html-to-markdown 3.4.0 → 3.6.0.pre.rc.23

data/vendor/html-to-markdown-rs/src/converter/semantic/definition_list.rs

@@ -1,266 +0,0 @@
-//! Handlers for HTML5 definition list and heading group elements.
-//!
-//! Processes list and heading semantic elements:
-//! - `<hgroup>` - Groups related headings together
-//! - `<dl>` - Definition list container
-//! - `<dt>` - Definition term
-//! - `<dd>` - Definition description
-//! - `<menu>` - Semantic list (typically unordered)
-//!
-//! These elements have special formatting requirements for proper Markdown output.
-
-// Note: Context and DomContext are defined in converter.rs
-// walk_node is also defined there and must be called via the parent module
-use super::walk_node;
-
-/// Handles the `<hgroup>` element.
-///
-/// An hgroup element groups related headings together (e.g., a title and subtitle).
-/// In Markdown, we simply process all children sequentially, allowing nested
-/// headings to maintain their individual formatting.
-///
-/// # Behavior
-///
-/// - Children are processed sequentially in the current context
-/// - No special formatting is applied at the hgroup level
-pub fn handle_hgroup(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                walk_node(child_handle, parser, output, options, ctx, depth, dom_ctx);
-            }
-        }
-    }
-}
-
-/// Handles the `<dl>` element.
-///
-/// A definition list contains terms and their definitions. Terms and definitions
-/// are output as plain blocks without Pandoc-style colon syntax, since standard
-/// Markdown and GFM do not support definition lists.
-///
-/// # Behavior
-///
-/// - **Inline mode**: Children are processed inline without block spacing
-/// - **Block mode**: Content is collected and wrapped with proper spacing
-pub fn handle_dl(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        // In inline context, just process children inline
-        if ctx.convert_as_inline {
-            let children = tag.children();
-            {
-                for child_handle in children.top().iter() {
-                    walk_node(child_handle, parser, output, options, ctx, depth, dom_ctx);
-                }
-            }
-            return;
-        }
-
-        // Collect content from children
-        let mut content = String::new();
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                walk_node(child_handle, parser, &mut content, options, ctx, depth, dom_ctx);
-            }
-        }
-
-        // Output collected content with proper spacing
-        let trimmed = content.trim();
-        if !trimmed.is_empty() {
-            if !output.is_empty() && !output.ends_with("\n\n") {
-                output.push_str("\n\n");
-            }
-            output.push_str(trimmed);
-            output.push_str("\n\n");
-        }
-    }
-}
-
-/// Handles the `<dt>` element.
-///
-/// A dt element contains a term being defined. Terms are output on their own line,
-/// with definitions following on subsequent lines.
-///
-/// # Behavior
-///
-/// - **Inline mode**: Content is output as-is
-/// - **Block mode**: Content is followed by a newline
-pub fn handle_dt(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        let mut content = String::with_capacity(64);
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                walk_node(child_handle, parser, &mut content, options, ctx, depth + 1, dom_ctx);
-            }
-        }
-
-        let trimmed = content.trim();
-        if !trimmed.is_empty() {
-            if ctx.convert_as_inline {
-                output.push_str(trimmed);
-            } else {
-                output.push_str(trimmed);
-                output.push('\n');
-            }
-        }
-    }
-}
-
-/// Handles the `<dd>` element.
-///
-/// A dd element contains the definition for a term. It is output as a plain
-/// block since standard Markdown and GFM do not support definition list syntax.
-///
-/// # Behavior
-///
-/// - **Inline mode**: Content is output as-is
-/// - **Block mode**: Content is output as a block
-pub fn handle_dd(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        let mut content = String::with_capacity(128);
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                walk_node(child_handle, parser, &mut content, options, ctx, depth + 1, dom_ctx);
-            }
-        }
-
-        let trimmed = content.trim();
-
-        if ctx.convert_as_inline {
-            if !trimmed.is_empty() {
-                output.push_str(trimmed);
-            }
-        } else if !trimmed.is_empty() {
-            output.push_str(trimmed);
-            output.push_str("\n\n");
-        }
-    }
-}
-
-/// Handles the `<menu>` element.
