html-to-markdown 2.24.6 → 2.25.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0cc6e5553ec13a221908d7529f53e0caa8d190df56523e7dab5b5852d229c9ef
-  data.tar.gz: 9543fca98f17d95285d97e5421565d604ee6c3e29ff1007a9af1692af5742ffc
+  metadata.gz: deb434d4fa161dcd1df3b836eac2ce0133bac0124f88a890757042044501acf2
+  data.tar.gz: 76a70b3e1a1abafa465b0d076526d1b58d50f3a39453e406bc2d4b4c8b669ce6
 SHA512:
-  metadata.gz: 948a72a00eceee87691aa3a20965d8fdde6cefb21ad465b46faed0fbec148915cad1573d98689de75c9d183528d90f1a118969d1857940d886884948195f6b4b
-  data.tar.gz: 91425e85e5df7fc997e6b1a091103166da1270a55d700ecfdaa1c4273d7029f98d06501d593eac594b6fd1420fd55fd331b37a62e299dc330805c3c0e1fa031a
+  metadata.gz: 59404b56deed91b2f0c65bf358fd05b685042804b30a2ee92ebb5cdb220a6b89c9f3dbc62d5576ef7d818c6372c42c1f8a2cc8bfbd2b8db16c8ecdbcdd60d427
+  data.tar.gz: 30f12a84b9b2b32aa6a1c1c3499ff8777e22994c84a38a22d4127eb91274ba4c9923270c2e8583e56b6cf758202581c58b38315a38fbd8c28ec737251ae32288

data/Gemfile.lock

@@ -1,7 +1,7 @@
 PATH
   remote: .
   specs:
-    html-to-markdown (2.24.6)
+    html-to-markdown (2.25.1)
       rb_sys (>= 0.9, < 1.0)
 
 GEM
@@ -161,7 +161,7 @@ CHECKSUMS
   ffi (1.17.3-x86_64-darwin) sha256=1f211811eb5cfaa25998322cdd92ab104bfbd26d1c4c08471599c511f2c00bb5
   ffi (1.17.3-x86_64-linux-gnu) sha256=3746b01f677aae7b16dc1acb7cb3cc17b3e35bdae7676a3f568153fb0e2c887f
   fileutils (1.8.0) sha256=8c6b1df54e2540bdb2f39258f08af78853aa70bad52b4d394bbc6424593c6e02
-  html-to-markdown (2.24.6)
+  html-to-markdown (2.25.1)
   i18n (1.14.8) sha256=285778639134865c5e0f6269e0b818256017e8cde89993fdfcbfb64d088824a5
   json (2.18.1) sha256=fe112755501b8d0466b5ada6cf50c8c3f41e897fa128ac5d263ec09eedc9f986
   language_server-protocol (3.17.0.5) sha256=fd1e39a51a28bf3eec959379985a72e296e9f9acfce46f6a79d31ca8760803cc

data/README.md

@@ -18,7 +18,7 @@
     <img src="https://img.shields.io/maven-central/v/dev.kreuzberg/html-to-markdown?label=Java&color=007ec6" alt="Java">
   </a>
   <a href="https://pkg.go.dev/github.com/kreuzberg-dev/html-to-markdown/packages/go/v2/htmltomarkdown">
-    <img src="https://img.shields.io/badge/Go-v2.24.6-007ec6" alt="Go">
+    <img src="https://img.shields.io/badge/Go-v2.25.1-007ec6" alt="Go">
   </a>
   <a href="https://www.nuget.org/packages/KreuzbergDev.HtmlToMarkdown/">
     <img src="https://img.shields.io/nuget/v/KreuzbergDev.HtmlToMarkdown?label=C%23&color=007ec6" alt="C#">

data/ext/html-to-markdown-rb/native/Cargo.lock

@@ -424,7 +424,7 @@ dependencies = [
 
 [[package]]
 name = "html-to-markdown-rb"
-version = "2.24.6"
+version = "2.25.1"
 dependencies = [
  "html-to-markdown-rs",
  "magnus",
@@ -443,7 +443,6 @@ dependencies = [
  "html5ever",
  "image",
  "lru",
- "markup5ever_rcdom",
  "once_cell",
  "regex",
  "serde",
@@ -663,18 +662,6 @@ dependencies = [
  "web_atoms",
 ]
 
