DOC add pyodide-http to enable fetch_openml in jupyterlite

.gitignore

@@ -34,4 +34,7 @@ doc/sg_execution_times.rst
 doc/.jupyterlite.doit.db
 doc/jupyterlite_contents/
 doc/_contents/
-emsdk/
+doc/_output/
+emsdk/
+packages/
+pyodide/

doc/conf.py

@@ -10,6 +10,8 @@
 import importlib.metadata
 from datetime import datetime
 
+from sphinx_gallery.notebook import add_code_cell
+
 # General information about the project.
 project = "fastcan"
 copyright = f"2024 - {datetime.now().year}, fastcan developers (MIT License)"
@@ -44,7 +46,7 @@
     # For PlantUML diagrams
     "sphinxcontrib.plantuml",
     # For interactive examples via JupyterLite
-    'jupyterlite_sphinx',
+    "jupyterlite_sphinx",
 ]
 
 # List of patterns, relative to source directory, that match files and
@@ -77,11 +79,6 @@
     "sklearn": ("https://scikit-learn.org/stable", None),
 }
 
-sphinx_gallery_conf = {
-    "examples_dirs": ["../examples"],
-    "gallery_dirs": ["auto_examples"],
-}
-
 # -----------------------------------------------------------------------------
 # Interactive documentation examples via JupyterLite
 # -----------------------------------------------------------------------------
@@ -92,3 +89,47 @@
     "If you encounter any issues, please report them on the"
     " [fastcan issue tracker](https://github.com/scikit-learn-contrib/fastcan/issues)."
 )
+
+
+def notebook_modification_function(notebook_content, notebook_filename):
+    notebook_content_str = str(notebook_content)
+    dummy_notebook_content = {"cells": []}
+    code_lines = []
+
+    if "fetch_" in notebook_content_str:
+        code_lines.extend(
+            [
+                "import pyodide_http",
+                "pyodide_http.patch_all()",
+            ]
+        )
+    # always import matplotlib and pandas to avoid Pyodide limitation with
+    # imports inside functions
+    code_lines.extend(
+        [
+            "import matplotlib",
+            "import pandas",
+        ]
+    )
+
+    if code_lines:
+        code_lines = ["# JupyterLite-specific code"] + code_lines
+        code = "\n".join(code_lines)
+        add_code_cell(dummy_notebook_content, code)
+
+    notebook_content["cells"] = (
+        dummy_notebook_content["cells"] + notebook_content["cells"]
+    )
+
+
+# -----------------------------------------------------------------------------
+suppress_warnings = [
+    # Ignore new warning in Sphinx 7.3.0 while pickling environment:
+    #   WARNING: cannot cache unpickable configuration value: 'sphinx_gallery_conf'
+    "config.cache",
+]
+sphinx_gallery_conf = {
+    "examples_dirs": ["../examples"],
+    "gallery_dirs": ["auto_examples"],
+    "jupyterlite": {"notebook_modification_function": notebook_modification_function},
+}

doc/environment.yml

@@ -5,4 +5,7 @@ channels:
 dependencies:
   - xeus-python
   - matplotlib
-  - fastcan
+  - pandas
+  - fastcan
+  - pip:
+    - pyodide-http

examples/plot_forecasting.py

@@ -5,9 +5,6 @@
 
 .. currentmodule:: fastcan.narx
 
-.. note::
-    This example can NOT be run with JupyterLite so far.
-
 In this examples, we will demonstrate how to use :func:`make_narx` to build (nonlinear)
 AutoRegressive (AR) models for time-series forecasting.
 The time series used isthe monthly average atmospheric CO2 concentrations

fastcan/narx/tests/test_narx.py

@@ -647,6 +647,7 @@ def test_auto_reg_error():
     with pytest.raises(ValueError, match=r"X should be an array-like of shape.*"):
         model.predict(len(y), y_init=y[: model.max_delay_])
 
+
 def test_predict_ndim():
     """Test the ndim of predict output"""
     X = np.random.rand(10, 2)

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "fastcan"
 version = "0.4.1"
-description = "A fast canonical-correlation-based feature selection method"
+description = "A fast canonical-correlation-based greedy search algorithm"
 authors = [
     { name = "Matthew Sikai Zhang", email = "[email protected]" },
 ]
@@ -59,7 +59,7 @@ channels = ["conda-forge"]
 platforms = ["win-64", "linux-64", "osx-64", "osx-arm64", "linux-aarch64"]
 
 [tool.pixi.dependencies]
-python = ">=3.12.0,<3.13"
+python = ">=3.10"
 scikit-learn = ">=1.7.0,!=1.7.1"
 
 [tool.pixi.pypi-dependencies]
@@ -81,6 +81,7 @@ plantuml = "*"
 [tool.pixi.feature.wasm.dependencies]
 pip = "*"
 pyodide-build = "*"
+prettier = "*"
 
 [tool.pixi.feature.asv.dependencies]
 asv = "*"
@@ -156,19 +157,23 @@ spell = "codespell"
 