-///
-/// A menu element is a semantic list, typically used for command menus or
-/// navigation. It is rendered as an unordered list with dashes.
-///
-/// # Behavior
-///
-/// - **Inline mode**: Children are processed inline without list formatting
-/// - **Block mode**: Content is rendered as an unordered list
-/// - Uses `-` as the list bullet (overrides configured bullets)
-/// - Proper blank-line spacing is maintained
-pub fn handle_menu(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        let content_start = output.len();
-
-        // Create options with menu-specific bullet style
-        let menu_options = crate::options::ConversionOptions {
-            bullets: "-".to_string(),
-            ..options.clone()
-        };
-
-        // Create context for list rendering
-        let list_ctx = super::Context {
-            in_ordered_list: false,
-            list_counter: 0,
-            in_list: true,
-            list_depth: ctx.list_depth,
-            ..ctx.clone()
-        };
-
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                walk_node(child_handle, parser, output, &menu_options, &list_ctx, depth, dom_ctx);
-            }
-        }
-
-        // Ensure proper spacing after menu
-        if !ctx.convert_as_inline && output.len() > content_start {
-            if !output.ends_with("\n\n") {
-                if output.ends_with('\n') {
-                    output.push('\n');
-                } else {
-                    output.push_str("\n\n");
-                }
-            }
-        } else if ctx.convert_as_inline {
-            // In inline mode, remove trailing newlines
-            while output.ends_with('\n') {
-                output.pop();
-            }
-        }
-    }
-}
-
-/// Dispatcher for definition list and related elements.
-///
-/// Routes `<hgroup>`, `<dl>`, `<dt>`, `<dd>`, and `<menu>` elements
-/// to their respective handlers.
-pub fn handle(
-    tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    match tag_name {
-        "hgroup" => handle_hgroup(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        "dl" => handle_dl(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        "dt" => handle_dt(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        "dd" => handle_dd(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        "menu" => handle_menu(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        _ => {}
-    }
-}

data/vendor/html-to-markdown-rs/src/converter/semantic/figure.rs

@@ -1,391 +0,0 @@
-//! Handlers for HTML5 figure elements.
-//!
-//! Processes figure-related semantic elements:
-//! - `<figure>` - Self-contained illustration, diagram, photo, code listing, etc.
-//! - `<figcaption>` - Caption or legend for a figure
-//!
-//! Figure elements group an image (or other media) with its associated caption.
-//! The Markdown output preserves this relationship through content organization.
-
-// Note: Context and DomContext are defined in converter.rs
-// walk_node is also defined there and must be called via the parent module
-
-/// Handles the `<figure>` element.
-///
-/// A figure element contains content (typically images) and optionally a figcaption.
-/// The handler collects all content and cleans up extra line breaks.
-///
-/// # Behavior
-///
-/// - **Inline mode**: Children are processed inline without block spacing
-/// - **Block mode**: Content is collected, line breaks normalized, and wrapped with blank lines
-/// - **Image normalization**: Removes extra spaces before `![` to improve Markdown formatting
-///
-/// # Implementation Details
-///
-/// The handler performs the following on the collected content:
-/// 1. Normalizes newline + image sequences: `\n![` → `![`
-/// 2. Normalizes space + image sequences: ` ![` → `![`
-/// 3. Trims the final content and wraps it with blank lines
-#[cfg_attr(not(feature = "visitor"), allow(unused_variables))]
-pub fn handle_figure(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        // In inline context, just process children inline
-        if ctx.convert_as_inline {
-            let children = tag.children();
-            {
-                for child_handle in children.top().iter() {
-                    super::walk_node(child_handle, parser, output, options, ctx, depth, dom_ctx);
-                }
-            }
-            return;
-        }
-
-        // Check visit_figure_start before processing children.