-[[package]]
-name = "markup5ever_rcdom"
-version = "0.36.0+unofficial"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "3e5fc8802e8797c0dfdd2ce5c21aa0aee21abbc7b3b18559100651b3352a7b63"
-dependencies = [
- "html5ever",
- "markup5ever",
- "tendril",
- "xml5ever",
-]
-
 [[package]]
 name = "memchr"
 version = "2.8.0"
@@ -683,9 +670,9 @@ checksum = "f8ca58f447f06ed17d5fc4043ce1b10dd205e060fb3ce5b979b8ed8e59ff3f79"
 
 [[package]]
 name = "memmap2"
-version = "0.9.9"
+version = "0.9.10"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "744133e4a0e0a658e1374cf3bf8e415c4052a15a111acd372764c55b4177d490"
+checksum = "714098028fe011992e1c3962653c96b2d578c4b4bce9036e15ff220319b1e0e3"
 dependencies = [
  "libc",
 ]
@@ -841,9 +828,9 @@ dependencies = [
 
 [[package]]
 name = "png"
-version = "0.18.0"
+version = "0.18.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "97baced388464909d42d89643fe4361939af9b7ce7a31ee32a168f832a70f2a0"
+checksum = "60769b8b31b2a9f263dae2776c37b1b28ae246943cf719eb6946a1db05128a61"
 dependencies = [
  "bitflags 2.11.0",
  "crc32fast",
@@ -1217,9 +1204,9 @@ dependencies = [
 
 [[package]]
 name = "syn"
-version = "2.0.115"
+version = "2.0.116"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "6e614ed320ac28113fa64972c4262d5dbc89deacdfd00c34a3e4cea073243c12"
+checksum = "3df424c70518695237746f84cede799c9c58fcb37450d7b23716568cc8bc69cb"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -1272,9 +1259,9 @@ dependencies = [
 
 [[package]]
 name = "unicode-ident"
-version = "1.0.23"
+version = "1.0.24"
 source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "537dd038a89878be9b64dd4bd1b260315c1bb94f4d784956b81e27a088d9a09e"
+checksum = "e6e4313cd5fcd3dad5cafa179702e2b244f760991f45397d14d4ebf38247da75"
 
 [[package]]
 name = "unicode-xid"
@@ -1556,16 +1543,6 @@ dependencies = [
  "wasmparser",
 ]
 
-[[package]]
-name = "xml5ever"
-version = "0.36.1"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "f57dd51b88a4b9f99f9b55b136abb86210629d61c48117ddb87f567e51e66be7"
-dependencies = [
- "log",
- "markup5ever",
-]
-
 [[package]]
 name = "yansi"
 version = "1.0.1"

data/ext/html-to-markdown-rb/native/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "html-to-markdown-rb"
-version ="2.24.6"
+version ="2.25.1"
 edition = "2024"
 authors = ["Na'aman Hirschfeld <naaman@kreuzberg.dev>"]
 license = "MIT"

data/lib/html_to_markdown/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module HtmlToMarkdown
-  VERSION = '2.24.6'
+  VERSION = '2.25.1'
 end

data/rust-vendor/html-to-markdown-rs/Cargo.toml

@@ -35,7 +35,6 @@ ahash = "0.8"
 html-escape = "0.2.13"
 image = { version = "0.25", default-features = false, features = ["gif", "jpeg", "png", "bmp", "webp"], optional = true }
 html5ever = "0.36"
-markup5ever_rcdom = "0.36"
 lru = "0.16"
 serde = { version = "1.0", features = ["derive"], optional = true }
 serde_json = { version = "1.0", optional = true }

data/rust-vendor/html-to-markdown-rs/src/converter/main_helpers.rs

@@ -101,9 +101,9 @@ pub fn has_custom_element_tags(html: &str) -> bool {
 ///
 /// Returns Some(repaired_html) if repair was successful, None otherwise.
 pub fn repair_with_html5ever(input: &str) -> Option<String> {
+    use crate::rcdom::{RcDom, SerializableHandle};
     use html5ever::serialize::{SerializeOpts, serialize};
     use html5ever::tendril::TendrilSink;
-    use markup5ever_rcdom::{RcDom, SerializableHandle};
 
