React hook for getting a unique identifier for a component

useComponentId.js

import { useRef } from 'react';

let uniqueId = 0;
const getUniqueId = () => uniqueId++;

export function useComponentId() {
  const idRef = useRef(getUniqueId());
  return idRef.current;
}

Going by Dan Abramov's comment here, it sounds like the useRef implementation was on the right track. You can avoid calling getUniqueId() on every render by using a conditional on idRef.current:

let uniqueId = 0;
const getUniqueId = () => uniqueId++;

export function useComponentId() {
  const idRef = useRef(null);
  if (idRef.current === null) {
    idRef.current = getUniqueId()
  }
  return idRef.current;
}

But using useRef for everything else that isn't a component is strange, at least for me.
Is more semantic and clear just using useMemo instead of useRef and on new line if (ref.current === null) {, that is almost the same of useMemo.

@macabeus I think you did not understand my initial post. I posted a use case plus counter and you posted counter, telling its simpler. It does not make any sense. When I subtract the use case from my initial post, then you have posted the same counter I did, except you create the arrow function each time, while I don't - minor disadvantage, but no benefits, so what is the point. The creation of the arrow function has nothing to do with the arrow function being called once in this specific example. Its two different questions.

@TrevorBurnham
Thanks for the link. With the fix, it does make sense. I wanted to either point out an error in the gist (which it is as it seems) or get new insight in something I might not knew there. Got it via side-channel from you :). Conclusion: I guess useMemo is more generic and does argument iteration to check against the dependencies, so checking manually with an simple weak null check and useRef is more efficient. Makes sense :)

@akomm Okay, now I understand your point. Thank you for reply =]

Hey,

There is an alternative implementation with useState hook:

let uniqueId = 0;

export function useComponentId() {
  const [componentId] = useState(() => uniqueId++);
  return componentId;
}

@mbelsky You can do that, but your example will increment the ID on each render call, even thought no new ID is required. You can fix it making the initialValue a calback () => uniqueId++. I'd create a meaningful hook at this point :)

Also see: https://twitter.com/dan_abramov/status/1099842565631819776?s=20 regarding useRef & useState

@akomm I've tested my solution with this demo: https://codesandbox.io/s/determined-goldstine-qj761 A component rerenders there on click event and uniqueId stays same.

How should I change this demo to see behavior that you described?

@mbelsky in your example, its the componentId which is the same, not uniqueId.

@akomm could you explain please what the difference between them and how I can reproduce the issue that you described above?

but your example will increment the ID on each render call, even thought no new ID is required

I'm still not understanding what wrong with this solution

just log the uniqueId to console in useComponentId and see if it is the same? You tell uniqueId is the same and post an example where you render the componentId. :)

@akomm got it, thanks. I've updated originally post & demo. btw there is another interesting issue: uniqueId-s are odd numbers now. I'll try to find a solution for that

@mbelsky

Your proposal worked. Its just a question if you want to increment the id forever on each render or not. The solution would be, as I proposed above, to replace uniqueId++ with () => uniqueId++.

@mbelsky

codesandbox is flaky there, hence your odd numbers. Try it locally, it should work. I used to use codepen, then codesandbox came out and was better. Liked it. But recently it got more and more flaky. A pure bugfeast. