execa/lib/return/final-error.js

Process execution for humans

// When the subprocess fails, this is the error instance being returned.
// If another error instance is being thrown, it is kept as `error.cause`.
export const getFinalError = (originalError, message, isSync) => {
	const ErrorClass = isSync ? ExecaSyncError : ExecaError;
	const options = originalError instanceof DiscardedError ? {} : {cause: originalError};
	return new ErrorClass(message, options);
};

// Indicates that the error is used only to interrupt control flow, but not in the return value
export class DiscardedError extends Error {}

// Proper way to set `error.name`: it should be inherited and non-enumerable
const setErrorName = (ErrorClass, value) => {
	Object.defineProperties(ErrorClass.prototype, {
		name: {
			value,
			writable: true,
			enumerable: false,
			configurable: true,
		},
		[execaErrorSymbol]: {
			value: true,
			writable: false,
			enumerable: false,
			configurable: false,
		},
	});
};

// Unlike `instanceof`, this works across realms
export const isExecaError = error => isErrorInstance(error) && execaErrorSymbol in error;

const execaErrorSymbol = Symbol('isExecaError');

export const isErrorInstance = value => Object.prototype.toString.call(value) === '[object Error]';

// We use two different Error classes for async/sync methods since they have slightly different shape and types
export class ExecaError extends Error {}
// eslint-disable-next-line unicorn/no-top-level-side-effects -- `error.name` must be set as soon as the class is defined
setErrorName(ExecaError, ExecaError.name);

export class ExecaSyncError extends Error {}
// eslint-disable-next-line unicorn/no-top-level-side-effects -- `error.name` must be set as soon as the class is defined
setErrorName(ExecaSyncError, ExecaSyncError.name);