     let dom = html5ever::parse_document(RcDom::default(), Default::default())
         .from_utf8()

data/rust-vendor/html-to-markdown-rs/src/hocr/converter/hierarchy.rs

@@ -57,17 +57,30 @@ pub fn detect_heading_paragraph(element: &HocrElement, text: &str) -> Option<Str
         return None;
     }
 
-    let line_count = element
+    let line_children: Vec<&HocrElement> = element
         .children
         .iter()
         .filter(|child| matches!(child.element_type, HocrElementType::OcrLine | HocrElementType::OcrxLine))
-        .count();
+        .collect();
 
-    if line_count != 1 {
+    if line_children.len() != 1 {
         return None;
     }
 
-    if text.is_empty() || text.len() > 60 || text.contains(':') || text.contains('\n') {
+    // Determine effective font size from child line elements.
+    // First check x_fsize, then fall back to bbox height as a proxy.
+    let font_size = line_children.iter().find_map(|child| {
+        child
+            .properties
+            .x_fsize
+            .or_else(|| child.properties.bbox.map(|b| b.height()))
+    });
+
+    let has_large_font = font_size.is_some_and(|size| size >= 14);
+
+    let char_limit = if has_large_font { 80 } else { 60 };
+
+    if text.is_empty() || text.len() > char_limit || text.contains(':') || text.contains('\n') {
         return None;
     }
 
@@ -83,7 +96,9 @@ pub fn detect_heading_paragraph(element: &HocrElement, text: &str) -> Option<Str
         }
     }
 
-    if word_count < 2 {
+    // Allow single-word headings when font size is large
+    let min_words = if has_large_font { 1 } else { 2 };
+    if word_count < min_words {
         return None;
     }
 

data/rust-vendor/html-to-markdown-rs/src/lib.rs

@@ -68,6 +68,7 @@ pub mod wrapper;
 mod convert_api;
 mod exports;
 pub mod prelude;
+mod rcdom;
 mod validation;
 
 // ============================================================================

data/rust-vendor/{markup5ever_rcdom/lib.rs → html-to-markdown-rs/src/rcdom.rs}

@@ -1,43 +1,39 @@
-// Copyright 2014-2017 The html5ever Project Developers. See the
-// COPYRIGHT file at the top-level directory of this distribution.
+// Vendored from markup5ever_rcdom v0.36.0+unofficial
+// Original source: https://github.com/servo/html5ever (rcdom/)
+// Copyright (c) 2014 The html5ever Project Developers
+// Licensed under MIT OR Apache-2.0 (see ATTRIBUTIONS.md)
 //
-// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
-// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
-// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
-// option. This file may not be copied, modified, or distributed
-// except according to those terms.
+// Vendored to:
+// - Remove unused xml5ever transitive dependency
+// - Eliminate pinned external dependency on "+unofficial" crate
+// - Gain full control over this small, critical module
+//
+// Changes from upstream:
+// - Replaced `extern crate markup5ever` / `extern crate tendril` with
+//   `use` imports through `html5ever` (edition 2024 compatibility)
+// - Added module-level clippy allows for vendored code style
+
+#![allow(
+    clippy::panic,
+    clippy::expect_used,
+    clippy::missing_panics_doc,
+    clippy::must_use_candidate,
+    clippy::return_self_not_must_use,
+    clippy::module_name_repetitions,
+    clippy::redundant_else,
+    clippy::match_wildcard_for_single_variants,
+    clippy::similar_names,
+    clippy::items_after_statements,
+    clippy::use_self,
+    clippy::missing_fields_in_debug,
+    clippy::semicolon_if_nothing_returned,
+    missing_docs
+)]
 
