fix: correct conversions for unitless patterns

Author: etienneguevel

edsnlp/pipes/misc/quantities/patterns.py

@@ -1841,6 +1841,20 @@
         "terms": ["gt", "goutte", "gouttes"],
         "followed_by": None,
     },
+    "µl": {
+        "dim": "length",
+        "degree": 3,
+        "scale": 0.001,
+        "terms": [
+            "micro litre",
+            "micro-litre",
+            "microlitre",
+            "microlitres",
+            "microlitres",
+            "µl",
+        ],
+        "followed_by": None,
+    },
     "ml": {
         "dim": "length",
         "degree": 3,

edsnlp/pipes/misc/quantities/quantities.py

@@ -116,7 +116,9 @@ def parse_unit(self, unit: str) -> Tuple[str, float]:
         for part in regex.split("(?<!per)_", unit):
             unit_config = self.config[unicodedata.normalize("NFKC", part)]
             degrees[unit_config["dim"]].append(unit_config["degree"])
-            scale *= unit_config["scale"] ** abs(unit_config["degree"])
+
+            # degree minimal 1 -> accompts for unitless patterns such as x10*9
+            scale *= unit_config["scale"] ** max(1, abs(unit_config["degree"]))
         degrees = {
             k: sum(v) if len(set(v)) > 1 else v[0]
             for k, v in degrees.items()

tests/pipelines/misc/test_quantities.py

@@ -367,6 +367,7 @@ def test_conversions(blank_nlp, matcher: QuantitiesMatcher):
         ("10 l", "cm3", 10000),
         ("10 l", "cl", 1000),
         ("25 kg/m2", "kg_per_cm2", 0.0025),
+        ("2.4 x10*9µl", "l", 2400),
     ]
 
     for text, unit, expected in tests: