Fix ZeroDivisionError when an optimization batch extracts nothing - #846
Open
GeorgWa wants to merge 3 commits into
Open
Fix ZeroDivisionError when an optimization batch extracts nothing#846GeorgWa wants to merge 3 commits into
GeorgWa wants to merge 3 commits into
Conversation
An optimization batch that extracts nothing sets total_elution_groups to 0, and the three optimizers computing precursor_proportion_detected each divided by it unguarded. The crash landed inside proceed_with_insufficient_precursors, the branch that exists to absorb exactly this case. Move the feature into workflow/optimizers/features.py, where a class owns its history column name, its computation, and its definition of undefined. The guard lives inside the feature, so it cannot be forgotten by a fifth optimizer, and name and computation can no longer drift apart. A round without a measurement is skipped rather than recorded as 0.0: a sentinel would make the relative comparisons in _just_converged and the axis limits in plot() non-finite, trading a loud failure for a silent one. _update_workflow keeps the current parameter when no round produced a measurement, since _find_index_of_optimum raises on an empty history. Behaviour is unchanged for any batch with total_elution_groups > 0. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Skipping the history row was not enough: step() carried on to propose a new parameter from the same empty frame. An unfitted estimator's ci() returns 0, so the parameter was silently set to 0.0 and every later search extracted nothing — the loud failure this branch removed, traded for a silent one. _update_history now reports whether it recorded a measurement, and step() leaves the parameter alone when it did not. Also: - guard plot() against the empty history that _update_workflow already handles, so re-enabling its commented-out call site cannot reintroduce the crash - raise at construction when a subclass declares no _feature, restoring the fail-fast that the abstract _get_feature_value used to provide - treat an all-NaN isotope_intensity_correlation column as unmeasurable, the NaN hazard MeanIsotopeIntensityCorrelation's own docstring describes - word the skip warning for both features, not just the zero-elution-group case Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Remove the hasattr check in AutomaticOptimizer.__init__. A subclass that does not declare _feature now fails in the same way as one that does not declare parameter_name or _estimator_name. Write the new docstrings and comments in simplified technical English (ASD-STE100): short sentences, active voice, and simple words. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
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
1 participant
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.
A search stops with a
ZeroDivisionErrorwhen an optimization batch extracts nothing. The failure occurs in the code that must recover from a batch with too few precursors.Three optimizers measure the same quantity. Each one calculated it separately, and none of them handled a batch with no data. This change gives the measurement its own class. The class holds the name, the calculation, and the rule for a batch with no data. All optimizers now use one rule, and a new optimizer gets the rule with it.
A round with no data is left out of the optimization history. The alternative is to record a zero. But zero is also a valid measurement, and it would corrupt the convergence test without an error message.
Optimization keeps the current search parameter when a round gives no data. It also completes normally when no round gives data at all.
A batch with data behaves as before.
🤖 Generated with Claude Code