fix: add unified regex w/ float support and tests

changelog.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## Unreleased
+
+### Fixed
+
+- Measurements now correctly match "0.X", "0.XX", ... numbers
+- Typo in "celsius" measurement unit
+
 ## v0.10.2
 
 ### Changed

edsnlp/pipes/misc/measurements/factory.py

@@ -7,6 +7,7 @@
     measurements=list(patterns.common_measurements.keys()),  # noqa: E501
     units_config=patterns.units_config,
     number_terms=patterns.number_terms,
+    number_regex=patterns.number_regex,
     stopwords=patterns.stopwords,
     unit_divisors=patterns.unit_divisors,
     ignore_excluded=True,

edsnlp/pipes/misc/measurements/measurements.py

@@ -481,6 +481,7 @@ def __init__(
           measurements: Union[str, List[Union[str, MsrConfig]], Dict[str, MsrConfig]] = list(patterns.common_measurements.keys()),  # noqa: E501
           units_config: Dict[str, UnitConfig] = patterns.units_config,
           number_terms: Dict[str, List[str]] = patterns.number_terms,
+          number_regex: str = patterns.number_regex,
           stopwords: List[str] = patterns.stopwords,
           unit_divisors: List[str] = patterns.unit_divisors,
           ignore_excluded: bool = True,
@@ -574,16 +575,7 @@ def __init__(
                     self.unitless_patterns[pattern_name] = {"name": name, **pattern}
 
         # NUMBER PATTERNS
-        one_plus = "[1-9][0-9]*"
-        self.regex_matcher.add(
-            "number",
-            [
-                rf"(?<![0-9][.,]?){one_plus}([ ]\d{{3}})*[ ]+(?:[,.][ ]+\d+)?",
-                rf"(?<![0-9][.,]?){one_plus}([ ]\d{{3}})*(?:[,.]\d+)?",
-                rf"(?<![0-9][.,]?){one_plus}([ ]/[ ]|/){one_plus}",
-                r"(?<![0-9][.,]?)00?([,.]\d+)?",                
-            ],
-        )
+        self.regex_matcher.add("number",[number_regex])
         self.number_label_hashes = {nlp.vocab.strings["number"]}
         for number, terms in number_terms.items():
             self.term_matcher.build_patterns(nlp, {number: terms})

edsnlp/pipes/misc/measurements/patterns.py

@@ -38,6 +38,25 @@
     "1000": ["mille", "milles"],
 }
 
+
+number_regex = r"""(?x)
+# no digit or floating point number prefix before
+(?<![0-9][.,]?)
+# integer part like 123 or 1 234
+(?:
+    0
+    |[1-9][0-9]*(?:\ \d{3})*
+)
+(?:
+    # floating point surounded by spaces
+    \ +[,.]\ +\d+
+    # floating point w/o space
+    | [,.]\d+
+    # fractions
+    | (?:\ /\ |/)[1-9][0-9]*(?:\ \d{3})*
+)?"""
+
+
 units_config = {
     # Lengths
     "µm": {

tests/pipelines/misc/test_measurements.py

@@ -240,3 +240,21 @@ def test_merge_intersect(blank_nlp, matcher: MeasurementsMatcher):
     assert len(doc.spans["measurements"]) == 2
     assert [doc.ents[0].text, doc.ents[1].text] == ["2.0cm", "3cm"]
     assert [doc.ents[0]._.value.cm, doc.ents[1]._.value.cm] == [2.0, 3]
+
+
+def test_measurement_snippets(blank_nlp, matcher: MeasurementsMatcher):
+    for text, result in [
+        ("0.50g", ["0.5 g"]),
+        ("0.050g", ["0.05 g"]),
+        ("1 m 50", ["1.5 m"]),
+        ("1.50 m", ["1.5 m"]),
+        ("1,50m", ["1.5 m"]),
+        ("2.0cm x 3cm", ["2.0 cm", "3 cm"]),
+        ("2 par 1mm", ["2 mm", "1 mm"]),
+        ("8, 13 et 15dm", ["8 dm", "13 dm", "15 dm"]),
+        ("1 / 50  kg", ["0.02 kg"]),
+    ]:
+        doc = blank_nlp(text)
+        doc = matcher(doc)
+
+        assert [str(span._.value) for span in doc.spans["measurements"]] == result