working on removing dependencies clean_dependencies

CHANGELOG.md

@@ -4,6 +4,68 @@ All notable changes to Celldega are documented here. This project follows
 [Keep a Changelog](https://keepachangelog.com/) conventions and
 [semantic versioning](https://semver.org/).
 
+## [0.24.1] - 2026-08-06
+
+Patch release fixing a NumPy binary-incompatibility crash on import in
+environments that already had NumPy 1.x installed.
+
+### Fixed
+
+- **`import celldega` failed with `ValueError: numpy.dtype size changed`** in
+  environments with a preinstalled NumPy 1.x (e.g. a conda base). The `numpy>=1.23`
+  floor introduced in 0.24.0 was satisfied by the existing 1.x, so it was left in
+  place while NumPy-2-built wheels (`h5py`, `anndata`, ...) were installed,
+  producing a dtype-ABI mismatch. The floor is now a hard `numpy>=2`, forcing the
+  runtime NumPy to match the wheels.
+
+## [0.24.0] - 2026-08-06
+
+Removes Scanpy from the default install path and modernizes the spatial-data
+stack. Cluster/gene coloring is now generated without Scanpy, several viral
+(GPL) and unused dependencies are dropped, and a fresh-environment install no
+longer fails on `pkg_resources`.
+
+### Added
+
+- **`scanpy` optional extra** — Scanpy is now installed only via
+  `pip install celldega[scanpy]`, needed solely for
+  `celldega.clust.Matrix.downsample_to()` (already behind a lazy import with a
+  clear error message).
+
+### Changed
+
+- **Cluster and gene coloring no longer depend on Scanpy.** The color palettes
+  in `Landscape`/`Clustergram` widgets and in `pre` (`_create_cluster_colors`,
+  `make_meta_gene`) and neighborhood gradients (`_ring_colors`) are now built
+  from a deterministic `colorsys` HSV generator instead of triggering a
+  `sc.pl.umap` call to populate `adata.uns[...]_colors`. Widgets no longer mutate
+  the caller's `AnnData`.
+- **`qc.orthogonal_expression_calc` returns data instead of plotting.** It now
+  returns `(results, orthogonal_summary)` DataFrames rather than rendering
+  Seaborn/Matplotlib figures, so callers can plot with any library.
+- **Modernized the spatial-data stack** — `spatialdata>=0.7.2,<0.8`,
+  `spatialdata-io>=0.7.1`, `ome-zarr>=0.12.2`, and `zarr>=3` (with `open_zarr`
+  migrated to the Zarr v3 storage API). `numpy>=2` is now required by the updated
+  stack.
+
+### Removed
+
+- **Dropped viral/copyleft dependencies** `igraph` (GPL-2.0-or-later) and
+  `leidenalg` (GPL-3.0-or-later), plus the `pytest-html` dev dependency
+  (MPL-2.0) — none were imported by celldega.
+- **Dropped unused runtime dependencies** `squidpy`, `dask`, and `datashader`
+  (not imported by celldega; `dask`/`datashader` remain available transitively
+  through `spatialdata`). Scanpy moved to an optional extra (see Added).
+
+### Fixed
+
+- **Fresh-environment install failed with `ModuleNotFoundError: pkg_resources`**
+  ([#292]). The old `spatialdata` pin held back `xarray_schema`, which imports
+  `pkg_resources` (removed in setuptools 82+). Bumping to `spatialdata>=0.7.2`
+  drops `xarray_schema` entirely, fixing the import at its root.
+
+[#292]: https://github.com/broadinstitute/celldega/issues/292
+
 ## [0.23.1] - 2026-08-05
 
 Fixes a batch of `Yearbook` selection, coloring, and layout bugs so that

package.json

@@ -1,6 +1,6 @@
 {
   "name": "celldega",
-  "version": "0.23.1",
+  "version": "0.24.1",
   "type": "module",
   "engines": {
     "node": ">=16",

pyproject.toml

@@ -7,37 +7,32 @@ build-backend = "hatchling.build"
 