-        #[cfg(feature = "visitor")]
-        if let Some(ref visitor_handle) = ctx.visitor {
-            use crate::converter::utility::content::collect_tag_attributes;
-            use crate::visitor::{NodeContext, NodeType, VisitResult};
-
-            let attributes = collect_tag_attributes(tag);
-            let node_id = node_handle.get_inner();
-            let parent_tag = dom_ctx.parent_tag_name(node_id, parser);
-            let index_in_parent = dom_ctx.get_sibling_index(node_id).unwrap_or(0);
-            let node_ctx = NodeContext {
-                node_type: NodeType::Figure,
-                tag_name: "figure".to_string(),
-                attributes,
-                depth,
-                index_in_parent,
-                parent_tag,
-                is_inline: false,
-            };
-            let visit_result = {
-                let mut visitor = visitor_handle.borrow_mut();
-                visitor.visit_figure_start(&node_ctx)
-            };
-            match visit_result {
-                VisitResult::Continue => {}
-                VisitResult::Skip => return,
-                VisitResult::Custom(custom) => {
-                    // Record where custom content starts so visit_figure_end can see it
-                    let start_pos = output.len();
-                    if !output.is_empty() && !output.ends_with("\n\n") {
-                        output.push_str("\n\n");
-                    }
-                    output.push_str(&custom);
-                    // Still call visit_figure_end with the custom content
-                    let safe_start = start_pos.min(output.len());
-                    let figure_output = output[safe_start..].to_owned();
-                    let end_result = {
-                        let mut visitor = visitor_handle.borrow_mut();
-                        visitor.visit_figure_end(&node_ctx, &figure_output)
-                    };
-                    match end_result {
-                        VisitResult::Continue => {
-                            if !output.ends_with('\n') {
-                                output.push_str("\n\n");
-                            } else if !output.ends_with("\n\n") {
-                                output.push('\n');
-                            }
-                        }
-                        VisitResult::Custom(end_custom) => {
-                            output.truncate(safe_start);
-                            output.push_str(&end_custom);
-                        }
-                        VisitResult::Skip => {
-                            output.truncate(safe_start);
-                        }
-                        VisitResult::PreserveHtml => {
-                            use crate::converter::utility::serialization::serialize_node;
-                            output.truncate(safe_start);
-                            output.push_str(&serialize_node(node_handle, parser));
-                        }
-                        VisitResult::Error(err) => {
-                            if ctx.visitor_error.borrow().is_none() {
-                                *ctx.visitor_error.borrow_mut() = Some(err);
-                            }
-                        }
-                    }
-                    return;
-                }
-                VisitResult::PreserveHtml => {
-                    use crate::converter::utility::serialization::serialize_node;
-                    output.push_str(&serialize_node(node_handle, parser));
-                    return;
-                }
-                VisitResult::Error(err) => {
-                    if ctx.visitor_error.borrow().is_none() {
-                        *ctx.visitor_error.borrow_mut() = Some(err);
-                    }
-                    return;
-                }
-            }
-        }
-
-        // Ensure spacing before the figure
-        if !output.is_empty() && !output.ends_with("\n\n") {
-            output.push_str("\n\n");
-        }
-
-        let figure_start = output.len();
-
-        // Collect content in a separate buffer
-        let mut figure_content = String::new();
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                super::walk_node(child_handle, parser, &mut figure_content, options, ctx, depth, dom_ctx);
-            }
-        }
-
-        // Normalize image syntax
-        figure_content = figure_content.replace("\n![", "![");
-        figure_content = figure_content.replace(" ![", "![");
-
-        // Trim and output
-        let trimmed = figure_content.trim_matches(|c| c == '\n' || c == ' ' || c == '\t');
-        if !trimmed.is_empty() {
-            output.push_str(trimmed);
-            if !output.ends_with('\n') {
-                output.push('\n');
-            }
-            if !output.ends_with("\n\n") {
-                output.push('\n');
-            }
-        }
-
-        // Call visit_figure_end after collecting content.
