try catch wrapper for functional typescript programming
npm install ts-try


tryF() is a simple wrapper for code possibly throwing exceptions, based on the idea of scala.util.Try.
Given a function (that could throw), it returns either the result of the function, or the error thrown (see examples for details).
Same for async-function, i.e.: basically promises.
Given the general idea of pure functions, every function should return a value, pure functions should not interrupt the execution by throwing an error.
Functional programming languages implement this by some specific type (different programming languages use
different names, with sometimes different semantics, i.e. scala also has Either).
A very tangible advantage is, when using statically typed languages, the function-signature indicates that a function could fail.
There's no way (in typescript) to indicate that a function might throw an error. At least not checked by the compiler.
Leveraging typescript's union-type it's easy to bring this concept to JavaScript, since functions may now e.g. return the
original value or an error.
``typescript`
function getDataFromApi(id: number): Response | Error { }
the calling code now needs to check if the result is an instance of Error or the real result. But typescript forces you
to check this, so there's no way you can miss that.
A wrapper for types. If a function might fail, the return type may be Try. E is optional, if you want to further
specify the type of the error.
Instead of
`typescript`
function getDataFromApi(id: number): Response | Error { }
one can write
`typescript`
function getDataFromApi(id: number): Try
or even
`typescript`
function getDataFromApi(id: number): Try
Please be aware that no runtime error-type-checking happens.
trivial typeguard to dry up code a little bit, given a value const myValue: Try
`typescript `
if(myValue instanceof Error) { / error handling / }
one can also write
`typescript `
if(isError(myValue)) { / error handling / }
The actual wrapper. See examples.
The test-file also shows all use-cases, but let me add some more context here.
This is what you'll want to use to wrap third-party code — for your own code you'll want to avoid throw completely return new Error()
(and use with function-return-type Try), thus you won't need tryF at all.
`typescript
const myData: Try
if(isError(myData)) {
// no way to recover here, let the calling function know there's an error
return myData;
}
`
$3
Combined with async/await, passing promises to tryF is rather neat. This example wraps request-promise-native,simple: true
which rejects for non-2xx-status-codes (if is set — which is the default).
`typescript
const result: Try
if(isError(result)) {
// error-handling here,
return some-sane-default;
}
// use result here. Typescript "knows" it's a Response here
`
Let's break it down:
`typescript`
const requestP: Promise
const resultP: Promise
const result: Try
Compare this to the code without Try, but using await, it's so ugly, it makes me sad:
`typescript
let response: Response; // uag, "let"!!
try {
// awaiting a promise that rejects requires try/catch
response = await rp("http://example.com");
} catch (err) {
// error-handling here,
return some-sane-default;
}
// use result here
`
One can also wrap async blocks. I think you should not wrap giant lambdas in tryF, but refactor your code to justtryF
wrap the necessary calls/promises in . It's the very same reason you should narrow down try-blocks to the bare minimum.
However:
`typescript
const result: Try
const a = getA();
const b = getB();
// this could of course throw
const data = JSON.parse(a);
const body = await rp("http://example.com");
return transform(body);
});
`
I didn't find a module on npmjs which does exactly what this does, but there are still alternatives.
Of course it's also an option to do this manually. For promises it's basically:
`typescript`
promise.then(x => x + x).catch(err => err);
for try-catch, you probably end up with almost the same wrapper-function.
fp-ts emulates Either from
Scala. Either wraps values into objects. Thus it has
more convenient access (via isLeft() etc.), but involves runtime overhead. tryF/Try is compile-time overhead only.
Besides that, Try is fixed to errors, Either allows generic types for "left" (which is the "failure"-case by convention).Try
When programming scala, there is no is always better then Either or vice versa, it depends on the use case.
I also played around with type Try