Merge pull request #12

Improve the parsing of list of quantities, and allow the parsing of percentage values.

Author: dylanwal

src/unit_parse/config.py

@@ -112,7 +112,6 @@ def __init__(self):
             ["cu m", "m**3"],  # pint gets confused
             ["cu cm", "cm**3"],  # pint gets confused
             ["cu mm", "mm**3"],  # pint gets confused
-            ["[0-9]{1,5} ?%", ""]
         ]
 
         self.pre_proc_split = [";"]

src/unit_parse/pre_processing_multiple.py

@@ -3,7 +3,7 @@
 
 from unit_parse.config import config
 from unit_parse.logger import log_debug, log_info
-from unit_parse.utils import remove_empty_str
+from unit_parse.utils import contains_number, remove_empty_str
 
 
 @log_info
@@ -52,6 +52,33 @@ def multiple_quantities(text_in: str, sep: list[str]) -> List[str]:
     result = re.split(sep, text_in)
     return [text.strip() for text in result]
 
+def split_on_quantities(text_in: str) -> list[str]:
+    """
+    Split the string into a list of strings, where each string contains a single quantity.
+
+    Examples
+    --------
+    '18 mm Hg @ 68 °F' --> ['18 mm Hg @', '68 °F']
+    'Melting point: 75% -17.5 °C' --> ['Melting point: 75%', '-17.5 °C']
+    'Pass me a 300 ml beer.' --> ['Pass me a 300 ml beer.']
+    """
+    # Use regular expression to split the input text into possible groups of quantities
+    # The pattern looks for spaces (\s) followed by a digit [-]?(\d)
+    # The positive lookahead (?=...) ensures that the split happens without
+    # consuming the digit
+    quantities = re.split(r'(\s+)(?=[-]?\d)', text_in)
+
+    # This regex will sometimes produce groups of just text, so merge subsequent groups until
+    # each group contains a number. This could be done in a more complex regex,
+    # but a loop is pretty simple.
+    results = []
+    for result in quantities:
+        if results and not contains_number(results[-1]):
+            results[-1] = results[-1] + result
+        else:
+            results.append(result)
+    return results
+
 
 @log_debug
 def condition_finder(text_in: str) -> List[str]:
@@ -89,7 +116,8 @@ def condition_finder(text_in: str) -> List[str]:
             result = re.split("@", text)
             out2 += [t.strip() for t in result]
         else:
-            out2.append(text)
+            result = split_on_quantities(text)
+            out2 += result
 
     return [text.strip() for text in out2]
 

tests/test_main.py

@@ -14,6 +14,7 @@
     ['40°F', Quantity('40 degF')],
     ['20.80 mmHg', Quantity('20.80 mmHg')],
     ['20.80 mm Hg', Quantity('20.80 mmHg')],  # correcting a unit that pint gets wrong
+    ['10%', Quantity('10 %')],
 
     # scientific notation
     ["15*10**2 s", Quantity("15*10**2 s")],  # standard
@@ -76,7 +77,11 @@
      [[Quantity('18 mmHg'), Quantity('68 degF')], [Quantity('20 mmHg'), Quantity('77 degF')]]],
     ["Low threshold= 13.1150 mg/cu m; High threshold= 26840 mg/cu m; Irritating concn= 22875 mg/cu m.",
      Quantity('22875 mg/m**3')],
-    ['Melting point: 75% -17.5 °C; 80% 4.6 °C; 85% 21 °C.', Quantity("4.6 degC")],
+    ['Melting point: 75% -17.5 °C; 80% 4.6 °C; 85% 21 °C.', [
+        [Quantity('75%'), Quantity('-17.5 degC')],
+        [Quantity("80%"), Quantity('4.6 degC')],
+        [Quantity("85%"), Quantity('21 degC')],
+    ]],
 
     # ranges
     ['115.2-115.3 °C', Quantity('115.2 degC')],

tests/test_pre_processing_multiple.py

@@ -35,6 +35,8 @@ def test_reduce_parenthesis(input_, output_):
     ['18 mm Hg @ 68 °F ', ['18 mm Hg', '68 °F']],
     ['20 mm Hg @ 77° F', ['20 mm Hg', '77° F']],
     [' 20 mm Hg @ 77° F (NTP, 1992)', ['20 mm Hg', '77° F', 'NTP, 1992']],
+    ['Melting point: 75% -17.5 °C', ['Melting point: 75%', '-17.5 °C']],
+    ['20.8 mm Hg 25 °C', ['20.8 mm Hg', '25 °C']],
 
     ['40 °F (4 °C) (Closed cup)', ['40 °F', '4 °C', 'Closed cup']],
     ['40 °F (4 °C)', ['40 °F', '4 °C']],
@@ -45,7 +47,6 @@ def test_reduce_parenthesis(input_, output_):
     ['(4 °C Closed cup)', ['4 °C Closed cup']],
 
     # negative control (fails)
-    ['20.8 mm Hg 25 °C', ['20.8 mm Hg 25 °C']],
     ['20.8 mm Hgat25 °C', ['20.8 mm Hgat25 °C']],
     ['Pass me a 300 ml beer.', ['Pass me a 300 ml beer.']],
     ["42.3 gcm-3", ["42.3 gcm-3"]],