feat: add date columns transformer (#68)

docs/README.md

@@ -3,10 +3,9 @@
 # sk-transformers
 ***A collection of various pandas & scikit-learn compatible transformers for all kinds of preprocessing and feature engineering steps*** 🛠
 
-[![testing](https://github.com/chrislemke/sk-transformers/actions/workflows/testing.yml/badge.svg?branch=main)](https://github.com/chrislemke/sk-transformers/actions/workflows/testing.yml)
-[![Codacy Badge](https://app.codacy.com/project/badge/Grade/44093b8f6b28454fb8e0336ccb83cdc5)](https://www.codacy.com/gh/chrislemke/sk-transformers/dashboard?utm_source=github.com&utm_medium=referral&utm_content=chrislemke/sk-transformers&utm_campaign=Badge_Grade)
+[![ChecksAndTesting](https://github.com/chrislemke/sk-transformers/actions/workflows/checks-testing.yml/badge.svg)](https://github.com/chrislemke/sk-transformers/actions/workflows/checks-testing.yml)
 [![codecov](https://codecov.io/github/chrislemke/sk-transformers/branch/main/graph/badge.svg?token=LJLXQXX6M8)](https://codecov.io/github/chrislemke/sk-transformers)
-[![deploy package](https://github.com/chrislemke/sk-transformers/actions/workflows/deploy-package.yml/badge.svg)](https://github.com/chrislemke/sk-transformers/actions/workflows/deploy-package.yml)
+[![Release](https://github.com/chrislemke/sk-transformers/actions/workflows/release.yml/badge.svg)](https://github.com/chrislemke/sk-transformers/actions/workflows/release.yml)
 [![pypi](https://img.shields.io/pypi/v/sk-transformers)](https://pypi.org/project/sk-transformers/)
 [![python version](https://img.shields.io/pypi/pyversions/sk-transformers?logo=python&logoColor=yellow)](https://www.python.org/)
 [![downloads](https://img.shields.io/pypi/dm/sk-transformers)](https://pypistats.org/packages/sk-transformers)
@@ -50,6 +49,7 @@ poetry install
 ## Available transformers
 | Module | Transformer | Description |
 | ------ | ----------- | ----------- |
+|[`Datetime transformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/datetime_transformer/)|[`DateColumnsTransformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/datetime_transformer/#sk_transformers.datetime_transformer.DateColumnsTransformer)|Splits a date column into multiple columns.|
 |[`Datetime transformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/datetime_transformer/)|[`DurationCalculatorTransformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/datetime_transformer/#sk_transformers.datetime_transformer.DurationCalculatorTransformer)|Calculates the duration between to given dates.|
 |[`Deep transformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/deep_transformer/)|[`ToVecTransformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/deep_transformer/#sk_transformers.deep_transformer.ToVecTransformer)|This transformer trains an [FT-Transformer](https://paperswithcode.com/method/ft-transformer) using the [pytorch-widedeep package](https://github.com/jrzaurin/pytorch-widedeep) and extracts the embeddings from its embedding layer.|
 |[`Encoder transformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/encoder_transformer/)|[`MeanEncoderTransformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/encoder_transformer/#sk_transformers.encoder_transformer.MeanEncoderTransformer)|Scikit-learn API for the [feature-engine MeanEncoder](https://feature-engine.readthedocs.io/en/latest/api_doc/encoding/MeanEncoder.html).|

examples/playground.ipynb

@@ -39,6 +39,30 @@
     "## [Datetime transformer](https://chrislemke.github.io/sk-transformers/API-reference/transformer/datetime_transformer/)"
    ]
   },
+  {
+   "attachments": {},
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "### [`DateColumnsTransformer`](https://chrislemke.github.io/sk-transformers/API-reference/transformer/datetime_transformer/#sk_transformers.datetime_transformer.DateColumnsTransformer)\n",
+    "\n",
+    "Splits a date column into multiple columns."
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "import pandas as pd\n",
+    "from sk_transformers import DateColumnsTransformer\n",
+    "\n",
+    "X = pd.DataFrame({\"foo\": [\"2021-01-01\", \"2022-02-02\", \"2023-03-03\"]})\n",
+    "transformer = DateColumnsTransformer([\"foo\"])\n",
+    "transformer.fit_transform(X)"
+   ]
+  },
   {
    "attachments": {},
    "cell_type": "markdown",