-        #[cfg(feature = "visitor")]
-        if let Some(ref visitor_handle) = ctx.visitor {
-            use crate::converter::utility::content::collect_tag_attributes;
-            use crate::visitor::{NodeContext, NodeType, VisitResult};
-
-            let attributes = collect_tag_attributes(tag);
-            let node_id = node_handle.get_inner();
-            let parent_tag = dom_ctx.parent_tag_name(node_id, parser);
-            let index_in_parent = dom_ctx.get_sibling_index(node_id).unwrap_or(0);
-            let node_ctx = NodeContext {
-                node_type: NodeType::Figure,
-                tag_name: "figure".to_string(),
-                attributes,
-                depth,
-                index_in_parent,
-                parent_tag,
-                is_inline: false,
-            };
-            let safe_start = figure_start.min(output.len());
-            let figure_output = output[safe_start..].to_owned();
-            let visit_result = {
-                let mut visitor = visitor_handle.borrow_mut();
-                visitor.visit_figure_end(&node_ctx, &figure_output)
-            };
-            match visit_result {
-                VisitResult::Continue => {}
-                VisitResult::Skip => {
-                    output.truncate(safe_start);
-                }
-                VisitResult::Custom(custom) => {
-                    output.truncate(safe_start);
-                    output.push_str(&custom);
-                }
-                VisitResult::PreserveHtml => {
-                    use crate::converter::utility::serialization::serialize_node;
-                    output.truncate(safe_start);
-                    output.push_str(&serialize_node(node_handle, parser));
-                }
-                VisitResult::Error(err) => {
-                    if ctx.visitor_error.borrow().is_none() {
-                        *ctx.visitor_error.borrow_mut() = Some(err);
-                    }
-                }
-            }
-        }
-    }
-}
-
-/// Handles the `<figcaption>` element.
-///
-/// A figcaption element contains text that describes or supplements the figure.
-/// It is rendered as emphasized (italic) text to distinguish it from regular content.
-///
-/// # Behavior
-///
-/// - Content is collected and trimmed
-/// - Non-empty content is wrapped in `*text*` (emphasis) markers
-/// - Proper spacing is maintained around the caption
-///
-/// # Implementation Details
-///
-/// The handler:
-/// 1. Collects and processes all children
-/// 2. Checks for existing output and adds spacing as needed
-/// 3. Wraps content in emphasis markers: `*caption*`
-/// 4. Ensures proper blank-line spacing after the caption
-#[cfg_attr(not(feature = "visitor"), allow(unused_variables))]
-pub fn handle_figcaption(
-    _tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    if let Some(tl::Node::Tag(tag)) = node_handle.get(parser) {
-        let mut text = String::new();
-        let children = tag.children();
-        {
-            for child_handle in children.top().iter() {
-                super::walk_node(child_handle, parser, &mut text, options, ctx, depth + 1, dom_ctx);
-            }
-        }
-
-        let text = text.trim().to_owned();
-        if text.is_empty() {
-            return;
-        }
-
-        #[cfg(feature = "visitor")]
-        if let Some(ref visitor_handle) = ctx.visitor {
-            use crate::converter::utility::content::collect_tag_attributes;
-            use crate::visitor::{NodeContext, NodeType, VisitResult};
-
-            let attributes = collect_tag_attributes(tag);
-            let node_id = node_handle.get_inner();
-            let parent_tag = dom_ctx.parent_tag_name(node_id, parser);
-            let index_in_parent = dom_ctx.get_sibling_index(node_id).unwrap_or(0);
-            let node_ctx = NodeContext {
-                node_type: NodeType::Figcaption,
-                tag_name: "figcaption".to_string(),
-                attributes,
-                depth,
-                index_in_parent,
-                parent_tag,
-                is_inline: false,
-            };
-            let visit_result = {
-                let mut visitor = visitor_handle.borrow_mut();
-                visitor.visit_figcaption(&node_ctx, &text)
