[#162] Add a rough draft of a RenameVectorAttributes transformer

Usually we don't care about the names of the vector attributes. But LightGBM
uses them as feature names and disallows some characters in the names.
Unfortunately, one of these characters is :, and Spark's Interaction names the
output of an interaction between A and B "A:B". I looked through the Spark code
and didn't see any way to configure the names of these output features. So I
think the easiest way forward here is to make a transformer that renames the
attributes of a vector by removing some characters and replacing them with
another.

hlink/linking/transformers/rename_vector_attributes.py

@@ -0,0 +1,21 @@
+from pyspark.ml import Transformer
+from pyspark.sql import DataFrame
+
+
+class RenameVectorAttributes(Transformer):
+    """
+    A custom transformer which renames the attributes or "slot names" of a
+    given input column of type vector. This is helpful when you don't have
+    complete control over the names of the attributes, but you need them to
+    look a certain way.
+
+    For example, LightGBM can't handle vector attributes with colons in their
+    names. But the Spark Interaction class creates vector attributes named with
+    colons. So we need to rename the attributes and remove the colons before
+    passing the feature vector to LightGBM for training.
+    """
+
+    def __init__(self) -> None: ...
+
+    def _transform(self, dataset: DataFrame) -> DataFrame:
+        return dataset

hlink/tests/transformers_test.py

@@ -0,0 +1,20 @@
+from pyspark.sql import SparkSession
+from pyspark.ml.feature import VectorAssembler
+
+from hlink.linking.transformers.rename_vector_attributes import RenameVectorAttributes
+
+
+def test_rename_vector_attributes(spark: SparkSession) -> None:
+    df = spark.createDataFrame(
+        [[0.0, 1.0], [1.0, 2.0], [3.0, 4.0]], schema=["A", "regionf_0:namelast_jw"]
+    )
+
+    assembler = VectorAssembler(
+        inputCols=["A", "regionf_0:namelast_jw"], outputCol="vectorized"
+    )
+    remove_colons = RenameVectorAttributes()
+    transformed = remove_colons.transform(assembler.transform(df))
+
+    attrs = transformed.schema["vectorized"].metadata["ml_attr"]["attrs"]["numeric"]
+    attr_names = [attr["name"] for attr in attrs]
+    assert attr_names == ["A", "regionf_0_namelast_jw"]