requirements.txt

@@ -1205,9 +1205,9 @@ smart-open==6.3.0 ; python_version >= "3.8" and python_version < "3.11" \
 smmap==5.0.0 ; python_version >= "3.8" and python_version < "3.11" \
     --hash=sha256:2aba19d6a040e78d8b09de5c57e96207b09ed71d8e55ce0959eeee6c8e190d94 \
     --hash=sha256:c840e62059cd3be204b0c9c9f74be2c09d5648eddd4580d9314c3ecde0b30936
-spacy-legacy==3.0.11 ; python_version >= "3.8" and python_version < "3.11" \
-    --hash=sha256:7b2a72bfe8e135c5885ecf22db946daa352c7a24639aaeda10a76d4c1c66196f \
-    --hash=sha256:bfaef03c377c323a3089b1885518e0dad489597b07a80af2499750b24fdbf54b
+spacy-legacy==3.0.12 ; python_version >= "3.8" and python_version < "3.11" \
+    --hash=sha256:476e3bd0d05f8c339ed60f40986c07387c0a71479245d6d0f4298dbd52cda55f \
+    --hash=sha256:b37d6e0c9b6e1d7ca1cf5bc7152ab64a4c4671f59c85adaf7a3fcb870357a774
 spacy-loggers==1.0.4 ; python_version >= "3.8" and python_version < "3.11" \
     --hash=sha256:e050bf2e63208b2f096b777e494971c962ad7c1dc997641c8f95c622550044ae \
     --hash=sha256:e6f983bf71230091d5bb7b11bf64bd54415eca839108d5f83d9155d0ba93bf28

src/sk_transformers/__init__.py

@@ -1,4 +1,5 @@
 from sk_transformers.datetime_transformer import (
+    DateColumnsTransformer,
     DurationCalculatorTransformer,
     TimestampTransformer,
 )

src/sk_transformers/datetime_transformer.py

@@ -7,6 +7,116 @@
 from sk_transformers.utils import check_ready_to_transform
 
 
+class DateColumnsTransformer(BaseTransformer):
+    """Splits a date column into multiple columns.
+
+    Example:
+    ```python
+    import pandas as pd
+    from sk_transformers import DateColumnsTransformer
+
+    X = pd.DataFrame({"foo": ["2021-01-01", "2022-02-02", "2023-03-03"]})
+    transformer = DateColumnsTransformer(["foo"])
+    transformer.fit_transform(X)
+    ```
+    ```
+             foo  foo_year  ...  foo_is_year_end  foo_is_weekend
+    0 2021-01-01      2021  ...            False           False
+    1 2022-02-02      2022  ...            False           False
+    2 2023-03-03      2023  ...            False           False
+    ```
+
+    Args:
+        features (List[str]): List of columns to transform.
+        date_format (str): Date format. Defaults to `%Y-%m-%d`.
+        errors (str): How to handle errors in `pd.to_datetime`. Defaults to `raise`.
+            available values: `ignore`, `raise`, `coerce`.
+            If `raise`, then invalid parsing will raise an exception.
+            If `coerce`, then invalid parsing will be set as `NaT`.
+            If `ignore`, then invalid parsing will return the input.
+        date_elements ([List[str]]): List of date elements to extract.
+    """
+
+    def __init__(  # pylint: disable=dangerous-default-value
+        self,
+        features: List[str],
+        date_format: str = "%Y-%m-%d",
+        errors: str = "raise",
+        date_elements: List[str] = [
+            "year",
+            "month",
+            "day",
+            "day_of_week",
+            "day_of_year",
+            "week_of_year",
+            "quarter",
+            "is_leap_year",
+            "is_month_start",
+            "is_month_end",
+            "is_quarter_start",
+            "is_quarter_end",
+            "is_year_start",
+            "is_year_end",
+            "is_weekend",
+        ],
+    ) -> None:
+        super().__init__()
+        self.features = features
+        self.date_format = date_format
+        self.date_elements = date_elements
+        self.errors = errors
+
+    def transform(  # pylint: disable=too-many-branches
+        self, X: pd.DataFrame
+    ) -> pd.DataFrame:
+        """Transforms columns from the provided dataframe.
+
+        Args:
+            X (pandas.DataFrame): Dataframe with columns to transform.
+
+        Returns:
+            pandas.DataFrame: Dataframe with transformed columns.
+        """
+
+        X = check_ready_to_transform(self, X, self.features)
+
+        for column in self.features:
+            X[column] = pd.to_datetime(
+                X[column], format=self.date_format, errors=self.errors
+            )
+            if "year" in self.date_elements:
+                X[f"{column}_year"] = X[column].dt.year
+            if "month" in self.date_elements:
+                X[f"{column}_month"] = X[column].dt.month
+            if "day" in self.date_elements:
+                X[f"{column}_day"] = X[column].dt.day
+            if "day_of_week" in self.date_elements:
+                X[f"{column}_day_of_week"] = X[column].dt.dayofweek
+            if "day_of_year" in self.date_elements:
+                X[f"{column}_day_of_year"] = X[column].dt.dayofyear
+            if "week_of_year" in self.date_elements:
+                X[f"{column}_week_of_year"] = X[column].dt.weekofyear
+            if "quarter" in self.date_elements:
+                X[f"{column}_quarter"] = X[column].dt.quarter
+            if "is_leap_year" in self.date_elements:
+                X[f"{column}_is_leap_year"] = X[column].dt.is_leap_year
+            if "is_month_start" in self.date_elements:
+                X[f"{column}_is_month_start"] = X[column].dt.is_month_start
+            if "is_month_end" in self.date_elements:
+                X[f"{column}_is_month_end"] = X[column].dt.is_month_end
+            if "is_quarter_start" in self.date_elements:
+                X[f"{column}_is_quarter_start"] = X[column].dt.is_quarter_start
+            if "is_quarter_end" in self.date_elements:
+                X[f"{column}_is_quarter_end"] = X[column].dt.is_quarter_end
+            if "is_year_start" in self.date_elements:
+                X[f"{column}_is_year_start"] = X[column].dt.is_year_start
+            if "is_year_end" in self.date_elements:
+                X[f"{column}_is_year_end"] = X[column].dt.is_year_end
+            if "is_weekend" in self.date_elements:
+                X[f"{column}_is_weekend"] = X[column].dt.dayofweek.isin([5, 6])
+        return X
+
+
 class DurationCalculatorTransformer(BaseTransformer):
     """Calculates the duration between to given dates.
 

