Module word_count 

Pure word/character counting over &str — no document, no store, no threads.

Mirrors djot_to_plain_text and the search matcher’s shape: a cheap primitive over &str with the policy (which counting method) passed in as a parameter, so a host app can count a manuscript of thousands of scenes without importing each one into a document.

Language is not a parameter: UAX #29 word segmentation is not locale-tailored the way case-folding is, and the CJK rule (CountMethod::CjkHybrid) is script-detected per character. A caller that wants “count this Chinese scene per character” selects CjkHybrid; it does not pass a language tag.

Structs

WordCharCounts

The three counts a caller might display. All three are always computed — “characters with spaces” vs “without” is a display choice, not a separate counting mode.

Enums

CountMethod

How words are delimited. Characters are always counted the same way (Unicode scalar values), independent of this choice — see WordCharCounts.

Functions

count

Count words and characters in already-extracted plain text.

count_djot

Extract prose from Djot source, then count. The two-step (strip then count) is kept deliberately simple: djot_to_plain_text carries a pinned contract (table-anchor sentinel, single-\n block joins) and fusing the parse-and-count into one AST walk is the easiest way to accidentally create a third, silently-diverging “what is the text”.