fix(docx): preserve drawing shapes, list order/nesting, code breaks & highlighting (#176)

README.md

@@ -11,7 +11,7 @@ No LibreOffice, no Chromium, no Docker — just a single binary powered by [Typs
 
 ## Features
 
-- **DOCX** — paragraphs, inline formatting (bold/italic/underline/color), tables, images, lists, headers/footers, page setup
+- **DOCX** — paragraphs, inline formatting (bold/italic/underline/color), tables, images, drawing shapes, ordered/nested lists, syntax-highlighted code, headers/footers, page setup
 - **PPTX** — slides, text boxes, shapes, tables (with theme-based table styles), images, slide masters, speaker notes, gradient backgrounds, shadow/reflection effects
 - **XLSX** — sheets, cell formatting, merged cells, column widths, row heights, conditional formatting (DataBar, IconSet)
 - **PDF/A-2b** — archival-compliant output via `--pdf-a`
@@ -128,7 +128,7 @@ Available functions: `convertToPdf(data, format)`, `convertDocxToPdf(data)`, `co
 
 | Format | Status | Key Features |
 |--------|--------|-------------|
-| DOCX | Supported | Text, tables, images, lists, headers/footers, page setup |
+| DOCX | Supported | Text, tables, images, drawing shapes, lists, code highlighting, headers/footers, page setup |
 | PPTX | Supported | Slides, text boxes, shapes, tables, images, masters, gradients, effects |
 | XLSX | Supported | Sheets, formatting, merged cells, column/row sizing, conditional formatting |
 

crates/office2pdf/src/ir/elements.rs

@@ -34,6 +34,7 @@ pub enum Block {
     Image(ImageData),
     FloatingImage(FloatingImage),
     FloatingTextBox(FloatingTextBox),
+    FloatingShape(FloatingShape),
     List(List),
     MathEquation(MathEquation),
     Chart(Chart),
@@ -125,6 +126,24 @@ pub struct FloatingTextBox {
     pub offset_y: f64,
 }
 
+/// A floating geometric shape (rectangle, line/arrow, ellipse, …) positioned
+/// with an anchor offset. Used for DrawingML word-processing shapes (`wps:wsp`)
+/// that carry geometry but no text box — these have no docx-rs representation
+/// and would otherwise be dropped (issue #176).
+#[derive(Debug, Clone)]
+pub struct FloatingShape {
+    pub shape: Shape,
+    /// On-page bounding-box width in points (from `wp:extent`).
+    pub width: f64,
+    /// On-page bounding-box height in points (from `wp:extent`).
+    pub height: f64,
+    /// Horizontal offset in points from the anchor reference.
+    pub offset_x: f64,
+    /// Vertical offset in points from the anchor reference.
+    pub offset_y: f64,
+    pub wrap_mode: WrapMode,
+}
+
 /// Vertical alignment for fixed text box content.
 #[derive(Debug, Clone, Copy, PartialEq, Eq, Default)]
 pub enum TextBoxVerticalAlign {

