Feature/var detected by cat upset by leventetn · Pull Request #40 · UKHD-NP/proteopy

leventetn · 2026-09-28T23:30:18Z

upset_test_spec.md
upset_function_spec.md

Tests: add spec-guided TDD support conftest Add tests/conftest.py from the spec-guided-tdd skill, formatted to ProteoPy's black/flake8 standards. It only affects test modules marked `pytestmark = pytest.mark.spec_guided` (e.g. tests/pl/test_upset.py); all other tests run unchanged. - `pytest` (default, CI) runs the holdout and randomized tests and deselects the `_IMP` mirrors - `pytest --implementer` runs the `_IMP` and randomized tests with obscured failure output, for implementation sessions - `--seed N` reproduces randomized inputs (seed printed in the header) - provides the `rng` and `report_input` fixtures - checks test naming and holdout/_IMP pairing

Tests: pl.var_detected_by_cat_upset() Add a spec-guided test suite for var_detected_by_cat_upset() (module marked `spec_guided`, run by tests/conftest.py): - TestVarDetectedByCatUpset: 75 holdout tests plus 9 randomized property and metamorphic tests - TestVarDetectedByCatUpsetIMP: 75 _IMP mirrors on different inputs, run only with `pytest --implementer` Coverage: intersection counts and thresholds (min_count / min_fraction, inclusivity, per-category denominators, empty categories, float edge cases), zero/NaN detection, category order and str coercion, sparse input, non-mutation (incl. views), UpSet construction and rendered axes, print_stats tables (incl. column-name collisions), verbose output, save/show, and argument validation. UpSet calls are observed through spies on the upsetplot class; randomized tests use the seeded rng fixture (reproduce with `--seed N`).

Feature: pl.var_detected_by_cat_upset() Add an UpSet plot of which features (.var) are detected in which categories of an .obs column, e.g. which proteins are found in which tissues. - A feature is a member of a category when it is detected (non-NaN in .X; with zero_to_na=True also non-zero) in enough of that category's observations: min_count (default 1) or min_fraction, inclusive, per category. Setting both raises ValueError. - Features that are members of no category form a "No category" set, always included, also with a count of 0. - Categories follow the default pl order: category order for a Categorical column (unused categories kept), else lexicographic order of the str-coerced values. Values colliding after str coercion raise ValueError. - print_stats prints global, per-intersection and per-category tables; verbose reports the input matrix, threshold and sizes. - Sparse .X is densified with a UserWarning; adata is never modified. - Returns the axes dict of upsetplot's UpSet.plot(). Add upsetplot>=0.9.0,<0.10 as a core dependency. upsetplot 0.9.0 is broken on pandas 3 / numpy >= 2.4, so proteopy.pl.upset patches it on import (dot style defaults, single-category aggregation, count-label positions, all-empty totals warning). The upper bound keeps these patches from hitting an untested upsetplot release. Export via pr.pl, list it under Quality Control in the plotting API docs, and add a HISTORY.md entry.

read-the-docs-community · 2026-09-28T23:32:08Z

Documentation build overview

📚 proteopy | 🛠️ Build #35071191 | 📁 Comparing a4164b5 against latest (3661567)

🔍 Preview build

1 file changed

± changelog.html

leventetn · 2026-10-08T14:35:39Z

Implemented sort_by with "degree", "cardinality", "-degree" and "-cardinality". New tests cover all modes, equal counts and degrees, repeatability, bar/dot alignment, and invalid inputs.

idf-io

PR review part 1. "Functionality"

idf-io · 2026-10-17T21:34:16Z

+    verbose: bool = False,
+    show: bool = True,
+    save: str | Path | None = None,
+    *,


Put * after cat_key, only adata and cat_key should be positional arguments.

idf-io · 2026-10-17T21:38:25Z

+        ProteoPy AnnData (peptide- or protein-level). Intensities are
+        read from ``adata.X`` only.


