Fix markdown output

marker/v2/providers/pdf.py

@@ -35,7 +35,10 @@ def __len__(self) -> int:
     def __del__(self):
         self.doc.close()
 
-    def font_flags_to_format(self, flags: int) -> Set[str]:
+    def font_flags_to_format(self, flags: int | None) -> Set[str]:
+        if flags is None:
+            return {"plain"}
+
         flag_map = {
             1: "FixedPitch",
             2: "Serif",
@@ -72,8 +75,11 @@ def font_flags_to_format(self, flags: int) -> Set[str]:
                 formats.add("plain")
         return formats
 
-    def font_names_to_format(self, font_name: str) -> Set[str]:
+    def font_names_to_format(self, font_name: str | None) -> Set[str]:
         formats = set()
+        if font_name is None:
+            return formats
+
         if "bold" in font_name.lower():
             formats.add("bold")
         if "ital" in font_name.lower():
@@ -97,16 +103,19 @@ def setup(self):
                 for line in block["lines"]:
                     spans: List[Span] = []
                     for span in line["spans"]:
-                        if not span["text"].strip():
+                        if not span["text"]:
                             continue
                         font_formats = self.font_flags_to_format(span["font"]["flags"]).union(self.font_names_to_format(span["font"]["name"]))
+                        font_name = span["font"]["name"] or "Unknown"
+                        font_weight = span["font"]["weight"] or 0
+                        font_size = span["font"]["size"] or 0
                         spans.append(
                             Span(
                                 polygon=PolygonBox.from_bbox(span["bbox"]),
-                                text=span["text"].strip(),
-                                font=span["font"]["name"],
-                                font_weight=span["font"]["weight"],
-                                font_size=span["font"]["size"],
+                                text=span["text"],
+                                font=font_name,
+                                font_weight=font_weight,
+                                font_size=font_size,
                                 minimum_position=span["char_start_idx"],
                                 maximum_position=span["char_end_idx"],
                                 formats=list(font_formats),

marker/v2/schema/blocks/base.py

@@ -100,21 +100,21 @@ def raw_text(self, document) -> str:
                 text += "\n"
         return text
 
-    def assemble_html(self, child_blocks):
+    def assemble_html(self, child_blocks, parent_structure=None):
         template = ""
         for c in child_blocks:
             template += f"<content-ref src='{c.id}'></content-ref>"
         return template
 
-    def render(self, document):
+    def render(self, document, parent_structure):
         child_content = []
         if self.structure is not None and len(self.structure) > 0:
             for block_id in self.structure:
                 block = document.get_block(block_id)
-                child_content.append(block.render(document))
+                child_content.append(block.render(document, self.structure))
 
         return BlockOutput(
-            html=self.assemble_html(child_content),
+            html=self.assemble_html(child_content, parent_structure),
             polygon=self.polygon,
             id=self.id,
             children=child_content

marker/v2/schema/blocks/listitem.py

@@ -4,6 +4,6 @@
 class ListItem(Block):
     block_type: str = "ListItem"
 
-    def assemble_html(self, child_blocks):
-        template = super().assemble_html(child_blocks)
+    def assemble_html(self, child_blocks, parent_structure):
+        template = super().assemble_html(child_blocks, parent_structure)
         return f"<li>{template}</li>"

marker/v2/schema/blocks/text.py

@@ -4,6 +4,7 @@
 class Text(Block):
     block_type: str = "Text"
 
-    def assemble_html(self, child_blocks):
-        template = super().assemble_html(child_blocks)
+    def assemble_html(self, child_blocks, parent_structure):
+        template = super().assemble_html(child_blocks, parent_structure)
+        template = template.replace("\n", " ")
         return f"<p>{template}</p>"

marker/v2/schema/document.py

@@ -35,7 +35,7 @@ def assemble_html(self, child_blocks):
     def render(self):
         child_content = []
         for page in self.pages:
-            child_content.append(page.render(self))
+            child_content.append(page.render(self, None))
 
         return DocumentOutput(
             children=child_content,

marker/v2/schema/groups/list.py

@@ -4,6 +4,6 @@
 class ListGroup(Block):
     block_type: str = "ListGroup"
 
-    def assemble_html(self, child_blocks):
-        template = super().assemble_html(child_blocks)
+    def assemble_html(self, child_blocks, parent_structure):
+        template = super().assemble_html(child_blocks, parent_structure)
         return f"<ul>{template}</ul>"

marker/v2/schema/text/line.py

@@ -1,24 +1,60 @@
+import re
+
+import regex
+
 from marker.v2.schema.blocks import Block, BlockOutput
 
+HYPHENS = r'-—¬'
+
+
+def remove_tags(text):
+    return re.sub(r'<[^>]+>', '', text)
+
+
+def replace_last(string, old, new):
+    matches = list(re.finditer(old, string))
+    if not matches:
+        return string
+    last_match = matches[-1]
+    return string[:last_match.start()] + new + string[last_match.end():]
+
+def strip_trailing_hyphens(line_text, next_line_text, line_html) -> str:
+    lowercase_letters = r'\p{Ll}|\d'
+
+    hyphen_regex = regex.compile(rf'.*[{HYPHENS}]\s?$', regex.DOTALL)
+    next_line_starts_lowercase = regex.match(rf"^\s?[{lowercase_letters}]", next_line_text)
+
+    if hyphen_regex.match(line_text) and next_line_starts_lowercase:
+        return replace_last(line_html, rf'[{HYPHENS}]', "")
+    return line_html
+
 
 class Line(Block):
     block_type: str = "Line"
 
-    def assemble_html(self, child_blocks):
+    def assemble_html(self, document, child_blocks, parent_structure):
         template = ""
         for c in child_blocks:
             template += c.html
+
+        raw_text = remove_tags(template).strip()
+        structure_idx = parent_structure.index(self.id)
+        if structure_idx < len(parent_structure) - 1:
+            next_block_id = parent_structure[structure_idx + 1]
+            next_line = document.get_block(next_block_id)
+            next_line_raw_text = next_line.raw_text(document)
+            template = strip_trailing_hyphens(raw_text, next_line_raw_text, template)
         return template
 
-    def render(self, document):
+    def render(self, document, parent_structure):
         child_content = []
         if self.structure is not None and len(self.structure) > 0:
             for block_id in self.structure:
                 block = document.get_block(block_id)
-                child_content.append(block.render(document))
+                child_content.append(block.render(document, parent_structure))
 
         return BlockOutput(
-            html=self.assemble_html(child_content),
+            html=self.assemble_html(document, child_content, parent_structure),
             polygon=self.polygon,
             id=self.id,
             children=[]

marker/v2/schema/text/span.py

@@ -22,10 +22,21 @@ def bold(self):
     def italic(self):
         return 'italic' in self.formats
 
-    def assemble_html(self, child_blocks):
-        if len(self.text) > 3:
+    def assemble_html(self, child_blocks, parent_structure):
+        text = self.text
+        text = text.replace("-\n", "")  # Remove hyphenated line breaks
+
+        replaced_newline = False
+        while len(text) > 0 and text[-1] in ["\n", "\r"]:
+            text = text[:-1]
+            replaced_newline = True
+
+        if replaced_newline:
+            text += " "
+
+        if len(text) > 3:
             if self.italic:
-                return f"<i>{self.text}</i>"
+                return f"<i>{text}</i>"
             elif self.bold:
-                return f"<b>{self.text}</b>"
-        return self.text
+                return f"<b>{text}</b>"
+        return text