Skip text directly on a record or lap element instead of crashing

src/activity_parser/parse_tcx_gpx.py

@@ -92,7 +92,7 @@ def walk_fields(element: etree._Element) -> Iterator[tuple[FieldPath, str]]:
     """Yields (path, value) for every attribute and leaf element under ``element``.
 
     ``path`` matches the keys of the field tables in ``xml_fields``. XML comments are
-    skipped.
+    skipped. ``path`` is never empty.
     """
     yield from _walk(element, ())
 
@@ -116,7 +116,7 @@ def _walk(element: etree._Element, path: FieldPath) -> Iterator[tuple[FieldPath,
         yield path + ((namespace, "@" + localname),), cast(str, value)
 
     text = element.text
-    if text is not None and not text.isspace():
+    if path and text is not None and not text.isspace():
         yield path, text
     # "*" matches only true elements, so comments are skipped without special-casing.
     for child in element.iterchildren("*"):

tests/files/gpx/direct_text.gpx

@@ -0,0 +1,14 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!-- Text directly inside the trkpt element itself, not a child's text. Used to raise
+     IndexError; must be skipped without dropping the fields around it. -->
+<gpx creator="activity_parser tests" version="1.1"
+    xmlns="http://www.topografix.com/GPX/1/1">
+  <trk>
+    <trkseg>
+      <trkpt lat="37.0000" lon="-122.0000">stray-trkpt-text
+        <ele>10.0</ele>
+        <time>2026-01-05T08:00:00Z</time>
+      </trkpt>
+    </trkseg>
+  </trk>
+</gpx>

tests/files/tcx/direct_text.tcx

@@ -0,0 +1,21 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!-- Text directly inside the Lap and Trackpoint elements themselves, not a child's text.
+     Both used to raise IndexError; both must be skipped without dropping the fields
+     around them. -->
+<TrainingCenterDatabase xmlns="http://www.garmin.com/xmlschemas/TrainingCenterDatabase/v2">
+  <Activities>
+    <Activity Sport="Biking">
+      <Id>2026-01-05T08:00:00Z</Id>
+      <Lap StartTime="2026-01-05T08:00:00Z">stray-lap-text
+        <TotalTimeSeconds>1.0</TotalTimeSeconds>
+        <DistanceMeters>5.0</DistanceMeters>
+        <Track>
+          <Trackpoint>stray-trackpoint-text
+            <Time>2026-01-05T08:00:00Z</Time>
+            <AltitudeMeters>10.0</AltitudeMeters>
+          </Trackpoint>
+        </Track>
+      </Lap>
+    </Activity>
+  </Activities>
+</TrainingCenterDatabase>

tests/test_parse_tcx_gpx.py

@@ -31,6 +31,8 @@
 CADENCE_COLLISION_TCX = TCX_FILES / "cadence_collision.tcx"
 UNKNOWN_COLLISION_TCX = TCX_FILES / "unknown_collision.tcx"
 MIXED_CONTENT_TCX = TCX_FILES / "mixed_content.tcx"
+DIRECT_TEXT_TCX = TCX_FILES / "direct_text.tcx"
+DIRECT_TEXT_GPX = GPX_FILES / "direct_text.gpx"
 VENDOR_TYPE_ATTRIBUTE_TCX = TCX_FILES / "vendor_type_attribute.tcx"
 DUPLICATE_CADENCE_TCX = TCX_FILES / "duplicate_cadence.tcx"
 LAPS_ONLY_TCX = TCX_FILES / "laps_only.tcx"
@@ -389,6 +391,15 @@ def test_gpx_unknown_extension_kept_as_namespaced_string():
     assert column not in normalized.columns
 
 
+def test_gpx_direct_text_on_trkpt_is_skipped():
+    # Text directly inside trkpt itself used to raise IndexError.
+    records, _, _ = parse_gpx(DIRECT_TEXT_GPX)
+    assert records["latitude"].tolist() == pytest.approx([37.0000])
+    assert records["longitude"].tolist() == pytest.approx([-122.0000])
+    assert records["altitude"].tolist() == [10.0]
+    assert not records.isin(["stray-trkpt-text"]).any().any()
+
+
 def test_gpx_1_0_base_fields():
     # GPX 1.0 exposes course/speed directly; GPS-fix diagnostics aren't fitness data.
     records, _, _ = ActivityParser().parse(GPX10)
@@ -493,6 +504,16 @@ def test_tcx_mixed_content_keeps_both_text_and_child():
     assert records[f"{{{ns}}}Nested"].tolist() == ["should-not-be-lost"]
 
 
+def test_tcx_direct_text_on_lap_and_trackpoint_is_skipped():
+    # Text directly inside Lap/Trackpoint themselves used to raise IndexError.
+    records, laps, _ = parse_tcx(DIRECT_TEXT_TCX)
+    assert records["altitude"].tolist() == [10.0]
+    assert laps["total_elapsed_time"].tolist() == [1.0]
+    assert laps["total_distance"].tolist() == pytest.approx([0.005])
+    assert not records.isin(["stray-trackpoint-text"]).any().any()
+    assert not laps.isin(["stray-lap-text"]).any().any()
+
+
 def test_tcx_type_attribute_only_skipped_for_xsi_namespace():
     # Only genuine xsi:type is metadata; a same-named attribute elsewhere is real data.
     records, _, _ = parse_tcx(VENDOR_TYPE_ATTRIBUTE_TCX)