Merge branch 'main' into fix_issue_108_narwhals_pandas_polars_support

CONTRIBUTING.md

@@ -0,0 +1,84 @@
+# Contributing to `marginaleffects` for Python
+
+Thank you for your interest in contributing to the `marginaleffects` Python project!
+
+This document outlines guidelines and instructions for setting up the development environment, running tests, and ensuring your code follows the project standards.
+
+## Development Environment
+
+We manage the development environment using [`uv`](https://docs.astral.sh/uv/guides/projects/#creating-a-new-project) for dependency management, task execution, and testing. All necessary information is in the `.toml` and `.lock` files, and `uv` will handle dependency management automatically.
+
+### Setting Up the Environment
+
+Clone the repository
+
+```bash
+git clone [email protected]:vincentarelbundock/pymarginaleffects.git
+cd pymarginaleffects
+```
+
+Create and activate the virtual environment using `uv`:
+
+```bash
+uv venv .venv
+source .venv/bin/activate.sh
+```
+
+Install the project and its dependencies:
+
+```bash
+uv pip install .
+```
+
+### Running Tests
+
+We use `pytest` for testing. The testing environment is fully managed by `uv`. To run the tests, simply execute:
+
+```bash
+uv run --all-extras pytest
+```
+
+`uv` will automatically install any missing dependencies before running the tests.
+
+### Code Formatting and Linting
+
+This project follows strict code formatting and linting rules enforced by `ruff`. Before submitting any code, make sure to run `ruff` to format your files and check for linting errors.
+
+You can run `ruff` by executing:
+
+```bash
+uv run --all-extras ruff check marginaleffects
+uv run --all-extras ruff format marginaleffects
+```
+
+Make sure there are no linting errors before submitting your changes.
+
+## Contribution Guidelines
+
+0. Be nice!
+1. Fork the repository and create a new branch for your feature or bugfix.
+2. Ensure all your changes are well-documented in the code, documentation, and changelog (if necessary).
+3. Write tests for any new functionality or changes you make.
+4. Run the tests and ensure everything passes before submitting your pull request.
+5. Ensure your code is formatted and linted using `ruff`.
+6. Push your changes to your forked repository.
+7. Create a pull request (PR) on the main repository.
+8. Make sure to include a clear description of your changes and reference any related issues (if applicable).
+
+We appreciate your contributions and look forward to reviewing your pull requests!
+
+
+## Documentaion Guidelines
+
+* Follow the Numpydoc docstring style for the documentation, see [link](https://numpydoc.readthedocs.io/en/latest/format.html)
+
+## Makefile
+
+The `marginaleffects` repository includes a `Makefile` to facilitate some common tasks.
+
+### Windows
+
+1. Install make for Windows https://gnuwin32.sourceforge.net/packages/make.htm.
+2. Add it to your path variable.
+3. Then you can use scripts specified in the Makefile. e.g. `make test`
+

marginaleffects/classes.py

@@ -1,5 +1,4 @@
 import polars as pl
-import narwhals as nw
 from .narwhals_utils import is_nw
 
 class MarginaleffectsDataFrame(pl.DataFrame):
@@ -85,8 +84,13 @@ def __str__(self):
         tmp = self.select(valid).rename(self.mapping)
         for col in tmp.columns:
             if tmp[col].dtype.is_numeric():
-                tmp = tmp.with_columns(
-                    pl.col(col).map_elements(lambda x: f"{x:.3g}", return_dtype=pl.Utf8)
+
+                def fmt(x):
+                    out = pl.Series([f"{i:.3g}" for i in x])
+                    return out
+
+                tmp.with_columns(
+                    pl.col(col).map_batches(fmt, return_dtype=pl.Utf8).alias(col)
                 )
         out += tmp.__str__()
         out = out + f"\n\nColumns: {', '.join(self.columns)}\n"

marginaleffects/comparisons.py

@@ -202,7 +202,7 @@ def comparisons(
     # ugly hack, but polars is very strict and `value / 2`` is float
     nd, hi, lo, pad = [ to_from_native(upcast, df) for df in [nd, hi, lo, pad] ]
     nd, hi, lo = map(
-        lambda l: nw.from_native(pl.concat([df.to_native() for df in l], how="vertical_relaxed")), 
+        lambda el: nw.from_native(pl.concat([df.to_native() for df in el], how="vertical_relaxed")), 
         [nd, hi, lo]
         )
     # hi = pl.concat(hi, how="vertical_relaxed")
@@ -220,7 +220,7 @@ def comparisons(
         pad = nw.concat(pad).unique()
         # nd, hi, lo = [ to_from_native(pl.concat, upcast([pad, df]), how="diagonal") for df in [nd, hi, lo] ]
         nd, hi, lo = map(
-            lambda l: nw.from_native(pl.concat(upcast([nw.to_native(pad), nw.to_native(l)]), how="diagonal")), 
+            lambda el: nw.from_native(pl.concat(upcast([nw.to_native(pad), nw.to_native(el)]), how="diagonal")), 
             [nd, hi, lo])
         # hi = pl.concat(upcast([pad, hi]), how="diagonal")
         # lo = pl.concat(upcast([pad, lo]), how="diagonal")

marginaleffects/equivalence.py

@@ -29,19 +29,19 @@ def get_equivalence(
     if np.isinf(df):
         x = x.with_columns(
             pl.col("statistic_noninf")
-            .map_elements(lambda x: 1 - norm.cdf(x), return_dtype=pl.Float64)
+            .map_batches(lambda x: 1 - norm.cdf(x), return_dtype=pl.Float64)
             .alias("p_value_noninf"),
             pl.col("statistic_nonsup")
-            .map_elements(lambda x: norm.cdf(x), return_dtype=pl.Float64)
+            .map_batches(lambda x: norm.cdf(x), return_dtype=pl.Float64)
             .alias("p_value_nonsup"),
         )
     else:
         x = x.with_columns(
             pl.col("statistic_noninf")
-            .map_elements(lambda x: 1 - t.cdf(x), return_dtype=pl.Float64)
+            .map_batches(lambda x: 1 - t.cdf(x), return_dtype=pl.Float64)
             .alias("p_value_noninf"),
             pl.col("statistic_nonsup")
-            .map_elements(lambda x: t.cdf(x), return_dtype=pl.Float64)
+            .map_batches(lambda x: t.cdf(x), return_dtype=pl.Float64)
             .alias("p_value_nonsup"),
         )
 

marginaleffects/uncertainty.py

@@ -82,9 +82,6 @@ def get_z_p_ci(df, model, conf_level, hypothesis_null=0):
                     dtype=nw.Float64,
                     native_namespace=nw.get_native_namespace(df),
                 )
-                # nw.col("p_value")
-                # .map_elements(lambda x: -np.log2(x), return_dtype=nw.Float64)
-                # .alias("s_value")
             )
         except Exception as e:
             print(f"An exception occurred: {e}")

pyproject.toml

@@ -1,11 +1,12 @@
 [project]
 name = "marginaleffects"
-version = "0.0.13"
+version = "0.0.13.1"
 description = "Predictions, counterfactual comparisons, slopes, and hypothesis tests for statistical models."
 readme = "README.md"
 requires-python = ">=3.10"
 dependencies = [
     "numpy >=2.0.0",
+    "narwhals",
     "patsy >=0.5.6",
     "polars >=1.7.0",
     "pyarrow >=17.0.0",