TextParseConfig

TextParseConfig (crates/show/src/parse.rs) controls how extracted text is split into cues during import. It is serializable and can be passed to import_pdf_show from the frontend (camelCase field names in JSON).

Fields

Field	Type	Default	Description
maxCueLength	number	140	Target maximum cue length in characters. Longer passages are split.
minCueLength	number	40	Fragments shorter than this are merged into a neighboring cue.
splitPunctuation	string[]	[";", ":", "—", ","]	Split points, in priority order, for over-long sentences.

Rust constants: DEFAULT_MAX_CUE_LENGTH, DEFAULT_MIN_CUE_LENGTH, DEFAULT_SPLIT_PUNCTUATION.

Splitting algorithm

1. Sentences are split on ., !, and ? (the terminator stays with the sentence it ends).
2. Long sentences (over maxCueLength) are split at the strongest nearby punctuation. The splitter searches right-to-left within the window and prefers higher-priority punctuation, so a comma is used only when no semicolon, colon, or em-dash is available.
3. If no punctuation is available, it splits at the last whitespace in the window, so a word is never cut in two. A sentence with no whitespace at all (one giant word) is kept whole.
4. Small fragments (under minCueLength) are merged into a neighboring cue, so imports don't contain stunted one-word cues.

These steps are implemented by split_cue_text, which composes split_sentences, split_long, find_split_point, and merge_small_into_neighbours.

Example

import { importPdfShow } from "@/lib/tauri";

// (The frontend wrapper currently passes no config; call invoke directly to
// override the defaults.)
await invoke("import_pdf_show", {
  config: {
    maxCueLength: 100,
    minCueLength: 30,
    splitPunctuation: [";", ":", "—", ","],
  },
});

Notes

• Lengths are measured in characters, not bytes, so multi-byte characters split correctly.
• Setting minCueLength to 0 disables merging.
• Removing all splitPunctuation still yields word-boundary splitting.