Node

npm install laterite

laterite on npm is the same clean-room engine as the Python wheel, with a born-typed arrow-js decode. The verbs match the shared vocabulary — read, validate, buildAgs4 — and, like Python, the read handle chains.

Validate

import { validate } from "laterite";

const report = validate("delivery.ags", { warnings: true });
console.log(report.ok, report.count, report.dictVersion, report.resolution);
// true 0 "4.1.1" "exact"

validate(path) runs the numbered-rules engine and returns a Report. ok is the headline verdict (zero findings), count the number of findings, and dictVersion / resolution tell you which AGS edition the rules came from — selected automatically from the file's TRAN_AGS, never passed in. Group them by rule (the same shape as Python's by_rule()):

for (const [rule, findings] of Object.entries(report.byRule())) {
  console.log(rule, findings.length);
}

Read a group as typed data

read() returns an Ags4File; .table(code) hands back an arrow-js Table whose columns are already cast to their AGS4 types.

import { read } from "laterite";

const file = read("delivery.ags");
file.groups; // ["PROJ", "LOCA", "SAMP", …]
const loca = file.table("LOCA"); // arrow-js Table, born-typed
console.log(loca.numRows, file.headings("LOCA"));

The chain — validate · fix · diff · certify

The read handle is fluent, so a whole workflow stays on one object:

const file = read("delivery.ags").validate({ warnings: true }); // returns the file; verdict on .report

if (!file.report.ok) {
  const repaired = file.fix(); // returns a NEW repaired Ags4File
  const delta = file.diff(repaired); // what changed, group by group
  repaired.save("clean.ags");
}

• .fix(opts) → a new repaired Ags4File (the fix log on .fixReport).
• .diff(other) → a RevisionDelta of cell/row/group changes.
• .certify(path?) → mints the .ags.idx validity certificate beside the file (after a clean .validate()).

Query with SQL

.sql() loads the optional @duckdb/node-api peer on demand (the addon itself is DuckDB-free) and runs SQL across the file's groups:

const rows = await file.sql("SELECT loca_id FROM LOCA WHERE loca_gl < 0");

Produce AGS4

import { buildAgs4 } from "laterite";

const out = buildAgs4({ PROJ: [{ PROJ_ID: "P1" }] /* … */ });
out.save("out.ags"); // or out.text / out.bytes

Same engine, proven

Node's findings are asserted byte-identical to Python, wasm, and DuckDB by the cross-surface compliance harness — see One engine, every stack.