feat: [WIP] write tags

skip-checks:true

4 files changed, 182 insertions, 37 deletions

diff --git a/crates/typst-layout/src/pages/run.rs b/crates/typst-layout/src/pages/run.rs
index 6d2d29da..e9e4e110 100644
--- a/crates/typst-layout/src/pages/run.rs
+++ b/crates/typst-layout/src/pages/run.rs
@@ -185,6 +185,8 @@ fn layout_page_run_impl(
     )?;
 
     // Layouts a single marginal.
+    // TODO: add some sort of tag that indicates the marginals and use it to
+    // mark them as artifacts for PDF/UA.
     let mut layout_marginal = |content: &Option<Content>, area, align| {
         let Some(content) = content else { return Ok(None) };
         let aligned = content.clone().styled(AlignElem::set_alignment(align));
diff --git a/crates/typst-pdf/src/convert.rs b/crates/typst-pdf/src/convert.rs
index 94925756..cd165b12 100644
--- a/crates/typst-pdf/src/convert.rs
+++ b/crates/typst-pdf/src/convert.rs
@@ -10,11 +10,11 @@ use krilla::error::KrillaError;
 use krilla::geom::PathBuilder;
 use krilla::page::{PageLabel, PageSettings};
 use krilla::surface::Surface;
-use krilla::tagging::{Node, SpanTag, Tag, TagGroup, TagTree};
+use krilla::tagging::{ArtifactType, ContentTag, Node};
 use krilla::{Document, SerializeSettings};
 use krilla_svg::render_svg_glyph;
 use typst_library::diag::{bail, error, SourceDiagnostic, SourceResult};
-use typst_library::foundations::{NativeElement, Repr, StyleChain};
+use typst_library::foundations::{NativeElement, Repr};
 use typst_library::introspection::{self, Location};
 use typst_library::layout::{
     Abs, Frame, FrameItem, GroupItem, PagedDocument, Size, Transform,
@@ -31,6 +31,7 @@ use crate::metadata::build_metadata;
 use crate::outline::build_outline;
 use crate::page::PageLabelExt;
 use crate::shape::handle_shape;
+use crate::tags::{handle_close_tag, handle_open_tag, Tags};
 use crate::text::handle_text;
 use crate::util::{convert_path, display_font, AbsExt, TransformExt};
 use crate::PdfOptions;
@@ -49,6 +50,8 @@ pub fn convert(
         xmp_metadata: true,
         cmyk_profile: None,
         configuration: config,
+        // TODO: Should we just set this to false? If set to `false` this will
+        // automatically be enabled if the `UA1` validator is used.
         enable_tagging: true,
         render_svg_glyph_fn: render_svg_glyph,
     };
@@ -70,12 +73,7 @@ pub fn convert(
 
     document.set_outline(build_outline(&gc));
     document.set_metadata(build_metadata(&gc));
-
-    let mut tag_tree = TagTree::new();
-    for tag in gc.tags.drain(..) {
-        tag_tree.push(tag);
-    }
-    document.set_tag_tree(tag_tree);
+    document.set_tag_tree(gc.tags.take_tree());
 
     finish(document, gc, options.standards.config)
 }
@@ -115,6 +113,19 @@ fn convert_pages(gc: &mut GlobalContext, document: &mut Document) -> SourceResul
             let mut surface = page.surface();
             let mut fc = FrameContext::new(typst_page.frame.size());
 
+            // Marked-content may not cross page boundaries: reopen tag
+            // that was closed at the end of the last page.
+            if let Some((_, _, nodes)) = gc.tags.stack.last_mut() {
+                let tag = if gc.tags.in_artifact {
+                    ContentTag::Artifact(ArtifactType::Other)
+                } else {
+                    ContentTag::Other
+                };
+                // TODO: somehow avoid empty marked-content sequences
+                let id = surface.start_tagged(tag);
+                nodes.push(Node::Leaf(id));
+            }
+
             handle_frame(
                 &mut fc,
                 &typst_page.frame,
@@ -123,6 +134,11 @@ fn convert_pages(gc: &mut GlobalContext, document: &mut Document) -> SourceResul
                 gc,
             )?;
 
