Merge pull request #336 from UnravelSports/fix/tracab_meta

Fix for older/incomplete/other versions of Tracab Meta data files

kloppy/infra/serializers/tracking/tracab/tracab_dat.py

@@ -72,15 +72,18 @@ def _frame_from_line(cls, teams, period, line, frame_rate):
                 )
 
             player = team.get_player_by_jersey_number(jersey_no)
-            if player:
-                players_data[player] = PlayerData(
-                    coordinates=Point(float(x), float(y)), speed=float(speed)
-                )
-            else:
-                # continue
-                raise DeserializationError(
-                    f"Player not found for player jersey no {jersey_no} of team: {team.name}"
+
+            if not player:
+                player = Player(
+                    player_id=f"{team.ground}_{jersey_no}",
+                    team=team,
+                    jersey_no=int(jersey_no),
                 )
+                team.players.append(player)
+
+            players_data[player] = PlayerData(
+                coordinates=Point(float(x), float(y)), speed=float(speed)
+            )
 
         (
             ball_x,
@@ -160,8 +163,12 @@ def deserialize(self, inputs: TRACABInputs) -> TrackingDataset:
             meta_data = objectify.fromstring(inputs.meta_data.read())
             match = meta_data.match
             frame_rate = int(match.attrib["iFrameRateFps"])
-            pitch_size_width = float(match.attrib["fPitchXSizeMeters"])
-            pitch_size_height = float(match.attrib["fPitchYSizeMeters"])
+            pitch_size_width = float(
+                match.attrib["fPitchXSizeMeters"].replace(",", ".")
+            )
+            pitch_size_height = float(
+                match.attrib["fPitchYSizeMeters"].replace(",", ".")
+            )
 
             periods = []
             for period in match.iterchildren(tag="period"):
@@ -180,12 +187,22 @@ def deserialize(self, inputs: TRACABInputs) -> TrackingDataset:
                         )
                     )
 
-            home_team = self.create_team(
-                meta_data["HomeTeam"], Ground.HOME, start_frame_id
-            )
-            away_team = self.create_team(
-                meta_data["AwayTeam"], Ground.AWAY, start_frame_id
-            )
+            if hasattr(meta_data, "HomeTeam") and hasattr(
+                meta_data, "AwayTeam"
+            ):
+                home_team = self.create_team(
+                    meta_data["HomeTeam"], Ground.HOME, start_frame_id
+                )
+                away_team = self.create_team(
+                    meta_data["AwayTeam"], Ground.AWAY, start_frame_id
+                )
+            else:
+                home_team = Team(
+                    team_id="home", name="home", ground=Ground.HOME
+                )
+                away_team = Team(
+                    team_id="away", name="away", ground=Ground.AWAY
+                )
             teams = [home_team, away_team]
 
         with performance_logging("Loading data", logger=logger):

kloppy/tests/files/tracab_meta_2.xml

@@ -0,0 +1,16 @@
+<?xml version="1.0"?>
+<TracabMetaData sVersion="1.0">
+    <match 
+        iId="1"
+        dtDate="2023-12-15 20:32:20"
+        iFrameRateFps="25"
+        fPitchXSizeMeters="105.00"
+        fPitchYSizeMeters="68.00"
+        fTrackingAreaXSizeMeters="unknown"
+        fTrackingAreaYSizeMeters="unknown">
+    <period iId="1" iStartFrame="1848508" iEndFrame="1916408"/>
+    <period iId="2" iStartFrame="1942114" iEndFrame="2017933"/>
+    <period iId="3" iStartFrame="0" iEndFrame="0"/>
+    <period iId="4" iStartFrame="0" iEndFrame="0"/>
+</match>
+</TracabMetaData>

kloppy/tests/files/tracab_meta_3.xml

@@ -0,0 +1,12 @@
+<OptaMetaData sVersion="1.2">
+    <match 
+        iId="1"
+        dtDate="2023-12-15 20:32:20"
+        iFrameRateFps="25"
+        fPitchXSizeMeters="105,00"
+        fPitchYSizeMeters="68,00"
+    >
+    <period iId="1" iStartFrame="1848508" iEndFrame="1916408"/>
+    <period iId="2" iStartFrame="1942114" iEndFrame="2017933"/>
+</match>
+</OptaMetaData>

kloppy/tests/test_tracab.py

@@ -39,6 +39,16 @@ def xml_meta_data(base_dir: Path) -> Path:
     return base_dir / "files" / "tracab_meta.xml"
 
