use numpy histogram in scatter_density

CHANGELOG.md

@@ -0,0 +1,9 @@
+# Changelog
+
+All notable changes to this project will be documented in this file.
+
+## [0.1.6] - 2024-xx-xx
+
+### Changed
+- `pm.pl.scatter_density` now uses `np.histogram2d`
+- `datashader` dependency has been removed

pyproject.toml

@@ -24,7 +24,6 @@ dependencies = [
     "matplotlib",
     "readfcs >=1.1.0",
     "flowutils",
-    "datashader",
     "consensusclustering",
     "minisom"
 ]

pytometry/plotting/_scatter_density.py

@@ -1,13 +1,10 @@
 from typing import Literal  # noqa: TYP001
 from typing import List, Optional, Tuple, Union
 
-import datashader as ds
+import matplotlib.colors as mcolors
 import matplotlib.pyplot as plt
 import numpy as np
-import pandas as pd
-import scanpy as sc
 from anndata import AnnData
-from datashader.mpl_ext import dsshow
 from matplotlib.axes import Axes
 from matplotlib.colors import Colormap
 from matplotlib.scale import ScaleBase
@@ -25,6 +22,7 @@ def scatter_density(
     y_lim: Optional[Tuple[float, float]] = None,
     ax: Optional[Axes] = None,
     figsize: Optional[tuple[int, int]] = None,
+    bins: Union[int, tuple[int, int]] = 500,
     cmap: Union[str, List, Colormap] = "jet",
     vmin: Optional[float] = None,
     vmax: Optional[float] = None,
@@ -54,6 +52,8 @@ def scatter_density(
             draw into an existing figure.
         figsize (tuple), optional:
             Figure size (width, height) if ``ax`` not provided. Defaults to (10, 10).
+        bins (int or tuple), optional:
+            Number of bins for the `np.histogram2d` function
         cmap (str or list or :class:`matplotlib.colors.Colormap`), optional:
             For scalar aggregates, a matplotlib colormap name or instance.
             Alternatively, an iterable of colors can be passed and will be converted
@@ -69,41 +69,43 @@ def scatter_density(
     Returns:
         Scatter plot that displays cell density
     """
-    figsize = figsize if figsize is not None else (10, 10)
     ax = plt.subplots(figsize=figsize)[1] if ax is None else ax
-    if x_label is None:
-        x_label = x
-    if y_label is None:
-        y_label = y
-    # Create df from anndata object
-    markers = [x, y]
-    joined = sc.get.obs_df(adata, keys=[*markers], layer=layer)
 
-    # Convert variables to np.array
-    x = np.array(joined[x])
-    y = np.array(joined[y])
+    if isinstance(bins, int):
+        bins = (bins, bins)
 
-    # Plot density with datashader
-    df = pd.DataFrame(dict(x=x, y=y))
-    dsartist = dsshow(
-        df,
-        ds.Point("x", "y"),
-        ds.count(),
-        vmin=vmin,
-        vmax=vmax,
-        norm=None,
-        # aspect="auto",
-        ax=ax,
-        cmap=cmap,
+    hist, xedges, yedges = np.histogram2d(
+        adata.obs_vector(x, layer=layer), adata.obs_vector(y, layer=layer), bins=bins
     )
 
-    plt.colorbar(dsartist)
+    vmin = hist.min() if vmin is None else vmin
+    vmax = hist.max() if vmax is None else vmax
 
-    plt.xlim(x_lim)
-    plt.ylim(y_lim)
-    plt.yscale(x_scale)
-    plt.xscale(y_scale)
-    plt.xlabel(x_label)
-    plt.ylabel(y_label)
+    image = ax.imshow(
+        hist.T,
+        extent=[xedges[0], xedges[-1], yedges[0], yedges[-1]],
+        norm=mcolors.Normalize(vmin=vmin, vmax=vmax),
+        cmap=_get_cmap_white_background(cmap),
+        aspect="auto",
+        origin="lower",
+    )
+    plt.colorbar(image, ax=ax)
+
+    ax.set_xlim(x_lim)
+    ax.set_ylim(y_lim)
+    ax.set_yscale(x_scale)
+    ax.set_xscale(y_scale)
+    ax.set_xlabel(x if x_label is None else x_label)
+    ax.set_ylabel(y if y_label is None else y_label)
 
     plt.show()
+
+
+def _get_cmap_white_background(cmap: Union[str, List, Colormap]) -> Colormap:
+    if isinstance(cmap, str):
+        cmap = plt.cm.get_cmap(cmap)
+
+    colors = cmap(np.arange(cmap.N))
+    colors[0] = np.array([1, 1, 1, 1])
+
+    return mcolors.ListedColormap(colors)