+            // Marked-content may not cross page boundaries: close open tag.
+            if !gc.tags.stack.is_empty() {
+                surface.end_tagged();
+            }
+
             surface.finish();
 
             for annotation in fc.annotations {
@@ -235,8 +251,8 @@ pub(crate) struct GlobalContext<'a> {
     /// The languages used throughout the document.
     pub(crate) languages: BTreeMap<Lang, usize>,
     pub(crate) page_index_converter: PageIndexConverter,
-    pub(crate) tag_stack: Vec<Location>,
-    pub(crate) tags: Vec<Node>,
+    /// Tagged PDF context.
+    pub(crate) tags: Tags,
 }
 
 impl<'a> GlobalContext<'a> {
@@ -256,8 +272,8 @@ impl<'a> GlobalContext<'a> {
             image_spans: HashSet::new(),
             languages: BTreeMap::new(),
             page_index_converter,
-            tag_stack: Vec::new(),
-            tags: Vec::new(),
+
+            tags: Tags::new(),
         }
     }
 }
@@ -294,33 +310,10 @@ pub(crate) fn handle_frame(
             }
             FrameItem::Link(d, s) => handle_link(fc, gc, d, *s),
             FrameItem::Tag(introspection::Tag::Start(elem)) => {
-                let Some(heading) = elem.to_packed::<HeadingElem>() else { continue };
-                let Some(loc) = heading.location() else { continue };
-
-                let level = heading.resolve_level(StyleChain::default());
-                let name = heading.body.plain_text().to_string();
-                let heading_id = surface
-                    .start_tagged(krilla::tagging::ContentTag::Span(SpanTag::empty()));
-                let tag = match level.get() {
-                    1 => Tag::H1(Some(name)),
-                    2 => Tag::H2(Some(name)),
-                    3 => Tag::H3(Some(name)),
-                    4 => Tag::H4(Some(name)),
-                    5 => Tag::H5(Some(name)),
-                    _ => Tag::H6(Some(name)),
-                };
-                let mut tag_group = TagGroup::new(tag);
-                tag_group.push(Node::Leaf(heading_id));
-                gc.tags.push(Node::Group(tag_group));
-
-                gc.tag_stack.push(loc);
+                handle_open_tag(gc, surface, elem)
             }
             FrameItem::Tag(introspection::Tag::End(loc, _)) => {
-                // FIXME: support or split up content tags that span multiple pages
-                if gc.tag_stack.last() == Some(loc) {
-                    surface.end_tagged();
-                    gc.tag_stack.pop();
-                }
+                handle_close_tag(gc, surface, loc);
             }
         }
 
diff --git a/crates/typst-pdf/src/lib.rs b/crates/typst-pdf/src/lib.rs
index 88c6ee55..c3835d24 100644
--- a/crates/typst-pdf/src/lib.rs
+++ b/crates/typst-pdf/src/lib.rs
@@ -9,6 +9,7 @@ mod outline;
 mod page;
 mod paint;
 mod shape;
+mod tags;
 mod text;
 mod util;
 
