Lately, I've been experimenting more with the async
/await
keywords in JavaScript. I noticed that I sometimes struggle to reconcile the strategies I use with Promises with the way I need to write code in the newer syntax. Most recently, I was playing around with finally
in some try
/catch
blocks and came upon some behavior I didn't expect.
This post assumes a general understanding of how asynchronous JavaScript code works - particularly, how Promises work. (If you're looking for an in-depth explanation of async JS from callbacks to the async/await keywords, there's a pretty good overview on javascript.info - you can also check out Mostafa Gaafar's article for some of the neat features of async/await.)
For context - in the JavaScript codebase I spend a lot of my time in, we've historically dealt with asynchronous actions by using Promises heavily. Generally, this pattern is a lot more familiar to me:
const loadSomething = () => {
return fetchSomeData()
.then(data => doSomethingWith(data))
.catch(error => logAndReport(error))
}
And this is less familiar:
const loadSomething = async () => {
try {
const data = await fetchSomeData()
return doSomethingWith(data)
} catch (error) {
logAndReport(error)
}
}
finally
...?
You'll notice that a finally
callback/block is missing from both of the examples above. I don't use either often in my code, which led me to a misunderstanding (of both, really). Let's dive into the differences between this concept in Promises and in try/catch!
finally
in Promises
When you use the somePromise.then(x).catch(y).finally(z)
pattern, your business logic is generally happening in the then
callback (x
, above - what you want to do once somePromise
has resolved) or in the catch
callback (y
above - returns what you want to pass along in case something goes horribly wrong). You might never have even used finally
in your code - and that's fine.
According to the MDN docs, a finally
callback allows you to execute logic once your Promise has been settled - resolved or rejected - one way or the other. It has absolutely no impact on the value that your promise will resolve to - it doesn't even have access to it. In fact, the documentation states that:
Promise.resolve(2).finally(() => {})
will be resolved with2
.
This means (somewhat counterintuitively) you can sprinkle finally
callbacks liberally throughout your promise chain without changing the final result that it will resolve to:
// Please don't do this 😅
Promise.resolve({ some: 'data' })
.finally(() => { console.log('WHALE HELLO THERE 🐋') })
.then(data => ({ ...data, anAdditional: 'key' }))
.finally(() => { console.log('Looks like we made it past the first step 🙏') })
.then(data => ({ ...data, yetAnother: 'thing added' }))
.finally(() => { console.log("We're done I think 🙌") })
.then(data => {
console.log('Final result:', data)
})
If you run this code, you should see this:
finally
in try/catch blocks
The try/catch/finally pattern has been around for a long time in JavaScript - since version 1.4 (ES3 specification, around 1999). There are a couple of logical parallels I'd drawn between this pattern and how promises are handled:
try
/then
:
This is where our "happy path" logic goes - if nothing breaks, all the action happens here!
catch
:
This is where we end up when things go wrong, and gives us a chance to redeem ourselves 🙏
finally
:
This logic will execute after the try
/then
(and possibly catch
) logic has completed. This code runs no matter what, whether we've encountered an error or not.
The difference here that tripped me up is related to return
statements. If your finally
block does not include a return statement, it has no effect on the return value. However, if you return a value from a finally
block, that value will override all other returns and be the final result of your function. (Check out this example from the docs!)
// This worked as I expected.
const returnFromTryCatch = (someFunction) => {
try {
return someFunction()
} catch (error) {
return `Caught an error: ${error}`
} finally {
// This block has no effect on the return value.
console.log('All done!')
}
}
// This was a surprise to me!
const returnFromFinally = (someFunction) => {
try {
return someFunction()
} catch (error) {
return `Caught an error: ${error}`
} finally {
// Wait... so I'm just swallowing my return and error handling?
return 'All done!'
}
}
This makes sense, but it felt inconsistent to me. My experience with Promises reared its head - why should a finally
block ever be allowed to override the value a function returns?
Finding the Reason
Finally, I pinged my tech lead detailing my annoyance, and he sent me a link to a related StackOverflow discussion. Seeing the ECMAScript specification (emphasis mine) for this behavior helped it settle into place in my brain:
The production
TryStatement : try Block Finally
is evaluated as follows:Let B be the result of evaluating Block.
Let F be the result of evaluating Finally.
If F.type is normal, return B.
Return F.
(It's worth noting the "completion types" according to the ECMAScript spec are "One of normal, break, continue, return, or throw" - I have assumed that a function that does not include a break
, continue
, return
, or throw
keyword qualifies as "normal." Kind of weird semantics there.)
Note on Multiple Return Statements
The code samples in this post do not utilize a single return. I'm not going to get too far into the debate around multiple return statements - I will say that in general, having a single return for longer functions has served me well in the past, but I've found them less helpful in shorter blocks. It probably would have made my life easier in this case, though!