Update numba requirement (#143)

* Release 1190 (#139)

* Bump version

* Update CHANGELOG

* Fix flake8 version

Goal here is to keep the tests consistent and deterministic. Rellying in
major bumps would lead to unexpected behaviours, like previous code that
used to pass our lint stopping.

* Fix E741

* Pin numba version

* Pin distributed version

* Less restrictive numba requirements

CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## [1.19.1] - 2020-07-13
+- **Bug Fix**
+    - Be less restrictive with Numba requirements
+
+## [1.19.0] - 2020-06-17
+- **Enhancement**
+    - Improve `space_time_split_dataset` performance
+
 ## [1.18.0] - 2020-05-08
 - **Enhancement**
     - Allow users to inform a Placeholder value in imputer learner

requirements.txt

@@ -5,3 +5,4 @@ pandas>=0.24.1,<0.25
 scikit-learn>=0.21.2,<0.22.0
 statsmodels>=0.9.0,<1
 toolz>=0.9.0,<1
+numba>=0.48.0,<0.50

requirements_tools.txt

@@ -1,2 +1,3 @@
 shap>=0.31.0,<=0.34
 swifter>=0.284,<0.300
+distributed>=2.14.0,<2.17.0

scripts/lint.sh

@@ -7,7 +7,7 @@ source $cur_dir/helpers.sh
 
 activate_venv
 
-venv/bin/python3 -m pip install -q flake8
+venv/bin/python3 -m pip install -q flake8==3.8.0
 venv/bin/python3 -m flake8 \
 --ignore=E731,W503 \
 --filename=*.py \

src/fklearn/resources/VERSION

@@ -1 +1 @@
-1.18.0
+1.19.1

src/fklearn/training/pipeline.py

@@ -60,8 +60,8 @@ def _no_variable_args(learner: LearnerFnType, predict_fn: PredictFnType) -> None
                              .format(learner.__name__, ', '.join(var_args)))
 
     # Check for unfilled arguments of learners
-    for l in learners:
-        _has_one_unfilled_arg(l)
+    for learner in learners:
+        _has_one_unfilled_arg(learner)
 
     def pipeline(data: pd.DataFrame) -> LearnerReturnType:
         current_data = data.copy()

src/fklearn/validation/splitters.py

@@ -457,7 +457,7 @@ def spatial_learning_curve_splitter(train_data: pd.DataFrame,
 
     folds_indices = _lc_fold_to_indexes(folds)  # final formatting with idx
 
-    logs = [assoc(l, "percentage", p) for l, p in zip(map(_log_time_fold, folds), train_percentages)]
+    logs = [assoc(learner, "percentage", p) for learner, p in zip(map(_log_time_fold, folds), train_percentages)]
 
     return folds_indices, logs