✨ feat(il_creater): improve font descent handling for character bound…

…ing boxes

- Add descent calculation for both vertical and horizontal text orientations
- Adjust character bounding boxes by incorporating font descent metrics
- Enhance font rendering accuracy by correctly applying descent offsets

yadt/document_il/frontend/il_creater.py

@@ -27,7 +27,8 @@ def __init__(self, translation_config: TranslationConfig):
         self.non_stroking_color_space_name = None
         self.passthrough_per_char_instruction: list[tuple[str, str]] = []
         self.translation_config = translation_config
-        self.passthrough_per_char_instruction_stack: list[list[tuple[str, str]]] = []
+        self.passthrough_per_char_instruction_stack: list[list[tuple[str, str]]] = [
+        ]
         self.xobj_id = 0
         self.xobj_inc = 0
         self.xobj_map: dict[int, il_version_1.PdfXobject] = {}
@@ -41,7 +42,8 @@ def is_passthrough_per_char_operation(self, operator: str):
 
     def on_passthrough_per_char(self, operator: str, args: list[str]):
         if not self.is_passthrough_per_char_operation(operator):
-            logger.error("Unknown passthrough_per_char operation: %s", operator)
+            logger.error(
+                "Unknown passthrough_per_char operation: %s", operator)
             return
         # logger.debug("xobj_id: %d, on_passthrough_per_char: %s ( %s )", self.xobj_id, operator, args)
         args = [self.parse_arg(arg) for arg in args]
@@ -50,7 +52,8 @@ def on_passthrough_per_char(self, operator: str, args: list[str]):
             if op == operator:
                 self.passthrough_per_char_instruction.remove(value)
                 break
-        self.passthrough_per_char_instruction.append((operator, " ".join(args)))
+        self.passthrough_per_char_instruction.append(
+            (operator, " ".join(args)))
         pass
 
     def remove_latest_passthrough_per_char_instruction(self):
@@ -154,7 +157,8 @@ def on_page_crop_box(
         x1: float | int,
         y1: float | int,
     ):
-        box = il_version_1.Box(x=float(x0), y=float(y0), x2=float(x1), y2=float(y1))
+        box = il_version_1.Box(x=float(x0), y=float(y0),
+                               x2=float(x1), y2=float(y1))
         self.current_page.cropbox = il_version_1.Cropbox(box=box)
 
     def on_page_media_box(
@@ -164,7 +168,8 @@ def on_page_media_box(
         x1: float | int,
         y1: float | int,
     ):
-        box = il_version_1.Box(x=float(x0), y=float(y0), x2=float(x1), y2=float(y1))
+        box = il_version_1.Box(x=float(x0), y=float(y0),
+                               x2=float(x1), y2=float(y1))
         self.current_page.mediabox = il_version_1.Mediabox(box=box)
 
     def on_page_number(self, page_number: int):
@@ -173,15 +178,17 @@ def on_page_number(self, page_number: int):
         self.current_page.page_number = page_number
 
     def on_page_base_operation(self, operation: str):
-        self.current_page.base_operations = il_version_1.BaseOperations(value=operation)
+        self.current_page.base_operations = il_version_1.BaseOperations(
+            value=operation)
 
     def on_page_resource_font(self, font: PDFFont, xref_id: int, font_id: str):
         font_name = font.fontname
         if isinstance(font_name, bytes):
             try:
                 font_name = font_name.decode("utf-8")
             except UnicodeDecodeError:
-                font_name = "BASE64:" + base64.b64encode(font_name).decode("utf-8")
+                font_name = "BASE64:" + \
+                    base64.b64encode(font_name).decode("utf-8")
         encoding_length = 1
         if isinstance(font, PDFCIDFont):
             try:
@@ -193,7 +200,8 @@ def on_page_resource_font(self, font: PDFFont, xref_id: int, font_id: str):
                 elif encoding == "/Identity-V":
                     encoding_length = 2
                 else:
-                    _, to_unicode_id = self.mupdf.xref_get_key(xref_id, "ToUnicode")
+                    _, to_unicode_id = self.mupdf.xref_get_key(
+                        xref_id, "ToUnicode")
                     to_unicode_bytes = self.mupdf.xref_stream(
                         int(to_unicode_id.split(" ")[0])
                     )
@@ -207,7 +215,8 @@ def on_page_resource_font(self, font: PDFFont, xref_id: int, font_id: str):
                 else:
                     encoding_length = 1
         try:
-            mupdf_font = pymupdf.Font(fontbuffer=self.mupdf.extract_font(xref_id)[3])
+            mupdf_font = pymupdf.Font(
+                fontbuffer=self.mupdf.extract_font(xref_id)[3])
             bold = mupdf_font.is_bold
             italic = mupdf_font.is_italic
             monospaced = mupdf_font.is_monospaced
@@ -264,7 +273,19 @@ def create_graphic_state(self, gs: pdfminer.pdfinterp.PDFGraphicState):
 
     def on_lt_char(self, char: LTChar):
         gs = self.create_graphic_state(char.graphicstate)
-        bbox = il_version_1.Box(char.bbox[0], char.bbox[1], char.bbox[2], char.bbox[3])
+        # Get font from current page or xobject
+        font = None
+        for pdf_font in (self.xobj_map.get(self.xobj_id,
+                                           self.current_page).pdf_font):
+            if pdf_font.font_id == char.aw_font_id:
+                font = pdf_font
+                break
+
+        # Get descent from font
+        descent = 0
+        if font and hasattr(font, 'descent'):
+            descent = font.descent * char.size / 1000
+
 
         char_id = char.cid
         char_unicode = char.get_text()
@@ -273,8 +294,21 @@ def on_lt_char(self, char: LTChar):
         advance = char.adv
         if char.matrix[0] == 0 and char.matrix[3] == 0:
             vertical = True
+            bbox = il_version_1.Box(
+                x=char.bbox[0] - descent,
+                y=char.bbox[1],
+                x2=char.bbox[2] - descent,
+                y2=char.bbox[3]
+            )
         else:
             vertical = False
+            # Add descent to y coordinates
+            bbox = il_version_1.Box(
+                x=char.bbox[0],
+                y=char.bbox[1] + descent,
+                x2=char.bbox[2],
+                y2=char.bbox[3] + descent
+            )
         pdf_style = il_version_1.PdfStyle(
             font_id=char.aw_font_id,
             font_size=char.size,