Merge pull request #88 from swisstopo/LGVISIUM-83-Extract-coordinates…

…-with-non-integer-values

Close #LGVISIUM-83: Extract coordinates with non integer values

src/stratigraphy/annotations/draw.py

@@ -138,10 +138,10 @@ def draw_metadata(
     """
     # TODO associate correctness with the extracted coordinates in a better way
     coordinate_color = "green" if is_coordinate_correct else "red"
-    coordinate_rect = fitz.Rect([5, 5, 200, 25])
+    coordinate_rect = fitz.Rect([5, 5, 250, 30])
 
     elevation_color = "green" if is_elevation_correct else "red"
-    elevation_rect = fitz.Rect([5, 25, 200, 45])
+    elevation_rect = fitz.Rect([5, 30, 250, 55])
 
     shape.draw_rect(coordinate_rect * derotation_matrix)
     shape.finish(fill=fitz.utils.getColor("gray"), fill_opacity=0.5)

src/stratigraphy/metadata/coordinate_extraction.py

@@ -14,14 +14,14 @@
 
 logger = logging.getLogger(__name__)
 
-COORDINATE_ENTRY_REGEX = r"(?:([12])[\.\s'‘’]{0,2})?(\d{3})[\.\s'‘’]{0,2}(\d{3})\.?\d?"
+COORDINATE_ENTRY_REGEX = r"(?:([12])[\.\s'‘’]{0,2})?(\d{3})[\.\s'‘’]{0,2}(\d{3})(?:\.(\d{1,}))?"
 
 
 @dataclass(kw_only=True)
 class CoordinateEntry:
     """Dataclass to represent a coordinate entry."""
 
-    coordinate_value: int
+    coordinate_value: float
 
     def __repr__(self):
         if self.coordinate_value > 1e5:
@@ -44,7 +44,7 @@ def __post_init__(self):
             self.north, self.east = self.east, self.north
 
     def __str__(self):
-        return f"E: {self.east}, N: {self.north}"
+        return f"E: {self.east.coordinate_value}, N: {self.north.coordinate_value}"
 
     def to_json(self) -> dict:
         """Converts the object to a dictionary.
@@ -64,7 +64,18 @@ def is_valid(self):
         pass
 
     @staticmethod
-    def from_values(east: int, north: int, rect: fitz.Rect, page: int) -> Coordinate | None:
+    def from_values(east: float, north: float, rect: fitz.Rect, page: int) -> Coordinate | None:
+        """Creates a Coordinate object from the given values.
+
+        Args:
+            east (float): The east coordinate value.
+            north (float): The north coordinate value.
+            rect (fitz.Rect): The rectangle that contains the extracted information.
+            page (int): The page number of the PDF document.
+
+        Returns:
+            Coordinate | None: The coordinate object.
+        """
         if 1e6 < east < 1e7:
             return LV95Coordinate(
                 east=CoordinateEntry(coordinate_value=east),
@@ -104,7 +115,9 @@ class LV95Coordinate(Coordinate):
 
     def is_valid(self):
         """Reference: https://de.wikipedia.org/wiki/Schweizer_Landeskoordinaten#Beispielkoordinaten."""
-        return 2324800 < self.east.coordinate_value < 2847500 and 1074000 < self.north.coordinate_value < 1302000
+        return (
+            2324800.0 < self.east.coordinate_value < 2847500.0 and 1074000.0 < self.north.coordinate_value < 1302000.0
+        )
 
 
 @dataclass
@@ -116,7 +129,7 @@ def is_valid(self):
 
         To account for uncertainties in the conversion of LV03 to LV95, we allow a margin of 2.
         """
-        return 324798 < self.east.coordinate_value < 847502 and 73998 < self.north.coordinate_value < 302002
+        return 324798.0 < self.east.coordinate_value < 847502.0 and 73998.0 < self.north.coordinate_value < 302002.0
 
 
 class CoordinateExtractor(DataExtractor):
@@ -224,12 +237,16 @@ def get_coordinates_from_lines(self, lines: list[TextLine], page: int) -> list[C
             list[Coordinate]: A list of potential coordinates
         """
         full_regex = regex.compile(
-            r"(?:[XY][=:\s]{0,2})?" + COORDINATE_ENTRY_REGEX + r".{0,4}?[XY]?[=:\s]{0,2}" + COORDINATE_ENTRY_REGEX
+            r"(?:[XY][=:\s]{0,2})?"
+            + COORDINATE_ENTRY_REGEX
+            + r".{0,4}?[XY]?[=:\s]{0,2}"
+            + COORDINATE_ENTRY_REGEX
+            + r"\b"
         )
         potential_coordinates = [
             Coordinate.from_values(
-                east=int("".join(match.groups(default="")[:3])),
-                north=int("".join(match.groups(default="")[3:])),
+                east=float("{}.{}".format("".join(match.groups(default="")[:3]), match.groups(default="")[3])),
+                north=float("{}.{}".format("".join(match.groups(default="")[4:-1]), match.groups(default="")[-1])),
                 rect=rect,
                 page=page,
             )

