TypeScript (AssemblyScript)

.ts files build with AssemblyScript 0.28.20 by default (--runtime stub -O3). Modules are a few KB and a small call costs around 12k gas. The pipeline is the same as ndsr compile: for the same source the produced module.wasm is byte-identical.

import { hive, storage } from 'hivekit'

function increment(input: string): string {
  const cur = storage.get("count");
  const next = (cur.length == 0 ? 0 : I64.parseInt(cur)) + 1;
  storage.set("count", next.toString());
  hive.emit("incremented", "{\"n\":" + next.toString() + "}");
  return next.toString();
}

function fingerprint(input: string): string {
  return hive.hash(input);   // keccak256, "0x…"
}

hive.define("increment", increment);
hive.define("fingerprint", fingerprint);

Rules

Handlers are (input: string): string: a top-level function or a non-capturing arrow (input: string): string => ….
hive.define("name", handler): the name must be a string literal matching [A-Za-z_][A-Za-z0-9_]{0,63}.
One source file. The only import allowed is import { … } from 'hivekit' (for editors; it is stripped at build time). Put the import on its own line, without a trailing comment.
There is no built-in JSON: build strings by hand (as above) or vendor a parser.
Rejected before compiling, with file:line: async/await/Promise, any, unknown, undefined, try/catch, JSON, other imports, require, Date.now/Math.random/new Date(), console, eval, delete, generators, Symbol. Closures that capture locals are not supported.
Identifiers containing __hive_ or __alloc are reserved.

Need any of those? Build the same file with --target js and it runs in the JavaScript engine instead (JavaScript & TypeScript (engine)), at a much higher gas cost.

API

Call	Meaning
`hive.define(name, fn)`	export `fn` as `name`
`hive.call(address, fn, input): string`	call another module; aborts this call if it fails
`hive.tryCall(address, fn, input): string \| null`	same, `null` on failure (callee state reverted)
`hive.tryCallRaw(…): i64`	raw packed result or error code
`hive.emit(name, dataJson)`	record an event (JSON string, integers only)
`hive.log(msg)`	node-local debug log
`hive.hash(data): string`	keccak256 hex of the UTF-8 bytes
`hive.fail(msg)`	fail the call; state changes and events are reverted
`storage.get(key): string`	this module’s value (`""` if unset)
`storage.set(key, value)`	write (`""` deletes)
`storage.del(key)`	delete

A module that uses only hive.hash, hive.emit, hive.log and hive.fail imports no storage functions and can be a committee worker. For example the fingerprint-only module above (without increment) imports only env.abort and crypto.hash, and fingerprint("hello") returns 0x1c8aff950685c2ed4bc3174f3472287b56d9517b9c948127319a09a7a36deac8 for 6,765 gas.

Build and run

hivec build counter.ts                     # → dist/counter.hbc
hivec inspect dist/counter.hbc
hivec run dist/counter.hbc increment --data-dir .node
hivec functions counter.ts

You can also compile AssemblyScript with the runtime itself: ndsr compile counter.ts -o counter.hbc (needs Node.js; installs the pinned compiler on first use). ndsr compile does not accept the import … from 'hivekit' line: remove it for that path. Its manifest names a different compiler (ndsr-asc/[email protected]), so the address differs from the hivec build.