 //! A simple reference-counted DOM.
 //!
 //! This is sufficient as a static parse tree, but don't build a
 //! web browser using it. :)
-//!
-//! A DOM is a [tree structure] with ordered children that can be represented in an XML-like
-//! format. For example, the following graph
-//!
-//! ```text
-//! div
-//!  +- "text node"
-//!  +- span
-//! ```
-//! in HTML would be serialized as
-//!
-//! ```html
-//! <div>text node<span></span></div>
-//! ```
-//!
-//! See the [document object model article on wikipedia][dom wiki] for more information.
-//!
-//! This implementation stores the information associated with each node once, and then hands out
-//! refs to children. The nodes themselves are reference-counted to avoid copying - you can create
-//! a new ref and then a node will outlive the document. Nodes own their children, but only have
-//! weak references to their parents.
-//!
-//! [tree structure]: https://en.wikipedia.org/wiki/Tree_(data_structure)
-//! [dom wiki]: https://en.wikipedia.org/wiki/Document_Object_Model
-
-extern crate markup5ever;
-extern crate tendril;
 
 use std::borrow::Cow;
 use std::cell::{Cell, RefCell};
@@ -48,16 +44,16 @@ use std::io;
 use std::mem;
 use std::rc::{Rc, Weak};
 
-use tendril::StrTendril;
+use html5ever::tendril::StrTendril;
 
-use markup5ever::interface::tree_builder;
-use markup5ever::interface::tree_builder::{ElementFlags, NodeOrText, QuirksMode, TreeSink};
-use markup5ever::serialize::TraversalScope;
-use markup5ever::serialize::TraversalScope::{ChildrenOnly, IncludeNode};
-use markup5ever::serialize::{Serialize, Serializer};
-use markup5ever::Attribute;
-use markup5ever::ExpandedName;
-use markup5ever::QualName;
+use html5ever::Attribute;
+use html5ever::ExpandedName;
+use html5ever::QualName;
+use html5ever::interface::tree_builder;
+use html5ever::interface::tree_builder::{ElementFlags, NodeOrText, QuirksMode, TreeSink};
+use html5ever::serialize::TraversalScope;
+use html5ever::serialize::TraversalScope::{ChildrenOnly, IncludeNode};
+use html5ever::serialize::{Serialize, Serializer};
 
 /// The different kinds of nodes in the DOM.
 #[derive(Debug)]
@@ -98,10 +94,7 @@ pub enum NodeData {
     },
 
     /// A Processing instruction.
-    ProcessingInstruction {
-        target: StrTendril,
-        contents: StrTendril,
-    },
+    ProcessingInstruction { target: StrTendril, contents: StrTendril },
 }
 
 /// A DOM node.
@@ -130,10 +123,9 @@ impl Drop for Node {
         let mut nodes = mem::take(&mut *self.children.borrow_mut());
         while let Some(node) = nodes.pop() {
             let children = mem::take(&mut *node.children.borrow_mut());
-            nodes.extend(children.into_iter());
+            nodes.extend(children);
             if let NodeData::Element {
-                ref template_contents,
-                ..
+                ref template_contents, ..
             } = node.data
             {
                 if let Some(template_contents) = template_contents.borrow_mut().take() {
@@ -193,7 +185,7 @@ fn append_to_existing_text(prev: &Handle, text: &str) -> bool {
         NodeData::Text { ref contents } => {
             contents.borrow_mut().push_slice(text);
             true
-        },
+        }
         _ => false,
     }
 }
@@ -240,8 +232,7 @@ impl TreeSink for RcDom {
 
     fn get_template_contents(&self, target: &Handle) -> Handle {
         if let NodeData::Element {
-            ref template_contents,
-            ..
+            ref template_contents, ..
         } = target.data
         {
             template_contents
@@ -287,10 +278,7 @@ impl TreeSink for RcDom {
     }
 
     fn create_pi(&self, target: StrTendril, data: StrTendril) -> Handle {
-        Node::new(NodeData::ProcessingInstruction {
-            target,
-            contents: data,
-        })
+        Node::new(NodeData::ProcessingInstruction { target, contents: data })
     }
 
     fn append(&self, parent: &Handle, child: NodeOrText<Handle>) {
@@ -315,8 +303,7 @@ impl TreeSink for RcDom {
     }
 
     fn append_before_sibling(&self, sibling: &Handle, child: NodeOrText<Handle>) {
-        let (parent, i) = get_parent_and_index(sibling)
-            .expect("append_before_sibling called on node without parent");
+        let (parent, i) = get_parent_and_index(sibling).expect("append_before_sibling called on node without parent");
 
         let child = match (child, i) {
             // No previous node.
@@ -334,7 +321,7 @@ impl TreeSink for RcDom {
                 Node::new(NodeData::Text {
                     contents: RefCell::new(text),
                 })
-            },
+            }
 