 [project]
 name = "celldega"
-version = "0.23.1"
+version = "0.24.1"
 readme = "README.md"
 requires-python = ">=3.11,<3.13"
 dependencies = [
     "anndata>=0.12,<0.13",
-    "scanpy[leiden]>=1.11,<1.12",
-    "squidpy>=1.6,<1.7",
-    "spatialdata>=0.2,<0.3",
-    "spatialdata-io>=0.1,<0.2",
-    "ome-zarr<0.10",
+    # spatialdata>=0.7.2 dropped the xarray_schema dep whose pkg_resources import
+    # broke installs under setuptools>=82 (github.com/broadinstitute/celldega/issues/292).
+    # Capped <0.8 because 0.8 requires Python>=3.12 and we still support 3.11.
+    "spatialdata>=0.7.2,<0.8",
+    "spatialdata-io>=0.7.1,<0.8",
+    "ome-zarr>=0.12.2",
     "mudata",
 
-    # xarray_schema still imports pkg_resources
-    "setuptools<81",
+    "zarr>=3",
 
-    # Keep Zarr 2 because ome-zarr/spatialdata stack still expects FSStore
-    "zarr>=2.18,<3",
-
-    # Pin <2: avoids numpy2/numpy1-ABI binary mismatches (e.g. h5py) in envs with preinstalled numpy1 wheels
-    "numpy>=1.23,<2",
-    "pandas>=2.2,<2.4",
+    # numpy>=2 is required by the modern spatialdata stack (multiscale-spatial-image),
+    # and must be a hard floor (not just >=1.23): in envs with a preinstalled numpy 1.x
+    # (e.g. conda base), a looser floor leaves numpy 1.x in place while numpy-2-built
+    # wheels (h5py, anndata, ...) get installed, causing a dtype-ABI mismatch on import.
+    "numpy>=2",
+    "pandas>=2.2",
     "scipy",
-    "dask",
-    "datashader",
     "polars>=1.10,<2",
     "pyarrow",
     "statsmodels>=0.14,<0.15",
 
-    "igraph",
-    "leidenalg",
-
     "geopandas",
     "libpysal",
     "shapely>=2.0,<2.2",
@@ -67,7 +62,6 @@ dev = [
     # Unit testing
     "pytest",
     "pytest-cov",
-    "pytest-html",
     "pytest-mock",
 
     # Git hooks automation
@@ -89,6 +83,10 @@ dev = [
     "nbconvert",
     "nbformat",
 ]
+# optional dependency for celldega.clust.Matrix.downsample_to()
+scanpy = [
+    "scanpy>=1.11,<1.12"
+]
 
 # Optional dependencies for pre-processing
 pre = [
@@ -195,18 +193,15 @@ dummy-variable-rgx = "^(_+|(_+[a-zA-Z0-9_]*[a-zA-Z0-9]+?))$"
 known-first-party = ["celldega"]
 known-third-party = [
     "anndata",
-    "dask",
     "geopandas",
     "matplotlib",
     "numpy",
     "pandas",
     "polars",
     "scanpy",
     "scipy",
-    "seaborn",
     "shapely",
     "spatialdata",
-    "squidpy",
     "zarr",
 ]
 lines-after-imports = 2

src/celldega/nbhd/gradient.py

@@ -11,6 +11,7 @@
 
 from __future__ import annotations
 
+import colorsys
 from typing import TYPE_CHECKING, Any
 
 import geopandas as gpd
@@ -165,15 +166,15 @@ def _resolve_clip_boundary(
     return None
 
 
-def _ring_colors(cmap_name: str, n: int) -> list[str]:
-    """Sample ``n`` hex colors from a matplotlib colormap (dark -> light)."""
-    import matplotlib.colors as mcolors
-    import matplotlib.pyplot as plt
-
+def _ring_colors(hue: float, n: int) -> list[str]:
+    """Generate ``n`` hex colors along a fixed hue (dark -> light)."""
     if n == 0:
         return []
-    cmap = plt.get_cmap(cmap_name)
-    return [mcolors.to_hex(cmap(v)) for v in np.linspace(0.8, 0.3, n)]
+    colors = []
+    for value in np.linspace(0.4, 0.95, n):
+        r, g, b = colorsys.hsv_to_rgb(hue, 0.75, value)
+        colors.append(f"#{int(r * 255):02x}{int(g * 255):02x}{int(b * 255):02x}")
+    return colors
 
 
 def _calc_gradient(
@@ -351,8 +352,8 @@ def inward_label(i: int) -> str:
         out_labels = gdf_rings.loc[gdf_rings["direction"] == "outward", "ring_range_um"].tolist()
         in_labels = gdf_rings.loc[gdf_rings["direction"] == "inward", "ring_range_um"].tolist()
         color_map = {
-            **dict(zip(out_labels, _ring_colors("Blues", len(out_labels)), strict=True)),
-            **dict(zip(in_labels, _ring_colors("Reds", len(in_labels)), strict=True)),
+            **dict(zip(out_labels, _ring_colors(0.6, len(out_labels)), strict=True)),  # blue
+            **dict(zip(in_labels, _ring_colors(0.0, len(in_labels)), strict=True)),  # red
         }
         gdf_rings["color"] = gdf_rings["ring_range_um"].map(color_map).fillna("#cccccc")
 

