Add pertpy's E-Distance (#454)

* init working version with pytest.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* naive implementation

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* error from merge

* integrate more tests

* save state with temporary scripts. now there is a reference code to improve

* vibe codded version + the version I simplied. it works roughly

* changes

* simplify kernels and create a separate file

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* move to float64 for precision.

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* bootstrapping

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* swich to f32

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* refactor

* push working state

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* add documentation and write a comparison script

* [pre-commit.ci] auto fixes from pre-commit.com hooks

for more information, see https://pre-commit.ci

* speed up kernel

* update

* update to faster implementation

* update kernel and tests

* add docs and fix test with smaller fucntions

* fix shared memory size issues

* avoid register pressure on old hardware

* add skipped test

* update docstring

* update docs

* finish docs

* update the page

---------

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Severin Dicks <37635888+Intron7@users.noreply.github.com>
Co-authored-by: Intron7 <severin.dicks@icloud.com>

Author: 3 people

.gitignore

@@ -42,7 +42,7 @@ __pycache__/
 .vscode/
 .cursor/
 .claude/
-Claude.md
+CLAUDE.md
 
 # tmp_scripts
 tmp_scripts/

docs/_templates/autosummary/class.rst

@@ -13,7 +13,7 @@ Attributes table
 
 .. autosummary::
 {% for item in attributes %}
-    ~{{ fullname }}.{{ item }}
+    ~{{ objname }}.{{ item }}
 {%- endfor %}
 {% endif %}
 {% endblock %}
@@ -26,7 +26,7 @@ Methods table
 .. autosummary::
 {% for item in methods %}
     {%- if item != '__init__' %}
-    ~{{ fullname }}.{{ item }}
+    ~{{ objname }}.{{ item }}
     {%- endif -%}
 {%- endfor %}
 {% endif %}

docs/api/index.md

@@ -12,6 +12,7 @@ import rapids_singlecell as rsc
 
 scanpy_gpu
 squidpy_gpu
+pertpy_gpu
 decoupler_gpu
 get
 ```

docs/api/pertpy_gpu.md

@@ -0,0 +1,39 @@
+# pertpy-GPU: `ptg`
+
+[pertpy](https://pertpy.readthedocs.io) contains tools for perturbation analysis. {mod}`rapids_singlecell.ptg` accelerates some of these methods.
+
+```{eval-rst}
+.. module:: rapids_singlecell.ptg
+.. currentmodule:: rapids_singlecell.ptg
+```
+
+## Distance
+
+```{eval-rst}
+.. autosummary::
+    :toctree: generated
+
+    Distance
+```
+
+```{eval-rst}
+.. autoclass:: Distance
+    :no-index:
+
+    .. rubric:: Methods
+
+    .. autosummary::
+
+        ~Distance.pairwise
+        ~Distance.onesided_distances
+        ~Distance.bootstrap
+
+    .. automethod:: __call__
+        :no-index:
+    .. automethod:: pairwise
+        :no-index:
+    .. automethod:: onesided_distances
+        :no-index:
+    .. automethod:: bootstrap
+        :no-index:
+```

docs/conf.py

@@ -8,6 +8,7 @@
 from pathlib import Path, PurePosixPath
 from typing import TYPE_CHECKING
 import anndata  # noqa
+import fast_array_utils  # noqa
 
 if TYPE_CHECKING:
     from sphinx.application import Sphinx
@@ -119,6 +120,7 @@
     "pymde": ("https://pymde.org", None),
     "scanpy": ("https://scanpy.readthedocs.io/en/stable/", None),
     "squidpy": ("https://squidpy.readthedocs.io/en/stable/", None),
+    "pertpy": ("https://pertpy.readthedocs.io/en/stable/", None),
     "seaborn": ("https://seaborn.pydata.org/", None),
     "decoupler": ("https://decoupler.readthedocs.io/en/latest/", None),
     "rmm": ("https://docs.rapids.ai/api/rmm/stable/", None),
@@ -172,6 +174,7 @@
     ("py:class", "scipy.sparse.base.spmatrix"),
     ("py:meth", "pandas.DataFrame.iloc"),
     ("py:meth", "pandas.DataFrame.loc"),
+    ("py:class", "pandas.core.series.Series"),
     ("py:class", "anndata._core.views.ArrayView"),
     ("py:class", "anndata._core.raw.Raw"),
     ("py:class", "scanpy._utils.Empty"),

docs/release-notes/0.13.6.md

@@ -0,0 +1,5 @@
+### 0.13.6 {small}`the-future`
+
+```{rubric} Features
+```
+* Adds `ptg.Distance` for GPU-accelerated distance computation between groups of cells, compatible with pertpy's Distance API {pr}`454` {smaller}`S Dicks & S Özleyen`

docs/release-notes/index.md

@@ -3,6 +3,8 @@
 # Release notes
 
 ## Version 0.13.0
+```{include} /release-notes/0.13.6.md
+```
 ```{include} /release-notes/0.13.5.md
 ```
 ```{include} /release-notes/0.13.4.md

src/rapids_singlecell/__init__.py

@@ -2,7 +2,7 @@
 
 import cuml.internals.logger as logger
 
-from . import dcg, get, gr, pp, tl
+from . import dcg, get, gr, pp, ptg, tl
 from ._version import __version__
 
 logger.set_level(2)

src/rapids_singlecell/pertpy_gpu/__init__.py

@@ -0,0 +1,3 @@
+from __future__ import annotations
+
+from ._distance import Distance, MeanVar