             // The tree builder promises we won't have a text node after
             // the insertion point.
@@ -366,12 +353,7 @@ impl TreeSink for RcDom {
         }
     }
 
-    fn append_doctype_to_document(
-        &self,
-        name: StrTendril,
-        public_id: StrTendril,
-        system_id: StrTendril,
-    ) {
+    fn append_doctype_to_document(&self, name: StrTendril, public_id: StrTendril, system_id: StrTendril) {
         append(
             &self.document,
             Node::new(NodeData::Doctype {
@@ -389,15 +371,8 @@ impl TreeSink for RcDom {
             panic!("not an element")
         };
 
-        let existing_names = existing
-            .iter()
-            .map(|e| e.name.clone())
-            .collect::<HashSet<_>>();
-        existing.extend(
-            attrs
-                .into_iter()
-                .filter(|attr| !existing_names.contains(&attr.name)),
-        );
+        let existing_names = existing.iter().map(|e| e.name.clone()).collect::<HashSet<_>>();
+        existing.extend(attrs.into_iter().filter(|attr| !existing_names.contains(&attr.name)));
     }
 
     fn remove_from_parent(&self, target: &Handle) {
@@ -461,27 +436,17 @@ impl Serialize for SerializableHandle {
         let mut ops = VecDeque::new();
         match traversal_scope {
             IncludeNode => ops.push_back(SerializeOp::Open(self.0.clone())),
-            ChildrenOnly(_) => ops.extend(
-                self.0
-                    .children
-                    .borrow()
-                    .iter()
-                    .map(|h| SerializeOp::Open(h.clone())),
-            ),
+            ChildrenOnly(_) => ops.extend(self.0.children.borrow().iter().map(|h| SerializeOp::Open(h.clone()))),
         }
 
         while let Some(op) = ops.pop_front() {
             match op {
                 SerializeOp::Open(handle) => match handle.data {
                     NodeData::Element {
-                        ref name,
-                        ref attrs,
-                        ..
+                        ref name, ref attrs, ..
                     } => {
-                        serializer.start_elem(
-                            name.clone(),
-                            attrs.borrow().iter().map(|at| (&at.name, &at.value[..])),
-                        )?;
+                        serializer
+                            .start_elem(name.clone(), attrs.borrow().iter().map(|at| (&at.name, &at.value[..])))?;
 
                         ops.reserve(1 + handle.children.borrow().len());
                         ops.push_front(SerializeOp::Close(name.clone()));
@@ -489,7 +454,7 @@ impl Serialize for SerializableHandle {
                         for child in handle.children.borrow().iter().rev() {
                             ops.push_front(SerializeOp::Open(child.clone()));
                         }
-                    },
+                    }
 
                     NodeData::Doctype { ref name, .. } => serializer.write_doctype(name)?,
 
@@ -507,7 +472,7 @@ impl Serialize for SerializableHandle {
 
                 SerializeOp::Close(name) => {
                     serializer.end_elem(name)?;
-                },
+                }
             }
         }
 