tests/conftest.py

@@ -97,6 +97,11 @@ def X_time_values() -> pd.DataFrame:
     )
 
 
+@pytest.fixture()
+def tiny_date_df() -> pd.DataFrame:
+    return pd.DataFrame({"a": ["2021-01-01", "2022-02-02", "2023-03-03"]})
+
+
 @pytest.fixture()
 def X_nan_values() -> pd.DataFrame:
     return pd.DataFrame(

tests/test_transformer/test_datetime_transformer.py

@@ -2,11 +2,98 @@
 import pytest
 from sklearn.pipeline import make_pipeline
 
-from sk_transformers import DurationCalculatorTransformer, TimestampTransformer
+from sk_transformers import (
+    DateColumnsTransformer,
+    DurationCalculatorTransformer,
+    TimestampTransformer,
+)
 
 # pylint: disable=missing-function-docstring, missing-class-docstring
 
 
+def test_date_columns_transformer_in_pipeline(tiny_date_df):
+    pipeline = make_pipeline(DateColumnsTransformer(["a"]))
+    X = pipeline.fit_transform(tiny_date_df).drop("a", axis=1)
+    expected = np.array(
+        [
+            [
+                2021,
+                1,
+                1,
+                4,
+                1,
+                53,
+                1,
+                False,
+                True,
+                False,
+                True,
+                False,
+                True,
+                False,
+                False,
+            ],
+            [
+                2022,
+                2,
+                2,
+                2,
+                33,
+                5,
+                1,
+                False,
+                False,
+                False,
+                False,
+                False,
+                False,
+                False,
+                False,
+            ],
+            [
+                2023,
+                3,
+                3,
+                4,
+                62,
+                9,
+                1,
+                False,
+                False,
+                False,
+                False,
+                False,
+                False,
+                False,
+                False,
+            ],
+        ],
+        dtype=object,
+    )
+
+    assert np.array_equal(X.to_numpy(), expected)
+    assert pipeline.steps[0][0] == "datecolumnstransformer"
+    assert pipeline.steps[0][1].features == ["a"]
+    assert pipeline.steps[0][1].date_format == "%Y-%m-%d"
+    assert pipeline.steps[0][1].date_elements == [
+        "year",
+        "month",
+        "day",
+        "day_of_week",
+        "day_of_year",
+        "week_of_year",
+        "quarter",
+        "is_leap_year",
+        "is_month_start",
+        "is_month_end",
+        "is_quarter_start",
+        "is_quarter_end",
+        "is_year_start",
+        "is_year_end",
+        "is_weekend",
+    ]
+
+
 def test_duration_calculator_transformer_in_pipeline_seconds(X_time_values) -> None:
     pipeline = make_pipeline(
         DurationCalculatorTransformer(