crates/office2pdf/src/parser/docx.rs

@@ -17,10 +17,10 @@ use crate::parser::Parser;
 #[cfg(test)]
 use self::contexts::scan_table_headers;
 use self::contexts::{
-    BidiContext, ChartContext, DocxConversionContext, DrawingTextBoxContext, DrawingTextBoxInfo,
-    MathContext, NoteContext, SmallCapsContext, TableHeaderContext, VmlTextBoxContext,
-    VmlTextBoxInfo, WrapContext, build_chart_context_from_xml, build_math_context_from_xml,
-    build_note_context_from_xml, build_wrap_context_from_xml,
+    BidiContext, ChartContext, DocxConversionContext, DrawingShapeContext, DrawingTextBoxContext,
+    DrawingTextBoxInfo, MathContext, NoteContext, SmallCapsContext, TableHeaderContext,
+    VmlTextBoxContext, VmlTextBoxInfo, WrapContext, build_chart_context_from_xml,
+    build_math_context_from_xml, build_note_context_from_xml, build_wrap_context_from_xml,
     extract_column_layout_from_section_property, is_note_reference_run, read_zip_text,
     scan_column_layouts,
 };
@@ -42,9 +42,9 @@ use self::styles::{
 };
 use self::tables::convert_table;
 use self::text::{
-    extract_doc_default_text_style, extract_paragraph_style, extract_run_style, extract_run_text,
-    extract_run_text_skip_column_breaks, extract_tab_stop_overrides, is_column_break,
-    parse_hex_color, resolve_hyperlink_url,
+    extract_doc_default_text_style, extract_paragraph_style, extract_run_style,
+    extract_run_style_id, extract_run_text, extract_run_text_skip_column_breaks,
+    extract_tab_stop_overrides, is_column_break, parse_hex_color, resolve_hyperlink_url,
 };
 #[cfg(test)]
 use self::text::{extract_tab_stops, resolve_highlight_color};
@@ -112,6 +112,7 @@ fn build_zip_preparse_assets(data: &[u8]) -> ZipPreParseAssets {
             let notes = build_note_context_from_xml(doc_xml.as_deref(), &mut archive);
             let wraps = build_wrap_context_from_xml(doc_xml.as_deref());
             let drawing_text_boxes = DrawingTextBoxContext::from_xml(doc_xml.as_deref());
+            let drawing_shapes = DrawingShapeContext::from_xml(doc_xml.as_deref());
             let table_headers = TableHeaderContext::from_xml(doc_xml.as_deref());
             let vml_text_boxes = VmlTextBoxContext::from_xml(doc_xml.as_deref());
             let math = build_math_context_from_xml(doc_xml.as_deref());
@@ -127,6 +128,7 @@ fn build_zip_preparse_assets(data: &[u8]) -> ZipPreParseAssets {
                 notes,
                 wraps,
                 drawing_text_boxes,
+                drawing_shapes,
                 table_headers,
                 vml_text_boxes,
                 bidi,
@@ -147,6 +149,7 @@ fn build_zip_preparse_assets(data: &[u8]) -> ZipPreParseAssets {
                 notes: NoteContext::empty(),
                 wraps: WrapContext::empty(),
                 drawing_text_boxes: DrawingTextBoxContext::from_xml(None),
+                drawing_shapes: DrawingShapeContext::from_xml(None),
                 table_headers: TableHeaderContext::from_xml(None),
                 vml_text_boxes: VmlTextBoxContext::from_xml(None),
                 bidi: BidiContext::from_xml(None),
@@ -386,6 +389,7 @@ fn build_text_run(
     run_property: &docx_rs::RunProperty,
     is_small_caps: bool,
     resolved_style: Option<&ResolvedStyle>,
+    style_map: &StyleMap,
     href: Option<String>,
 ) -> Option<Run> {
     if text.is_empty() {
@@ -395,6 +399,14 @@ fn build_text_run(
     if is_small_caps {
         explicit_style.small_caps = Some(true);
     }
+    // Layer the referenced character style (`<w:rStyle>`, e.g. a syntax
+    // highlighting token) beneath the run's explicit properties so its color
+    // and weight apply while explicit run formatting still wins (issue #176).
+    if let Some(char_style) = extract_run_style_id(run_property).and_then(|id| style_map.get(&id)) {
+        let mut combined: TextStyle = char_style.text.clone();
+        combined.merge_from(&explicit_style);
+        explicit_style = combined;
+    }
     Some(Run {
         text,
         style: merge_text_style(&explicit_style, resolved_style),
@@ -435,6 +447,16 @@ fn extract_run_children_media(
                 drawing, images, hyperlinks, style_map, ctx,
             ));
         }
+        // A `<w:drawing>` that docx-rs cannot classify as a picture or a text box
+        // (geometry-only `wps:wsp` shapes) leaves `data == None`. Pair each such
+        // drawing, in document order, with the geometry scanned from the raw XML
+        // so rectangles, lines and arrows are not dropped (issue #176).
+        if let docx_rs::RunChild::Drawing(drawing) = run_child
+            && drawing.data.is_none()
+            && let Some(shape) = ctx.drawing_shapes.consume_next()
+        {
+            text_box_blocks.push(Block::FloatingShape(shape));
+        }
         if let docx_rs::RunChild::Shape(shape) = run_child {
             let vml_text_box: VmlTextBoxInfo = ctx.vml_text_boxes.consume_next();
             if let Some(floating_text_box) = extract_vml_shape_text_box(shape, &vml_text_box) {
@@ -465,6 +487,7 @@ fn process_hyperlink_runs(
     hyperlink: &docx_rs::Hyperlink,
     hyperlinks: &HyperlinkMap,
     resolved_style: Option<&ResolvedStyle>,
+    style_map: &StyleMap,
     ctx: &DocxConversionContext,
     runs: &mut Vec<Run>,
 ) {
@@ -478,6 +501,7 @@ fn process_hyperlink_runs(
                 &run.run_property,
                 hl_small_caps,
                 resolved_style,
+                style_map,
                 href.clone(),
             ) {
                 runs.push(ir_run);
@@ -514,7 +538,9 @@ fn convert_paragraph_blocks(
     // Collect text runs and detect inline images
     let mut runs: Vec<Run> = Vec::new();
     let mut inline_images: Vec<Block> = Vec::new();
-    let mut emitted_text_box_blocks: bool = false;
+    let mut emitted_paragraph: bool = false;
+    let mut emitted_media_blocks: bool = false;
+    let mut emitted_floating_anchor: bool = false;
 