data/rust-vendor/html-to-markdown-rs/tests/hocr_compliance_test.rs

@@ -350,3 +350,160 @@ fn test_container_elements() {
     assert!(markdown.contains("Layout analysis"));
     assert!(markdown.contains("Block content"));
 }
+
+#[test]
+fn test_ocr_header_renders_as_italic_not_heading() {
+    // OcrHeader is a "page running header" (repeated at top of pages),
+    // NOT a section heading. It must render as italic (*text*), not as # heading.
+    let hocr = r#"<div class="ocr_page" title="bbox 0 0 1000 1000">
+    <div class="ocr_header" title="bbox 0 0 1000 50">
+        <span class="ocr_line" title="bbox 0 0 500 30">
+            <span class="ocrx_word" title="bbox 0 0 100 30; x_wconf 95">Chapter</span>
+            <span class="ocrx_word" title="bbox 110 0 200 30; x_wconf 95">One</span>
+        </span>
+    </div>
+    <p class="ocr_par" title="bbox 0 100 900 200">
+        <span class="ocr_line" title="bbox 0 100 800 130">
+            <span class="ocrx_word" title="bbox 0 100 50 130; x_wconf 95">Some</span>
+            <span class="ocrx_word" title="bbox 60 100 120 130; x_wconf 95">body</span>
+            <span class="ocrx_word" title="bbox 130 100 180 130; x_wconf 95">text</span>
+            <span class="ocrx_word" title="bbox 190 100 240 130; x_wconf 95">here</span>
+        </span>
+    </p>
+</div>"#;
+
+    let dom = tl::parse(hocr, tl::ParserOptions::default()).unwrap();
+    let (elements, _) = extract_hocr_document(&dom);
+    let markdown = convert_to_markdown(&elements, true);
+
+    // OcrHeader must render as italic
+    assert!(
+        markdown.contains("*Chapter One*"),
+        "OcrHeader should render as italic (*text*), got: {markdown}"
+    );
+    // It must NOT render as a markdown heading
+    assert!(
+        !markdown.contains("# Chapter One"),
+        "OcrHeader must NOT render as a markdown heading, got: {markdown}"
+    );
+}
+
+#[test]
+fn test_heading_detection_with_x_fsize_on_line_child() {
+    // A paragraph containing a single ocr_line child with x_fsize 18 (large font)
+    // and short capitalized text should be detected as a heading.
+    let hocr = r#"<div class="ocr_page" title="bbox 0 0 1000 1000">
+    <div class="ocr_carea" title="bbox 0 0 1000 500">
+        <p class="ocr_par" title="bbox 0 0 500 40">
+            <span class="ocr_line" title="bbox 0 0 500 30; x_fsize 18">
+                <span class="ocrx_word" title="bbox 0 0 120 30; x_wconf 95">Important</span>
+                <span class="ocrx_word" title="bbox 130 0 250 30; x_wconf 95">Section</span>
+                <span class="ocrx_word" title="bbox 260 0 350 30; x_wconf 95">Title</span>
+            </span>
+        </p>
+        <p class="ocr_par" title="bbox 0 60 900 200">
+            <span class="ocr_line" title="bbox 0 60 800 90; x_fsize 12">
+                <span class="ocrx_word" title="bbox 0 60 50 90; x_wconf 95">This</span>
+                <span class="ocrx_word" title="bbox 60 60 90 90; x_wconf 92">is</span>
+                <span class="ocrx_word" title="bbox 100 60 200 90; x_wconf 98">regular</span>
+                <span class="ocrx_word" title="bbox 210 60 280 90; x_wconf 98">body</span>
+                <span class="ocrx_word" title="bbox 290 60 340 90; x_wconf 98">text</span>
+                <span class="ocrx_word" title="bbox 350 60 430 90; x_wconf 98">content.</span>
+            </span>
+        </p>
+    </div>
+</div>"#;
+
+    let dom = tl::parse(hocr, tl::ParserOptions::default()).unwrap();
+    let (elements, _) = extract_hocr_document(&dom);
+    let markdown = convert_to_markdown(&elements, true);
+
+    // The large-font paragraph should be detected as a heading
+    assert!(
+        markdown.contains("# Important Section Title"),