src/celldega/pre/__init__.py

@@ -8,15 +8,14 @@
     pyvips = None
 
 import base64
+import colorsys
 import hashlib
 import json
 from pathlib import Path
 import subprocess
 import warnings
 import xml.etree.ElementTree as ET
 
-from matplotlib.colors import to_hex
-import matplotlib.pyplot as plt
 import numpy as np
 import pandas as pd
 from scipy.sparse import csr_matrix, issparse
@@ -93,6 +92,12 @@ def _load_xenium_cluster_data(data_dir, meta_cell):
     return default_clustering, clusters, ser_counts
 
 
+def _hsv_to_hex(h: float) -> str:
+    """Convert HSV color to hex string."""
+    r, g, b = colorsys.hsv_to_rgb(h, 0.65, 0.9)
+    return f"#{int(r * 255):02x}{int(g * 255):02x}{int(b * 255):02x}"
+
+
 def _create_cluster_colors(clusters):
     """
     Create color mapping for clusters.
@@ -103,13 +108,11 @@ def _create_cluster_colors(clusters):
     Returns:
     - List of colors for clusters
     """
-    palettes = [plt.get_cmap(name).colors for name in plt.colormaps() if "tab" in name]
-    flat_colors = [color for palette in palettes for color in palette]
-    flat_colors_hex = [to_hex(color) for color in flat_colors]
+    n = len(clusters)
+    palette = [_hsv_to_hex(i / n) for i in range(n)]
 
     return [
-        (flat_colors_hex[i % len(flat_colors_hex)] if "Blank" not in cluster else "#FFFFFF")
-        for i, cluster in enumerate(clusters)
+        (palette[i] if "Blank" not in cluster else "#FFFFFF") for i, cluster in enumerate(clusters)
     ]
 
 
@@ -1032,14 +1035,7 @@ def make_meta_gene(cbg, path_output):
     print("\n========Write meta gene files========")
     genes = cbg.columns.tolist()
 
-    palettes = [plt.get_cmap(name).colors for name in plt.colormaps() if "tab" in name]
-    flat_colors = [color for palette in palettes for color in palette]
-    flat_colors_hex = [to_hex(color) for color in flat_colors]
-
-    colors = [
-        flat_colors_hex[i % len(flat_colors_hex)] if "Blank" not in gene else "#FFFFFF"
-        for i, gene in enumerate(genes)
-    ]
+    colors = _create_cluster_colors(genes)
 
     ser_color = pd.Series(colors, index=genes)
     meta_gene = calc_meta_gene_data(cbg)
@@ -1546,9 +1542,11 @@ def write_xenium_transform(
     # Function to open a Zarr file
     def open_zarr(path: str) -> zarr.Group:
         store = (
-            zarr.ZipStore(path, mode="r") if path.endswith(".zip") else zarr.DirectoryStore(path)
+            zarr.storage.ZipStore(path, mode="r")
+            if path.endswith(".zip")
+            else zarr.storage.LocalStore(path, read_only=True)
         )
-        return zarr.group(store=store)
+        return zarr.open_group(store=store, mode="r")
 
     try:
         # Open the cells Zarr file
@@ -1761,10 +1759,6 @@ def add_clustering_from_adata(
     The Landscape widget can use the custom clustering by setting the
     `segmentation` parameter to match the `segmentation_name`.
     """
-    from contextlib import suppress
-
-    import scanpy as sc
-
     path_lf = Path(path_dega_files)
 
     # Determine output directory
@@ -1788,15 +1782,7 @@ def add_clustering_from_adata(
     clusters = cluster_counts.index.tolist()
 
     color_key = f"{cluster_key}_colors"
-    colors = None
-    if color_key in adata.uns:
-        colors = adata.uns[color_key]
-    else:
-        # Try to generate colors using scanpy
-        with suppress(Exception):
-            sc.pl.umap(adata, color=cluster_key, show=False)
-            plt.close()
-            colors = adata.uns.get(color_key)
+    colors = adata.uns.get(color_key)
 
     # Fallback to generated colors
     if colors is None: