Struct ActualTextIndex 

pub struct ActualTextIndex {
    pub mcid_to_actual_text: HashMap<(McidScope, u32), Arc<str>>,
    pub covered_mcids: HashSet<(McidScope, u32)>,
    pub suppress_only: HashSet<(McidScope, u32)>,
}

Pre-computed index of structure-element /ActualText replacements resolved over the entire structure tree.

Per ISO 32000-1:2008 §14.9.4, a structure element may carry an /ActualText entry that replaces the entire content of its descendants for text-extraction purposes. The replacement scope is the subtree rooted at the bearing element. Nested ActualText declarations override their ancestors for the spans they cover.

Data model: keyed by (McidScope, mcid)

Per ISO 32000-1:2008 §14.7.4.3, MCIDs are NOT globally unique — they are unique only within a single content stream (page content stream, Form XObject content stream, or Tiling Pattern content stream). Two Form XObjects on the same page can both emit MCID 0 referring to distinct structure elements. Keying by (McidScope, mcid) keeps every namespace separate.

For page-scoped MCIDs, McidScope::Page(page_index) plays exactly the role the previous (page, mcid) keying did; the cross-page emit-once rule still applies via Self::suppress_only. Form- and Pattern-scoped MCIDs are intrinsically per-stream — every covered MCID emits at its anchor (the form/pattern’s content stream covers a single namespace).

Emission model: mcid-driven, consecutive-run dedup

Consumers iterate per-page MCIDs in structure-tree order and emit mcid_to_actual_text[(page, mcid)] whenever the replacement changes between consecutive covered MCIDs. A subtree covering [5, 6, 7] with one replacement string therefore emits ONE replacement; a subtree where a nested ActualText overrides MCID 6 gives THREE emissions (outer at 5, inner at 6, outer at 7), which is the correct inner-wins shape. Run-end is detected at the next non-covered MCID or at a covered MCID with a different replacement.

Visibility (OCG filtering)

When a covered MCID has been filtered out by an excluded OCG layer, the consumer skips emission for that MCID but does NOT break the consecutive-run dedup — so partial OCG visibility still emits the replacement at the first visible covered MCID in the run.

Fields

mcid_to_actual_text: HashMap<(McidScope, u32), Arc<str>>

Map from (McidScope, mcid) → its innermost replacement text.

The “innermost” resolution honours nesting: when a descendant element redeclares /ActualText, the descendant’s text replaces the ancestor’s for (scope, mcid) keys in the inner subtree.

covered_mcids: HashSet<(McidScope, u32)>

(McidScope, mcid) pairs whose raw glyph spans must be suppressed during assembly.

Every key in Self::mcid_to_actual_text is present here, and keys in Self::suppress_only are present here too. Suppression prevents duplicate output: the replacement is emitted via the mcid-driven walk, and raw glyphs for the same (scope, mcid) are dropped.

suppress_only: HashSet<(McidScope, u32)>

(McidScope, mcid) pairs that are covered (raw glyphs suppressed) but must NOT emit a replacement.

For page-scoped subtrees, this encodes “emit-once on the first page” semantics: the bearing element’s first-page (Page(p), mcid) entries land in Self::mcid_to_actual_text; every other page’s (Page(p), mcid) entries land here so the raw glyphs are still suppressed but no second emission fires. Form- and Pattern- scoped subtrees do not populate this set: each Form / Pattern covers a single namespace.

Implementations

impl ActualTextIndex

pub fn new() -> Self

Construct an empty index.

pub fn is_empty(&self) -> bool

Returns true when no ActualText scopes were discovered.

Trait Implementations

impl Clone for ActualTextIndex

fn clone(&self) -> ActualTextIndex

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ActualTextIndex

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for ActualTextIndex

fn default() -> ActualTextIndex

Returns the “default value” for a type.