 3.13.15 Linux (x86_64 and aarch64, with CI's pybammsolvers wheels) showed that the solver statistics #5782 adds cost 72 B per step, one SolverStatistics. That is within run-to-run noise: main and #5782 both measured 1.39–1.40, so the gate was already at the edge on main. Most of each step's memory was Solution overhead. An experiment keeps four solutions per step (the step, its cycle, and the cycle's cached first and last states), and each carried a full-size __dict__ (1,584 B on 3.13). Solution set more instance attributes than CPython's 30-key shared layout holds, and Python 3.14 pre-fills those keys from __static_attributes__. Whether a solution's attributes fit therefore also depended on which solutions the process had made before, so the test's ratio moved with test order.

• Solution and SolutionBase declare __slots__, so every solution gets the same compact layout whatever the Python version or process history. __dict__ stays, so callers can still attach their own attributes.
• first_state and last_state cache in slots instead of through functools.cached_property, which needs __dict__.
• Solution.__setstate__ loads solutions pickled before this change: a plain dict state, without solver_statistics, with the cached states under their property names.
• test_cycle_count_memory_scaling's gate drops from 1.4 to 1.3, so a regression to main's level fails it.
• An editable pybammsolvers install, which uv sync creates, can be imported on its own. idaklu's only CasADi rpath was ../casadi, which holds for a wheel but not for the build tree an editable install loads from, so import pybammsolvers failed to load libcasadi unless casadi had been imported first. Editable builds also search the environment's CasADi directory; wheels are unchanged.

Run after the rest of the memory suite, as CI runs it:

Platform	Ratio	Growth per 4-step cycle
Linux x86_64, Python 3.13 (CI)	1.242–1.245	32.9 KB, from 55.9 KB on #5782
Linux aarch64, Python 3.13	1.240–1.253	33.1 KB
macOS arm64, Python 3.14	1.265–1.279	37.4 KB

Type of change

Add an entry under # [Unreleased] in CHANGELOG.md, in one of ## Breaking changes, ## Deprecated, ## Features, or ## Bug fixes (include PR number; breaking and deprecation entries need a one-line migration note). Internal-only PRs — refactor, docs, CI, tests — can skip this. See RELEASE.md for the full policy.

Important checks:

Please confirm the following before marking the PR as ready for review:

• No style issues: nox -s pre-commit
• All tests pass: nox -s tests
• The documentation builds: nox -s doctests
• Code is commented for hard-to-understand areas
• Tests added that prove fix is effective or that feature works

[github-actions]

Bencher Report

Project	PyBaMM
Branch	a2/solution-compact-attributes
Testbed	bare-metal

⚠️ WARNING: Truncated view!

The full continuous benchmarking report exceeds the maximum length allowed on this platform.

🐰 View full continuous benchmarking report in Bencher

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 98.09%. Comparing base (b3c168f) to head (423ed38).

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #5827   +/-   ##
=======================================
  Coverage   98.09%   98.09%           
=======================================
  Files         346      346           
  Lines       35323    35354   +31     
=======================================
+ Hits        34649    34680   +31     
  Misses        674      674           

☔ 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.

[martinjrobins]

is it possible to remove the dict if we are using slots? I don't think anyone else should be adding attributes to this class

[martinjrobins]

thanks @BradyPlanden !