 
+@pytest.fixture(scope="session")
+def xml_meta2_data(base_dir: Path) -> Path:
+    return base_dir / "files" / "tracab_meta_2.xml"
+
+
+@pytest.fixture(scope="session")
+def xml_meta3_data(base_dir: Path) -> Path:
+    return base_dir / "files" / "tracab_meta_3.xml"
+
+
 @pytest.fixture(scope="session")
 def dat_raw_data(base_dir: Path) -> Path:
     return base_dir / "files" / "tracab_raw.dat"
@@ -223,3 +233,125 @@ def test_correct_normalized_deserialization(
         assert dataset.records[0].players_data[
             player_home_1
         ].coordinates == Point(x=1.0019047619047619, y=0.49602941176470583)
+
+
+class TestTracabMeta2:
+    def test_correct_deserialization(
+        self, xml_meta2_data: Path, dat_raw_data: Path
+    ):
+        dataset = tracab.load(
+            meta_data=xml_meta2_data,
+            raw_data=dat_raw_data,
+            coordinates="tracab",
+            only_alive=False,
+        )
+
+        # Check metadata
+        assert dataset.metadata.provider == Provider.TRACAB
+        assert dataset.dataset_type == DatasetType.TRACKING
+        assert len(dataset.records) == 7
+        assert len(dataset.metadata.periods) == 2
+        assert dataset.metadata.orientation == Orientation.AWAY_HOME
+        assert dataset.metadata.periods[0].id == 1
+        assert dataset.metadata.periods[0].start_timestamp == timedelta(
+            seconds=73940, microseconds=320000
+        )
+        assert dataset.metadata.periods[0].end_timestamp == timedelta(
+            seconds=76656, microseconds=320000
+        )
+        assert dataset.metadata.periods[1].id == 2
+        assert dataset.metadata.periods[1].start_timestamp == timedelta(
+            seconds=77684, microseconds=560000
+        )
+        assert dataset.metadata.periods[1].end_timestamp == timedelta(
+            seconds=80717, microseconds=320000
+        )
+
+        # No need to check frames, since we do that in TestTracabDATTracking
+        # The only difference in this test is the meta data file structure
+
+        # make sure player data is only in the frame when the player is at the pitch
+        assert "home_20" in [
+            player.player_id
+            for player in dataset.records[0].players_data.keys()
+        ]
+        assert "home_20" not in [
+            player.player_id
+            for player in dataset.records[6].players_data.keys()
+        ]
+
+    def test_correct_normalized_deserialization(
+        self, xml_meta2_data: Path, dat_raw_data: Path
+    ):
+        dataset = tracab.load(
+            meta_data=xml_meta2_data, raw_data=dat_raw_data, only_alive=False
+        )
+
+        player_home_1 = dataset.metadata.teams[0].get_player_by_jersey_number(
+            1
+        )
+
+        assert dataset.records[0].players_data[
+            player_home_1
+        ].coordinates == Point(x=1.0019047619047619, y=0.49602941176470583)
+
+
+class TestTracabMeta3:
+    def test_correct_deserialization(
+        self, xml_meta3_data: Path, dat_raw_data: Path
+    ):
+        dataset = tracab.load(
+            meta_data=xml_meta3_data,
+            raw_data=dat_raw_data,
+            coordinates="tracab",
+            only_alive=False,
+        )
+
+        # Check metadata
+        assert dataset.metadata.provider == Provider.TRACAB
+        assert dataset.dataset_type == DatasetType.TRACKING
+        assert len(dataset.records) == 7
+        assert len(dataset.metadata.periods) == 2
+        assert dataset.metadata.orientation == Orientation.AWAY_HOME
+        assert dataset.metadata.periods[0].id == 1
+        assert dataset.metadata.periods[0].start_timestamp == timedelta(
+            seconds=73940, microseconds=320000
+        )
+        assert dataset.metadata.periods[0].end_timestamp == timedelta(
+            seconds=76656, microseconds=320000
+        )
+        assert dataset.metadata.periods[1].id == 2
+        assert dataset.metadata.periods[1].start_timestamp == timedelta(
+            seconds=77684, microseconds=560000
+        )
+        assert dataset.metadata.periods[1].end_timestamp == timedelta(
+            seconds=80717, microseconds=320000
+        )
+
+        # No need to check frames, since we do that in TestTracabDATTracking
+        # The only difference in this test is the meta data file structure
+
+        # make sure player data is only in the frame when the player is at the pitch
+        assert "home_20" in [
+            player.player_id
+            for player in dataset.records[0].players_data.keys()
+        ]
+        assert "home_20" not in [
+            player.player_id
+            for player in dataset.records[6].players_data.keys()
+        ]
+
+    def test_correct_normalized_deserialization(
+        self, xml_meta3_data: Path, dat_raw_data: Path
+    ):
+        dataset = tracab.load(
+            meta_data=xml_meta3_data, raw_data=dat_raw_data, only_alive=False
+        )
+
+        player_home_1 = dataset.metadata.teams[0].get_player_by_jersey_number(
+            1
+        )
+
+        assert dataset.records[0].players_data[
+            player_home_1
+        ].coordinates == Point(x=1.0019047619047619, y=0.49602941176470583)