+        "Large font paragraph should be detected as heading, got: {markdown}"
+    );
+}
+
+#[test]
+fn test_single_word_heading_with_large_font() {
+    // A single-word paragraph with large font size should be detected as a heading.
+    // Without font size awareness, single-word paragraphs are rejected.
+    let hocr = r#"<div class="ocr_page" title="bbox 0 0 1000 1000">
+    <div class="ocr_carea" title="bbox 0 0 1000 500">
+        <p class="ocr_par" title="bbox 0 0 300 40">
+            <span class="ocr_line" title="bbox 0 0 300 30; x_fsize 24">
+                <span class="ocrx_word" title="bbox 0 0 200 30; x_wconf 95">Introduction</span>
+            </span>
+        </p>
+        <p class="ocr_par" title="bbox 0 60 900 200">
+            <span class="ocr_line" title="bbox 0 60 800 90; x_fsize 12">
+                <span class="ocrx_word" title="bbox 0 60 50 90; x_wconf 95">Some</span>
+                <span class="ocrx_word" title="bbox 60 60 120 90; x_wconf 92">body</span>
+                <span class="ocrx_word" title="bbox 130 60 180 90; x_wconf 98">text</span>
+                <span class="ocrx_word" title="bbox 190 60 280 90; x_wconf 98">follows.</span>
+            </span>
+        </p>
+    </div>
+</div>"#;
+
+    let dom = tl::parse(hocr, tl::ParserOptions::default()).unwrap();
+    let (elements, _) = extract_hocr_document(&dom);
+    let markdown = convert_to_markdown(&elements, true);
+
+    // Single word with large font should be detected as heading
+    assert!(
+        markdown.contains("# Introduction"),
+        "Single word with large font should be detected as heading, got: {markdown}"
+    );
+}
+
+#[test]
+fn test_single_word_without_large_font_not_heading() {
+    // A single-word paragraph without large font should NOT be detected as heading.
+    // This ensures we haven't broken the existing behavior.
+    let hocr = r#"<div class="ocr_page" title="bbox 0 0 1000 1000">
+    <div class="ocr_carea" title="bbox 0 0 1000 500">
+        <p class="ocr_par" title="bbox 0 0 300 20">
+            <span class="ocr_line" title="bbox 0 0 300 12; x_fsize 10">
+                <span class="ocrx_word" title="bbox 0 0 100 12; x_wconf 95">Word</span>
+            </span>
+        </p>
+    </div>
+</div>"#;
+
+    let dom = tl::parse(hocr, tl::ParserOptions::default()).unwrap();
+    let (elements, _) = extract_hocr_document(&dom);
+    let markdown = convert_to_markdown(&elements, true);
+
+    // Single word with small font should NOT be a heading
+    assert!(
+        !markdown.contains("# Word"),
+        "Single word with small font should not be detected as heading, got: {markdown}"
+    );
+}
+
+#[test]
+fn test_heading_detection_with_bbox_height_proxy() {
+    // When x_fsize is absent, bbox height should serve as a font-size proxy.
+    // A bbox height of 30 pixels (>= 14) indicates large text.
+    let hocr = r#"<div class="ocr_page" title="bbox 0 0 1000 1000">
+    <div class="ocr_carea" title="bbox 0 0 1000 500">
+        <p class="ocr_par" title="bbox 0 0 500 40">
+            <span class="ocr_line" title="bbox 0 0 500 30">
+                <span class="ocrx_word" title="bbox 0 0 200 30; x_wconf 95">Summary</span>
+            </span>
+        </p>
+    </div>
+</div>"#;
+
+    let dom = tl::parse(hocr, tl::ParserOptions::default()).unwrap();
+    let (elements, _) = extract_hocr_document(&dom);
+    let markdown = convert_to_markdown(&elements, true);
+
+    // bbox height of 30 (y2=30 - y1=0) should serve as proxy for large font
+    assert!(
+        markdown.contains("# Summary"),
+        "Single word with tall bbox (height=30) should be detected as heading via bbox proxy, got: {markdown}"
+    );
+}

