L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Productivité · SKILL.md

legacy-factories

legacy-factories

Routes legacy scikit-plot factory-injection and deprecated plotters-module requests to safe migration and compatibility guidance.

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: legacy-factories description: "Routes legacy scikit-plot factory-injection and deprecated plotters-module requests to safe migration and compatibility guidance." disable-model-invocation: true metadata: disco-role: operating license: MIT

Legacy factories

Use this sub-skill when the user is maintaining older scikit-plot code that calls classifier_factory, clustering_factory, or imports plotting helpers from the deprecated scikitplot.plotters module.

Route here

  • Inject plotting methods into a classifier instance with classifier_factory.
  • Inject plotting methods into a clusterer instance with clustering_factory.
  • Diagnose warnings about missing methods, method-name collisions, or deprecation.
  • Migrate old scikitplot.plotters imports to current modules.
  • Validate compatibility quickly with scripts/legacy_factories_smoke.py.

Prefer modern routes for new work

Use direct module functions for new code:

  • Metric curves and silhouette: ../metrics/SKILL.md.
  • Feature importance and learning curves: ../estimators/SKILL.md.
  • Elbow curves: ../clustering/SKILL.md.
  • PCA plots: ../decomposition/SKILL.md.

The factory layer is useful when the user must preserve old bound-method code, but it is deprecated in this snapshot.

Inputs to check first

  1. A classifier passed to classifier_factory must expose fit, score, and predict.
  2. Probability-based injected methods also need predict_proba.
  3. A clusterer passed to clustering_factory must expose fit and fit_predict.
  4. Existing methods with the injected names may be overwritten with warnings.
  5. The same SciPy/Matplotlib compatibility window from the root skill still applies.

Reroute

  • If the user wants exact behavior of the underlying plot, route to the owning sub-skill after identifying the injected method.
  • If the user only needs an import migration from scikitplot.plotters, read references/workflows.md.
  • If a factory warning is the main issue, read references/troubleshooting.md before changing model code.

Read next

  • references/api-reference.md for injected method names and verified contracts.
  • references/workflows.md for classifier, clusterer, and deprecated-module migration patterns.
  • references/troubleshooting.md for missing-method, override, and deprecation failures.
  • scripts/legacy_factories_smoke.py for a tiny Agg-backed compatibility smoke run.
Skills enregistrées