diff --git a/crates/typst-pdf/src/tags.rs b/crates/typst-pdf/src/tags.rs
new file mode 100644
index 00000000..70792dfe
--- /dev/null
+++ b/crates/typst-pdf/src/tags.rs
@@ -0,0 +1,149 @@
+use krilla::surface::Surface;
+use krilla::tagging::{ContentTag, Node, Tag, TagGroup, TagTree};
+use typst_library::foundations::{Content, StyleChain};
+use typst_library::introspection::Location;
+use typst_library::model::{HeadingElem, OutlineElem, OutlineEntry};
+
+use crate::convert::GlobalContext;
+
+pub(crate) struct Tags {
+    /// The intermediary stack of nested tag groups.
+    pub(crate) stack: Vec<(Location, Tag, Vec<Node>)>,
+    pub(crate) in_artifact: bool,
+
+    /// The output.
+    pub(crate) tree: TagTree,
+}
+
+impl Tags {
+    pub(crate) fn new() -> Self {
+        Self {
+            stack: Vec::new(),
+            in_artifact: false,
+            tree: TagTree::new(),
+        }
+    }
+
+    pub(crate) fn take_tree(&mut self) -> TagTree {
+        std::mem::take(&mut self.tree)
+    }
+
+    pub(crate) fn context_supports(&self, tag: &Tag) -> bool {
+        let Some((_, parent, _)) = self.stack.last() else { return true };
+
+        use Tag::*;
+
+        match parent {
+            Part => true,
+            Article => !matches!(tag, Article),
+            Section => true,
+            BlockQuote => todo!(),
+            Caption => todo!(),
+            TOC => matches!(tag, TOC | TOCI),
+            // TODO: NonStruct is allowed to but (currently?) not supported by krilla
+            TOCI => matches!(tag, TOC | Lbl | Reference | P),
+            Index => todo!(),
+            P => todo!(),
+            H1(_) => todo!(),
+            H2(_) => todo!(),
+            H3(_) => todo!(),
+            H4(_) => todo!(),
+            H5(_) => todo!(),
+            H6(_) => todo!(),
+            L(_list_numbering) => todo!(),
+            LI => todo!(),
+            Lbl => todo!(),
+            LBody => todo!(),
+            Table => todo!(),
+            TR => todo!(),
+            TH(_table_header_scope) => todo!(),
+            TD => todo!(),
+            THead => todo!(),
+            TBody => todo!(),
+            TFoot => todo!(),
+            InlineQuote => todo!(),
+            Note => todo!(),
+            Reference => todo!(),
+            BibEntry => todo!(),
+            Code => todo!(),
+            Link => todo!(),
+            Annot => todo!(),
+            Figure(_) => todo!(),
+            Formula(_) => todo!(),
+            Datetime => todo!(),
+            Terms => todo!(),
+            Title => todo!(),
+        }
+    }
+}
+
+pub(crate) fn handle_open_tag(
+    gc: &mut GlobalContext,
+    surface: &mut Surface,
+    elem: &Content,
+) {
+    if gc.tags.in_artifact {
+        return;
+    }
+
+    let Some(loc) = elem.location() else { return };
+
+    let tag = if let Some(heading) = elem.to_packed::<HeadingElem>() {
+        let level = heading.resolve_level(StyleChain::default());
+        let name = heading.body.plain_text().to_string();
+        match level.get() {
+            1 => Tag::H1(Some(name)),
+            2 => Tag::H2(Some(name)),
+            3 => Tag::H3(Some(name)),
+            4 => Tag::H4(Some(name)),
+            5 => Tag::H5(Some(name)),
+            // TODO: when targeting PDF 2.0 headings `> 6` are supported
+            _ => Tag::H6(Some(name)),
+        }
+    } else if let Some(_) = elem.to_packed::<OutlineElem>() {
+        Tag::TOC
+    } else if let Some(_outline_entry) = elem.to_packed::<OutlineEntry>() {
+        Tag::TOCI
+    } else {
+        return;
+    };
+
+    if !gc.tags.context_supports(&tag) {
+        // TODO: error or warning?
+    }
+
+    // close previous marked-content and open a nested tag.
+    if !gc.tags.stack.is_empty() {
+        surface.end_tagged();
+    }
+    let content_id = surface.start_tagged(krilla::tagging::ContentTag::Other);
+
+    gc.tags.stack.push((loc, tag, vec![Node::Leaf(content_id)]));
+}
+
+pub(crate) fn handle_close_tag(
+    gc: &mut GlobalContext,
+    surface: &mut Surface,
+    loc: &Location,
+) {
+    let Some((_, tag, nodes)) = gc.tags.stack.pop_if(|(l, ..)| l == loc) else {
+        return;
+    };
+    // TODO: contstruct group directly from nodes
+    let mut tag_group = TagGroup::new(tag);
+    for node in nodes {
+        tag_group.push(node);
+    }
+
+    surface.end_tagged();
+
+    if let Some((_, _, parent_nodes)) = gc.tags.stack.last_mut() {
+        parent_nodes.push(Node::Group(tag_group));
+
+        // TODO: somehow avoid empty marked-content sequences
+        let id = surface.start_tagged(ContentTag::Other);
+        parent_nodes.push(Node::Leaf(id));
+    } else {
+        gc.tags.tree.push(Node::Group(tag_group));
+    }
+}