     for child in &para.children {
         match child {
@@ -545,13 +571,17 @@ fn convert_paragraph_blocks(
                 );
 
                 if !media.text_box_blocks.is_empty() {
+                    emitted_media_blocks = true;
+                    emitted_floating_anchor |= media.text_box_blocks.iter().any(|block| {
+                        matches!(block, Block::FloatingShape(_) | Block::FloatingTextBox(_))
+                    });
                     if !runs.is_empty() {
                         out.append(&mut inline_images);
                         push_paragraph_from_runs(out, para, resolved_style, is_rtl, &mut runs);
+                        emitted_paragraph = true;
                     } else if !inline_images.is_empty() {
                         out.append(&mut inline_images);
                     }
-                    emitted_text_box_blocks = true;
                     out.extend(media.text_box_blocks);
                 }
 
@@ -560,27 +590,45 @@ fn convert_paragraph_blocks(
                     if !runs.is_empty() {
                         out.append(&mut inline_images);
                         push_paragraph_from_runs(out, para, resolved_style, is_rtl, &mut runs);
+                        emitted_paragraph = true;
                     }
                     out.push(Block::ColumnBreak);
 
                     // Still extract any text from this run (after the break)
                     let text: String = extract_run_text_skip_column_breaks(run);
-                    if let Some(ir_run) =
-                        build_text_run(text, &run.run_property, is_small_caps, resolved_style, None)
-                    {
+                    if let Some(ir_run) = build_text_run(
+                        text,
+                        &run.run_property,
+                        is_small_caps,
+                        resolved_style,
+                        style_map,
+                        None,
+                    ) {
                         runs.push(ir_run);
                     }
                 } else {
                     let text: String = extract_run_text(run);
-                    if let Some(ir_run) =
-                        build_text_run(text, &run.run_property, is_small_caps, resolved_style, None)
-                    {
+                    if let Some(ir_run) = build_text_run(
+                        text,
+                        &run.run_property,
+                        is_small_caps,
+                        resolved_style,
+                        style_map,
+                        None,
+                    ) {
                         runs.push(ir_run);
                     }
                 }
             }
             docx_rs::ParagraphChild::Hyperlink(hyperlink) => {
-                process_hyperlink_runs(hyperlink, hyperlinks, resolved_style, ctx, &mut runs);
+                process_hyperlink_runs(
+                    hyperlink,
+                    hyperlinks,
+                    resolved_style,
+                    style_map,
+                    ctx,
+                    &mut runs,
+                );
             }
             _ => {}
         }
@@ -589,7 +637,11 @@ fn convert_paragraph_blocks(
     // Emit image blocks before the paragraph (inline images are block-level in our IR)
     out.extend(inline_images);
 
-    if !runs.is_empty() || !emitted_text_box_blocks {
+    if !runs.is_empty() || !emitted_media_blocks || (emitted_floating_anchor && !emitted_paragraph)
+    {
+        // Keep paragraph marks for floating drawing anchors. The drawing itself
+        // is positioned by offsets, but the source paragraph still contributes
+        // to flow spacing between the drawing cluster and following content.
         push_paragraph_from_runs(out, para, resolved_style, is_rtl, &mut runs);
     }
 }