data/rust-vendor/memmap2/.cargo-checksum.json

@@ -1 +1 @@
-{"files":{".cargo_vcs_info.json":"175df596d2f8c352eaca47603bf5b23a0eab080ffb910b914fa2e3a8ad191159",".github/workflows/main.yml":"e2b3d5678a31325a616bae0c1ccb223f9ed2f5b43d39a134c5c45858e4fbf4ca","CHANGELOG.md":"63d0899dd592ddfedaf38832478ff0342726ab951021fb9b7bc237ce7be4066a","Cargo.lock":"ab24cb6544ffc0a82625e5bde17c6da152490577a0659851a1c77e767fdcd413","Cargo.toml":"c50a23ef2999b6a5ca3fd96d72b8932ebfc2f940acfed78a881c73f7eeaa91dd","Cargo.toml.orig":"11b7a3b11bea46362dd87d4371a2495f6ca0f1f7957011e8802d4c6d20fe4c4a","LICENSE-APACHE":"04ea4849dba9dcae07113850c6f1b1a69052c625210639914eee352023f750ad","LICENSE-MIT":"0d25d03b5ab49576178ad0cae7a2648d12c17ad0452fe49c07e55e4b59aa5257","README.md":"e3388f55065d69e076d90871c0a91dc97420bd0d07b4f154b08e40ac47b115eb","examples/cat.rs":"594b9457ca6eb4ce9b840133da5076fa7b96334953df03f894233169564622f6","src/advice.rs":"a4c023982a598a77c23b5a4e524de581329d42287d639be88e2ffda3bd929511","src/lib.rs":"50f44261aec0353976962d1314d2d885e9a34afda6215bb5b67d3afb336519e1","src/stub.rs":"656088d19012fce05de58420c8b84d59f5f1d10c3bb79a950c7f7394bca5fae1","src/unix.rs":"6133da835aff4aa7bf29f17e43d501c6409a6b343b00232fd082f4a0db035cdb","src/windows.rs":"0fbd1efc122a7e83defb5d0a401e973a3876e90c6a1f14f2ac1976462633dd79"},"package":"744133e4a0e0a658e1374cf3bf8e415c4052a15a111acd372764c55b4177d490"}
+{"files":{".cargo_vcs_info.json":"57a15fc6b2ddb894def64e09a11e5cc52e575be77740a6fc0aa3c451c8c0fa49",".github/workflows/main.yml":"e2b3d5678a31325a616bae0c1ccb223f9ed2f5b43d39a134c5c45858e4fbf4ca","CHANGELOG.md":"832a95917c80f443c113ede708a5602550d081bd79cc33f4152365357ab36bb8","Cargo.lock":"6ddd1fa91a86a6e8bbbc17161703f21799213d823ea26d7f7ce86181295a3fc8","Cargo.toml":"01ff6425d680ddac0a8efb1b41e6b0503d0153388fadcc4bf584ca38c4dca221","Cargo.toml.orig":"507519ce5facd6b78a80e3cfab1437c7b6b5a1fc99c65e24552525c884d4ed3e","LICENSE-APACHE":"04ea4849dba9dcae07113850c6f1b1a69052c625210639914eee352023f750ad","LICENSE-MIT":"0d25d03b5ab49576178ad0cae7a2648d12c17ad0452fe49c07e55e4b59aa5257","README.md":"e3388f55065d69e076d90871c0a91dc97420bd0d07b4f154b08e40ac47b115eb","examples/cat.rs":"594b9457ca6eb4ce9b840133da5076fa7b96334953df03f894233169564622f6","src/advice.rs":"a4c023982a598a77c23b5a4e524de581329d42287d639be88e2ffda3bd929511","src/lib.rs":"d93f73dd80b5bfdecc10836a7ebcd04c124f6283f9a104686fe48a18d34764ab","src/stub.rs":"beccccb0233903df5de1773674b2dcd9b0991889a10c23719f5aee8f7496f958","src/unix.rs":"fea7c7c21a6082bc77052e5e40a2bff1311103a19c2cd281ff383604fa799b35","src/windows.rs":"0fbd1efc122a7e83defb5d0a401e973a3876e90c6a1f14f2ac1976462633dd79"},"package":"714098028fe011992e1c3962653c96b2d578c4b4bce9036e15ff220319b1e0e3"}

data/rust-vendor/memmap2/.cargo_vcs_info.json

@@ -1,6 +1,6 @@
 {
   "git": {
-    "sha1": "cc94666eabb82a3a1647f8dc1d949a9ffc6babff"
+    "sha1": "1f0196ab31cf1e6f4eb4b3633c51a4d3ba920788"
   },
   "path_in_vcs": ""
 }