Reword: AnnData in ProteoData format.

idf-io · 2026-10-17T21:40:08Z

+    show : bool
+        Call ``plt.show()`` at the end.
+    save : str | Path | None
+        Path to save the figure to; None skips saving.


"; None skips saving" is redundant.

idf-io · 2026-10-17T21:43:34Z

+        Call ``plt.show()`` at the end.
+    save : str | Path | None
+        Path to save the figure to; None skips saving.
+    sort_by : str


rename to "intersection_order".
order_by is reserved for a plot when you order the samples/vars by a category and color by them. e.g.

idf-io · 2026-10-17T21:44:33Z

+
+    Returns
+    -------
+    dict[str, Axes]


If pyupsetplot coopts the matplotlib figure, return None

idf-io · 2026-10-17T21:53:20Z

+        order after conversion to strings. Use an ordered
+        :class:`pandas.Categorical` to customize the order.
+    min_count : int | None
+        Minimum detections per category, as a nonnegative integer.


Make it a tiny bit clearer by wording like:
Minimum number of non-missing var values (detections) within a category for the membership to be counted.
Make same improvement to min_fraction

idf-io · 2026-10-17T21:56:37Z

+        )
+
+    # -- Plot
+    axes = _plot_upset(counts, sort_by)


Remove axes return. Return None.

idf-io · 2026-10-18T18:50:05Z

+def _print_verbose_report(
+    cat_key: str,
+    threshold_name: str,
+    threshold_value: int | float,
+    n_vars: int,
+    n_categories: int,
+) -> None:
+    """Print a short report about the input of the plot."""
+    print(
+        f"Using the .X matrix.\n"
+        f"Categories from .obs['{cat_key}'].\n"
+        f"Threshold: {threshold_name} = {str(threshold_value)}\n"
+        f"Features: {n_vars}\n"
+        f"Categories: {n_categories}"
+    )


Too obvious verbose statements. They don't bring any new information. i dont think this plot needs any verbose output.
Guideline for plotting functions:

verbose: non-obvious useful printouts e.g. if several heavy steps are involved, announce them. If expensive looping algorithm, print at which loop we are etc.

print_stats: print what we see in the plot so that the user can have tangible numbers to report in a paper. This includes sometimes making stats of what you see: mean of data points...

idf-io · 2026-10-18T18:50:36Z

+    counts = _intersection_counts(membership, names)
+
+    # -- Report
+    if verbose:


Remove verbose parameter, see other comment why,

idf-io · 2026-10-18T18:59:41Z

+    axes = _plot_upset(counts, sort_by)
+
+    if save is not None:
+        axes["matrix"].figure.savefig(save)


Does this actually work? From the code I would think that if there are several axes and you only save one of them, we are not saving the entire upset figure. Please check in practice.

leventetn added 4 commits September 28, 2026 17:35

fixed/removed legend from var_detected_by_cat_upset

4d51cf7

leventetn force-pushed the feature/var_detected_by_cat_upset branch from bee29ba to 2c85530 Compare October 8, 2026 21:39

leventetn added 2 commits October 11, 2026 11:15

fix: docstrings and history

8ef5887

fix: Implemented sort_by for var_detected_by_cat_upset

a4164b5

Implemented sort_by with "degree", "cardinality", "-degree" and "-cardinality". New tests cover all modes, equal counts and degrees, repeatability, bar/dot alignment, and invalid inputs.

leventetn force-pushed the feature/var_detected_by_cat_upset branch from 2c85530 to a4164b5 Compare October 11, 2026 09:32

idf-io reviewed Oct 26, 2026

View reviewed changes

		ProteoPy AnnData (peptide- or protein-level). Intensities are
		read from ``adata.X`` only.

Conversation

leventetn commented Sep 28, 2026

Uh oh!

read-the-docs-community Bot commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Documentation build overview

Uh oh!

leventetn commented Oct 8, 2026

Uh oh!

idf-io left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

read-the-docs-community Bot commented Sep 28, 2026 •

edited

Loading

idf-io left a comment •

edited

Loading