 [tool.pixi.feature.docs.tasks]
 doc = { cmd = "{{ SPHINXBUILD }} -M {{ CMD }} {{ SOURCEDIR }} {{ BUILDDIR }} {{ SPHINXOPTS }} --fail-on-warning", cwd = "doc", args = [{ arg = "SPHINXBUILD", default = "sphinx-build" }, { arg = "CMD", default = "html" }, { arg = "SOURCEDIR", default = "." }, { arg = "BUILDDIR", default = "_build" }, { arg = "SPHINXOPTS", default = "" }] }
-doc-clean = { cmd = "rm -rf {{ BUILDDIR }} generated auto_examples jupyterlite_contents .jupyterlite.doit.db _contents", cwd = "doc", args = [{ arg = "BUILDDIR", default = "_build" }] }
+doc-clean = { cmd = "rm -rf {{ BUILDDIR }} generated auto_examples jupyterlite_contents .jupyterlite.doit.db _contents _output .cache", cwd = "doc", args = [{ arg = "BUILDDIR", default = "_build" }] }
 doc-deploy = { cmd = "python -m http.server" , cwd = "doc/_build/html" }
 
 [tool.pixi.feature.nogil.tasks]
 nogil-h = "python -Xgil=0 -m timeit -n 5 -s 'import numpy as np; from fastcan import FastCan; X = np.random.rand(3000, 100); y = np.random.rand(3000, 20)' 's = FastCan(100, verbose=0).fit(X, y)'"
 nogil-eta = "python -Xgil=0 -m timeit -n 5 -s 'import numpy as np; from fastcan import FastCan; X = np.random.rand(3000, 100); y = np.random.rand(3000, 20)' 's = FastCan(100, eta=True, verbose=0).fit(X, y)'"
 
 [tool.pixi.feature.wasm.tasks]
-wasm-clone-emsdk = "bash -c '[ -d emsdk ] || git clone https://github.com/emscripten-core/emsdk.git emsdk'"
-wasm-compatible = "pyodide xbuildenv search -a"
-wasm-toolchain = { cmd = "LATEST_COMPATIBLE=$(pyodide xbuildenv search -a | grep '│ Yes' | head -n 1 | awk -F '│' '{print $2}' | xargs) && pyodide xbuildenv install $LATEST_COMPATIBLE" }
-wasm-setup-emsdk = { cmd = "PYODIDE_EMSCRIPTEN_VERSION=$(pyodide config get emscripten_version) && ./emsdk install $PYODIDE_EMSCRIPTEN_VERSION && ./emsdk activate $PYODIDE_EMSCRIPTEN_VERSION", cwd = "emsdk", depends-on = ["wasm-clone-emsdk", "wasm-toolchain"] }
-wasm-build = { cmd = "bash -c 'source emsdk/emsdk_env.sh && pyodide build'", depends-on = ["wasm-setup-emsdk"] }
+emsdk-clone = "bash -c '[ -d emsdk ] || git clone https://github.com/emscripten-core/emsdk.git emsdk'"
+pyodide-compatible = "pyodide xbuildenv search -a"
+pyodide-toolchain = { cmd = "LATEST_COMPATIBLE=$(pyodide xbuildenv search -a | grep '│ Yes' | head -n 1 | awk -F '│' '{print $2}' | xargs) && pyodide xbuildenv install $LATEST_COMPATIBLE" }
+emsdk-setup = { cmd = "PYODIDE_EMSCRIPTEN_VERSION=$(pyodide config get emscripten_version) && ./emsdk install $PYODIDE_EMSCRIPTEN_VERSION && ./emsdk activate $PYODIDE_EMSCRIPTEN_VERSION", cwd = "emsdk", depends-on = ["emsdk-clone", "pyodide-toolchain"] }
+pyodide-build = { cmd = "bash -c 'source emsdk/emsdk_env.sh && pyodide build'", depends-on = ["emsdk-setup"] }
+pyodide-create-recipe = "rm -rf packages && pyodide skeleton pypi fastcan"
+pyodide-build-recipe = { cmd = "rm -rf dist && bash -c 'source emsdk/emsdk_env.sh && pyodide build-recipes fastcan --install'", depends-on = ["emsdk-setup"] }
+pyodide-download = '''bash -c "[ -d pyodide ] || (LATEST_TAG=$(curl -s https://api.github.com/repos/pyodide/pyodide/releases/latest | grep \"tag_name\" | cut -d \" -f4) && curl -L https://github.com/pyodide/pyodide/releases/download/${LATEST_TAG}/pyodide-${LATEST_TAG}.tar.bz2 | tar -xjf -)"'''
+pyodide-test-recipe = { cmd = "mv ../dist/* . && python -m http.server --directory .", cwd = "pyodide", depends-on = ["pyodide-download", "pyodide-build-recipe"] }
 
 [tool.pixi.environments]
 dev = ["docs", "test", "build", "jupyter", "asv"]

rtd-environment.yml

@@ -5,13 +5,13 @@ dependencies:
   - micromamba
   - pip
   - furo
-  - matplotlib
-  - pandas
   - sphinx-gallery
   - sphinx-design
   - sphinxcontrib-plantuml
   - jupyterlite-sphinx
   - plantuml
+  - matplotlib
+  - pandas
   - pip:
       - .
-      - jupyterlite-xeus >=2.1.2
+      - jupyterlite-xeus

tests/test_fastcan.py

@@ -52,7 +52,7 @@ def test_select_kbest_classif():
     ssc = correlation_filter.scores_.sum()
     # Test whether the ssc from the fastcan is consistent
     # with the mcc from the linear regression
-    assert_almost_equal(actual=ssc, desired=gtruth_ssc)
+    assert abs(ssc - gtruth_ssc) < 1e-5
 
     support = correlation_filter.get_support()
     gtruth = np.zeros(n_features)