plot: fail fast with guidance when interactive backend fails to load - #449
Open
yarikoptic-gitmate wants to merge 8 commits into
Open
yarikoptic-gitmate wants to merge 8 commits into
yarikoptic-gitmate wants to merge 8 commits into
Conversation
Previously, con-duct only checked that MPLBACKEND named a known interactive backend, not that it actually loaded. A backend that matched by name but failed to import (e.g. TkAgg with a broken/missing tkinter) only surfaced as a raw traceback deep inside plt.subplots(), with no guidance (#447). Move the interactivity check earlier, before any data loading/plot building, and actually try to load the backend module there so a failure is caught with an actionable message pointing at --output or a different MPLBACKEND value, instead of an unguided traceback. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
When the current backend isn't recognized as interactive, name the built-in interactive backends matplotlib knows about (one of which needs a GUI toolkit installed) instead of just pointing at MPLBACKEND in the abstract. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
If MPLBACKEND isn't set and the resolved matplotlib backend can't display interactively, try loading each builtin interactive backend ourselves (native GUI toolkits first, then browser/cairo variants matplotlib's own auto-fallback doesn't consider, e.g. webagg) and use the first one that actually works, instead of just reporting the non-interactive backend matplotlib gave up on. When MPLBACKEND is set explicitly, that choice is respected as before: report the failure rather than silently overriding it. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
backend_registry.load_backend_module() doesn't consistently raise ImportError for a missing dependency -- e.g. webagg raises RuntimeError when tornado is missing. The pinned-backend branch only caught ImportError, so a pinned backend failing this way (e.g. MPLBACKEND=webagg without tornado installed) leaked a raw traceback instead of the intended friendly message. Widen it to Exception, same as the auto-probe helper already does for the same reason. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #449 +/- ##
=======================================
Coverage 92.02% 92.02%
=======================================
Files 15 15
Lines 1316 1342 +26
Branches 179 182 +3
=======================================
+ Hits 1211 1235 +24
- Misses 70 71 +1
- Partials 35 36 +1 ☔ View full report in Codecov by Harness. 🚀 New features to boost your workflow:
|
The oldestdeps CI job (Python 3.10 + matplotlib 3.5, our documented floor per PR #322/issue #320 for Debian Bookworm compat) was failing: tests that decorate a test function with @patch("matplotlib.backends.backend_registry....") have that patch target resolved before the test body runs, so an inline `if not hasattr(...): pytest.skip(...)` inside the function never gets a chance to run on matplotlib < 3.9, which has no backend_registry module at all -- the decorator itself raises first. Add a shared `requires_backend_registry` skipif marker, evaluated at collection time before pytest calls the (patch-wrapped) test function at all, and apply it to every test that patches backend_registry directly. Verified against a real matplotlib 3.5.3 install: previously 5 errors, now those tests skip cleanly and the rest of the suite passes. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
for more information, see https://pre-commit.ci
Collaborator
Author
|
The Generated by Claude Code |
_INTERACTIVE_BACKEND_PROBE_ORDER's own entries are always builtin interactive backend names, so "candidate not in interactive_backends" can't fire for them in practice, and the loop's broad except Exception already handles a candidate that fails to resolve to a real backend module (load_backend_module on a bad name just raises ModuleNotFoundError, which that clause catches). Dead code identified by codecov's patch coverage flagging it as the one uncovered line/branch in the PR. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf
Collaborator
Author
|
Generated by Claude Code |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Previously, con-duct only checked that MPLBACKEND named a known
interactive backend, not that it actually loaded. A backend that
matched by name but failed to import (e.g. TkAgg with a broken/missing
tkinter) only surfaced as a raw traceback deep inside plt.subplots(),
with no guidance (#447).
Move the interactivity check earlier, before any data loading/plot
building, and actually try to load the backend module there so a
failure is caught with an actionable message pointing at --output or
a different MPLBACKEND value, instead of an unguided traceback.
Co-Authored-By: Claude Sonnet 5 noreply@anthropic.com
Claude-Session: https://claude.ai/code/session_01Hv454Dwv5QSLfUArTtdYyf