tree-sitter-recipe

Tree-sitter grammar for .recipe files — a digital notation for pharmacological prescriptions in the Dutch/Latin KNMP/FNA tradition.

What it parses

Three section markers (case-insensitive):

| Marker | Meaning | Content | | ------ | ----------- | ------------------------------------ | | R/ | recipe | active ingredient(s), form, strength | | Da/ | da/dispense | quantity to dispense | | S/ | signa | directions to the patient |

Plus Latin abbreviations, modern frequencies (1 dd, 3 dd), numbers with Dutch (0,1) or international (0.1) decimals, and a four-tier comment system.

Example

/**
 * Patient: 45yo female, chronic URTI
 * Allergies: none known
 */
R/ claritromycin tablet 500mg
Da/ 14 tablets
S/ b.d.d. 1 tablet p.c.

#! empirical coverage for suspected CAP
R/ amoxicilline 500mg
Da/ 21 caps
S/ 3 dd 1 caps

# pharmacist: check interaction table before dispensing

Install

bun install        # or: npm install
bun test           # or: npm test

Development

npx tree-sitter generate                   # regenerate parser after grammar changes
npx tree-sitter test                       # run corpus tests
npx tree-sitter parse file.recipe          # inspect parse tree
npx tree-sitter highlight file.recipe      # visualize syntax highlighting
npx tree-sitter build --wasm               # build WASM for playground

Comment system

| Syntax | Purpose | AST role | | ---------- | ----------------- | -------------------- | | # … | line comment | extra (no structure) | | #! … | doc line comment | extra (sibling node) | | /* … */ | block comment | extra (no structure) | | /** … */ | doc block comment | extra (sibling node) |

All four live in extras, so they parse anywhere without affecting structure. Doc variants (#!, /**) are intended for tooling (LSP, formatter, docgen) to associate them with the next section by source position.

Supported abbreviations

Frequency: d.d., b.d.d., t.d.d., q.i.d., s.d.d. plus modern N dd

Dutch prose frequency: 3 keer per dag, driemaal daags, om de 8 uur (structured into count/every/period fields)

Timing: a.c., p.c., a.n., m. et v.

Route: i.m., i.v., s.c., subling., o.d., o.s., a.u.e., u.e.

Compounding: ad (fill-to), dtd / d.t.d. (dispense-this-dose)

Dispensing/warning: d.s.p., u.c., z.n., d.c.p., d.s. monit., d.c. prohib., d.i.m.m., CITO, PIM

Forms: aq. pur., collut., lin., supp., gtt aur., gtt nas., gtt ophth.

Conditional: si nec. sit

Not yet supported

• International abbreviations (Rx, Rp, sig., prn, qd, qid, bid, tid, hs, stat) — reserved for #!dialect international pragma.
• iter N x repeat directive — dropped due to lexical conflict with x appearing in signa phrases (2 x daags).
• ISMP error-prone-abbreviation linting — future tooling, not a grammar concern.
• Nested language injection inside doc comments — reserved for a later release.

License

MIT © Kaj Kowalski