feat: tests for materials

pipeline/beam_etl/helpers/materials.py

@@ -0,0 +1,81 @@
+"""Functions to parse the materials string into a list of materials that we can
+score later
+
+Scoreable materials are:
+  * metal
+  * wood
+  * glass
+  * resin
+  * fabric
+  * plastic
+
+I will try to match materials found into the specification into one of the
+scoreable ones. However, I found a few, like "stoneware" and "cardboard" that I
+can't fid there, they'll have to remain unscored for now
+"""
+
+from typing import Optional, List
+import re
+
+MATERIAL_MAPPING = {
+    "polyester": "fabric",
+    "spandex": "fabric",
+    "leather": "fabric",
+    "crystal": "glass",
+    "hardwood": "wood",
+    "plywood": "wood",
+    "mdf": "wood",
+    "wood": "wood",
+    "steel": "metal",
+    "polycarbonate": "plastic",
+    "polypropylene": "plastic",
+    "pvc": "plastic",
+    "resin": "plastic",
+    "stoneware": "stoneware",
+    "cardboard": "cardboard",
+    "paper": "cardboard",
+}
+
+
+def material_classifier(material: str) -> str:
+    """I will to match materials to one of the following:
+    * metal
+    * wood
+    * glass
+    * resin
+    * fabric
+    * plastic
+    * cardboard
+    * paper
+    There's a fair amount of variation in the names, so I just try to see if
+    a given keyword identifying a material is found.
+    If no match found, just return the input string itself.
+    """
+
+    for key, value in MATERIAL_MAPPING.items():
+        if key in material:
+            return value
+    return material
+
+
+def clean_material_name(material: str) -> str:
+    """Do some cleaning to material names like:
+      * removing annotations found inside parentheses
+      * removing mounts and percentages
+      * keep the names as lowercase
+    So that everything is more homogenious"""
+    no_paren_annotations = re.sub(r"\(.*\)", "", material)
+    no_amounts = re.sub(r"\d+%?", "", no_paren_annotations)
+    return no_amounts.strip().lower()
+
+
+def parse_materials(materials: Optional[str]) -> Optional[List[str]]:
+    """Parse a string of materials as specified in raw_specifications into a
+    list of standardized material names.
+    Return a sorted list of unique materials for more consistency"""
+    if materials is None:
+        return None
+    material_ls = [
+        material_classifier(clean_material_name(x)) for x in materials.split(",")
+    ]
+    return sorted(set(material_ls))

pipeline/beam_etl/tests/test_materials.py

@@ -0,0 +1,37 @@
+"""Test the `parse_materials` function and its helpers"""
+
+from helpers.materials import parse_materials, clean_material_name, material_classifier
+
+
+def test_none():
+    """Test None value"""
+    assert parse_materials(None) is None
+
+
+def test_amounts():
+    """Test a materials string containing amounts"""
+    assert parse_materials(" 83% Recycled Polyester, 17% Spandex") == ["fabric"]
+    assert clean_material_name(" 83% Recycled Polyester") == "recycled polyester"
+    assert material_classifier("recycled polyester") == "fabric"
+
+
+def test_annotations():
+    """Test a materials string containing annotations between parentheses"""
+    assert parse_materials(" Cardboard (Frame)") == ["cardboard"]
+    assert clean_material_name(" Cardboard (Frame)") == "cardboard"
+    assert material_classifier("cardboard") == "cardboard"
+
+
+def test_keyword():
+    """Test a string where the material is a word inside the string"""
+    assert parse_materials(" Walnut Wood (Frame)") == ["wood"]
+    assert clean_material_name("walnut wood") == "walnut wood"
+    assert material_classifier("walnut wood") == "wood"
+
+
+def test_multiple_materials():
+    """Test a string with two different materials"""
+    assert parse_materials(" MDF (Medium-Density Fiberboard), Metal (Frame)") == [
+        "metal",
+        "wood",
+    ]