Parse TAT text and gaps

tetras_extraction/script/src/extract_page.py

@@ -183,6 +183,14 @@ class ChoiceGroup:
         self.items: list[Choice]
 
 
+class Gap:
+    """A gap in a gap-fill text exercise"""
+
+    def __init__(self, id: str):
+        self.id = id
+        self.choices: list[Choice] = []
+
+
 class ExerciceQC(Exercice):
     def __init__(self, is_qcm: bool = False) -> None:
         super().__init__()
@@ -275,8 +283,41 @@ class ExerciceQM(Exercice):
 class ExerciceTAT(Exercice):
     def __init__(self):
         super().__init__()
-        self.text: str  # can be HTML
-        self.gaps: list[ChoiceGroup]
+        self.segments: list[str | Gap] = []
+
+    @override
+    def parse_html(self, root: HtmlElement):
+        super().parse_html(root)
+        # Find the text container
+        try:
+            container = root.find_class("STY_texteTAT")[0]
+        except IndexError as e:
+            raise ParseError("ExerciceTAT: text container not found") from e
+
+        # Text buffer accumulates the text found
+        text_segment_buf = container.text or ""
+        for elem in container:
+            if elem.tag == "select" and "STY_selectTAT" in elem.classes:
+                # It's a gap
+                # Time to "close" the text segment and add it
+                self.segments.append(text_segment_buf)
+                # Add the gap
+                gap_id = elem.attrib["id"].replace("champTrou", "")
+                self.segments.append(Gap(gap_id))
+                # New text segment starts with the tail text of this element
+                text_segment_buf = elem.tail or ""
+            else:
+                text_segment_buf += to_html(elem)
+
+        self.segments.append(text_segment_buf)
+
+        nb_total_gaps = len(container.find_class("STY_selectTAT"))
+        nb_found_gaps = len([e for e in self.segments if isinstance(e, Gap)])
+        if nb_found_gaps != nb_total_gaps:
+            log.warning(
+                f"{self.id}: Text has {nb_total_gaps} gaps in total, but found {nb_found_gaps} gap elements, some might be missing"
+            )
+        pass
 
 
 class ExerciceGD(Exercice):