tests/test_coordinate_extraction.py

@@ -21,7 +21,7 @@ def test_strLV95():  # noqa: D103
         rect=fitz.Rect(),
         page=1,
     )
-    assert str(coord) == "E: 2'789'456, N: 1'123'012"
+    assert str(coord) == "E: 2789456, N: 1123012"
 
 
 def test_to_jsonLV95():  # noqa: D103
@@ -52,7 +52,7 @@ def test_strLV03():  # noqa: D103
         rect=fitz.Rect(),
         page=1,
     )
-    assert str(coord) == "E: 789'456, N: 123'012"
+    assert str(coord) == "E: 789456, N: 123012"
 
 
 def test_to_jsonLV03():  # noqa: D103
@@ -67,6 +67,7 @@ def test_to_jsonLV03():  # noqa: D103
 
 
 doc = fitz.open(DATAPATH.parent / "example" / "example_borehole_profile.pdf")
+doc_with_digits_in_coordinates = fitz.open(DATAPATH.parent / "example" / "A7367.pdf")
 extractor = CoordinateExtractor(doc)
 
 
@@ -76,8 +77,18 @@ def test_CoordinateExtractor_extract_coordinates():  # noqa: D103
     coordinates = extractor.extract_coordinates()
     # Check if the returned value is a list
     assert isinstance(coordinates, Coordinate)
-    assert repr(coordinates.east) == "615'790"
-    assert repr(coordinates.north) == "157'500"
+    assert repr(coordinates.east) == "615'790.0"
+    assert repr(coordinates.north) == "157'500.0"
+
+
+def test_CoordinateExtractor_extract_coordinates_with_digits_in_coordinates():  # noqa: D103
+    """Test the extraction of coordinates from a PDF document with digits in the coordinates."""
+    # Assuming there is a method called 'extract' in CoordinateExtractor class
+    coordinates = CoordinateExtractor(doc_with_digits_in_coordinates).extract_coordinates()
+    # Check if the returned value is a list
+    assert isinstance(coordinates, Coordinate)
+    assert repr(coordinates.east) == "607'562.0"
+    assert repr(coordinates.north) == "187'087.5"
 
 
 def _create_simple_lines(text_lines: list[str]) -> list[TextLine]:
@@ -211,3 +222,35 @@ def test_CoordinateExtractor_get_coordinates_from_lines_rect():  # noqa: D103
     expected_rect.include_rect(lines[2].rect)
     assert coordinates[0].rect == expected_rect
     assert coordinates[0].page == 1
+
+    # Example from 269126143-bp.pdf (a slash in the middle of the coordinates as misread by OCR as the digit 1)
+    lines = _create_simple_lines(["269578211260032"])
+    coordinates = extractor.get_coordinates_from_lines(lines, page=1)
+    assert coordinates[0].east.coordinate_value == 2695782
+    assert coordinates[0].north.coordinate_value == 1260032
+
+
+def test_get_single_decimal_coordinates():
+    """Test the extraction of decimal coordinates from a list of text lines."""
+    lines = _create_simple_lines(["615.790.6 / 157.500.5"])
+    coordinates = extractor.get_coordinates_from_lines(lines, page=1)
+    assert coordinates[0].east.coordinate_value == 615790.6
+    assert coordinates[0].north.coordinate_value == 157500.5
+
+    lines = _create_simple_lines(["2600000.6 / 1200000.5"])
+    coordinates = extractor.get_coordinates_from_lines(lines, page=1)
+    assert coordinates[0].east.coordinate_value == 2600000.6
+    assert coordinates[0].north.coordinate_value == 1200000.5
+
+
+def test_get_double_decimal_coordinates():
+    """Test the extraction of decimal coordinates from a list of text lines."""
+    lines = _create_simple_lines(["615.790.64 / 157.500.55"])
+    coordinates = extractor.get_coordinates_from_lines(lines, page=1)
+    assert coordinates[0].east.coordinate_value == 615790.64
+    assert coordinates[0].north.coordinate_value == 157500.55
+
+    lines = _create_simple_lines(["2600000.64 / 1200000.55"])
+    coordinates = extractor.get_coordinates_from_lines(lines, page=1)
+    assert coordinates[0].east.coordinate_value == 2600000.64
+    assert coordinates[0].north.coordinate_value == 1200000.55