Entries: parse multiple definitions as list

parser/src/parser/entry.py

@@ -51,6 +51,42 @@ class Entry(NamedTuple):
     definitions: str
     status: EntryStatus
 
+    @property
+    def definitions_list(self) -> list[str]:
+        entry_starts = ["1)", "2)", "3)", "4)"]
+
+        # We only want to process "intact" looking definitions.
+        if "1) " not in self.definitions:
+            return [self.definitions]
+
+        lines = []
+        line = ""
+
+        words = self.definitions.split(" ")
+
+        # Walk through words, parsing new lines with new numbered indexes.
+        for word in words:
+            if word in entry_starts:
+                lines.append(line.strip())
+                line = ""
+
+            line = f"{line} {word}"
+
+        lines.append(line.strip())
+
+        # Sanity: we want each line to be properly numbered with no jumps.
+        expected_number = 1
+
+        for line in lines:
+            first = line[0]
+            if first.isnumeric():
+                if int(first) != expected_number:
+                    return [self.definitions]
+
+                expected_number += 1
+
+        return lines
+
     def to_json(self) -> dict[str, str]:
         return {
             "headword": self.headword,

parser/tests/test_entry.py

@@ -0,0 +1,62 @@
+from src.parser.entry import Entry, EntryStatus
+
+
+def test_parses_single_definition_as_list() -> None:
+    entry = Entry(headword="Foo", definitions="Foo bar baz", status=EntryStatus.VALID)
+
+    expected_definitions = ["Foo bar baz"]
+
+    assert entry.definitions_list == expected_definitions
+
+
+def test_parses_single_numbered_definition_as_list() -> None:
+    entry = Entry(
+        headword="Bable",
+        definitions="go. 1) at tale uforståe- ligt. Moth;",
+        status=EntryStatus.VALID,
+    )
+
+    expected_definitions = ["go.", "1) at tale uforståe- ligt. Moth;"]
+
+    assert entry.definitions_list == expected_definitions
+
+
+def test_parses_multiple_definitions_as_list() -> None:
+    entry = Entry(
+        headword="Afkom",
+        definitions=(
+            "go. 1) komme (bort) fra; kommer schiff paa grund oc kand ey med mindre affkomme. "
+            "N. D. Mag. VI. 104; som ey bygge oc boo paa wort oc kronens goilz oc ey ere aff- "
+            "komne met wore mynde (1506). Ro- senv., Gl. L. V. 202; po thet Hans Ure motte thets "
+            "bedre affkomme oc guit bliffae trette oc uenighed (1549). Rosenv., Gl. D. I. 71; N. D. Mag. 1. 814. "
+            "— 2) komme af; at ther will afikomme eth stort oprør och forderflue (1526). N. D. Mag. V. 215; der met "
+            "er det affkommen, mand neppe kiender slecten. Hvitf. VIII 365. — 3) aflægges (t. abkommen.); at the ismaa "
+            "markede ere aflagde, oc att ingen haffaer fordelle ther af, at the ere afkommen (1542). D. Mag. IV. 288. — "
+            "4) overkomme; at voris depu- terede samme miinstringer, naar de afkomme kand, self skall bjwaane (1890). "
+            "Geh. Ark. Årsb. IL 294."
+        ),
+        status=EntryStatus.VALID,
+    )
+
+    expected_definitions = [
+        "go.",
+        "1) komme (bort) fra; kommer schiff paa grund oc kand ey med mindre affkomme. N. D. Mag. VI. 104; som ey bygge oc boo paa wort oc kronens goilz oc ey ere aff- komne met wore mynde (1506). Ro- senv., Gl. L. V. 202; po thet Hans Ure motte thets bedre affkomme oc guit bliffae trette oc uenighed (1549). Rosenv., Gl. D. I. 71; N. D. Mag. 1. 814. —",  # noqa: E501
+        "2) komme af; at ther will afikomme eth stort oprør och forderflue (1526). N. D. Mag. V. 215; der met er det affkommen, mand neppe kiender slecten. Hvitf. VIII 365. —",  # noqa: E501
+        "3) aflægges (t. abkommen.); at the ismaa markede ere aflagde, oc att ingen haffaer fordelle ther af, at the ere afkommen (1542). D. Mag. IV. 288. —",  # noqa: E501
+        "4) overkomme; at voris depu- terede samme miinstringer, naar de afkomme kand, self skall bjwaane (1890). Geh. Ark. Årsb. IL 294.",  # noqa: E501
+    ]
+
+    assert entry.definitions_list == expected_definitions
+
+
+def test_parses_incorrectly_numbered_definitions_list() -> None:
+    entry = Entry(
+        headword="Foo",
+        definitions="Foo bar baz 1) bar bar 3) baz baz 2) foo foo ",
+        status=EntryStatus.VALID,
+    )
+
+    # Should not split if numbering does not make sense.
+    expected_definitions = ["Foo bar baz 1) bar bar 3) baz baz 2) foo foo "]
+
+    assert entry.definitions_list == expected_definitions