-            };
-            match visit_result {
-                VisitResult::Continue => {}
-                VisitResult::Skip => return,
-                VisitResult::Custom(custom) => {
-                    // Add spacing before caption if needed
-                    if !output.is_empty() {
-                        if output.ends_with("```\n") {
-                            output.push('\n');
-                        } else {
-                            while output.ends_with(' ') || output.ends_with('\t') {
-                                output.pop();
-                            }
-                            if output.ends_with('\n') && !output.ends_with("\n\n") {
-                                output.push('\n');
-                            } else if !output.ends_with('\n') {
-                                output.push_str("\n\n");
-                            }
-                        }
-                    }
-                    output.push_str(&custom);
-                    if !custom.ends_with('\n') {
-                        output.push_str("\n\n");
-                    }
-                    return;
-                }
-                VisitResult::PreserveHtml => {
-                    use crate::converter::utility::serialization::serialize_node;
-                    output.push_str(&serialize_node(node_handle, parser));
-                    return;
-                }
-                VisitResult::Error(err) => {
-                    if ctx.visitor_error.borrow().is_none() {
-                        *ctx.visitor_error.borrow_mut() = Some(err);
-                    }
-                    return;
-                }
-            }
-        }
-
-        // Add spacing before caption if needed
-        if !output.is_empty() {
-            if output.ends_with("```\n") {
-                output.push('\n');
-            } else {
-                // Trim trailing whitespace and ensure single blank line
-                while output.ends_with(' ') || output.ends_with('\t') {
-                    output.pop();
-                }
-                if output.ends_with('\n') && !output.ends_with("\n\n") {
-                    output.push('\n');
-                } else if !output.ends_with('\n') {
-                    output.push_str("\n\n");
-                }
-            }
-        }
-
-        // Output caption as emphasized text
-        output.push('*');
-        output.push_str(&text);
-        output.push_str("*\n\n");
-    }
-}
-
-/// Dispatcher for figure-related elements.
-///
-/// Routes `<figure>` and `<figcaption>` elements to their respective handlers.
-pub fn handle(
-    tag_name: &str,
-    node_handle: &tl::NodeHandle,
-    parser: &tl::Parser,
-    output: &mut String,
-    options: &crate::options::ConversionOptions,
-    ctx: &super::Context,
-    depth: usize,
-    dom_ctx: &super::DomContext,
-) {
-    match tag_name {
-        "figure" => handle_figure(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        "figcaption" => handle_figcaption(tag_name, node_handle, parser, output, options, ctx, depth, dom_ctx),
-        _ => {}
-    }
-}
-
-#[cfg(test)]
-mod tests {
-    #[test]
-    fn figure_caption_separated_from_image() {
-        let html = r#"<figure><img src="photo.jpg" alt="Photo"><figcaption>A nice photo</figcaption></figure>"#;
-        let result = crate::convert(html, None).unwrap();
-        let content = result.content.unwrap_or_default();
-        assert!(
-            content.contains("![Photo](photo.jpg)"),
-            "image should be present: {}",
-            content
-        );
-        assert!(
-            content.contains("A nice photo"),
-            "caption should be present: {}",
-            content
-        );
-        // Image and caption should not be on the same line
-        let lines: Vec<&str> = content.lines().filter(|l| !l.trim().is_empty()).collect();
-        let img_line = lines.iter().position(|l| l.contains("![")).unwrap_or(999);
-        let cap_line = lines.iter().position(|l| l.contains("A nice photo")).unwrap_or(999);
-        assert!(
-            cap_line > img_line,
-            "caption should be on a separate line after image, lines: {:?}",
-            lines
-        );
-    }
-}