Better developer experience when using "log.critical"

[PaulRBerg]

Problem

According to the docs for the Logging API, log.critical terminates the subgraph:

log.critical(fmt: string, args: Array<string>): void – logs a critical message and terminates the Subgraph

The trouble is that when I am using log.critical, the TypeScript compiler isn't able to infer that log.critical will terminate the execution of the function. Example:

export function getChainId(): BInt {
  const context = dataSource.context();
  const chainId = context.get("chainId");
  if (chainId == null) {
    log.critical("[SABLIER] chainId not found in context", []); // <--- PROBLEM HERE
  }
  return chainId.toBigInt();
}

Getting this type error:

'chainId' is possibly 'null'.

Solution

Improve the developer experience when using log.critical that lets the TypeScript compiler infer that the program execution will stop.

[PaulRBerg]

Until a solution is provided for this in the @graphprotocol-graph-ts package, we can use the unreachable global in AssemblyScript:

if (someVar == null) {
  unreachable();
}
return someOtherVar;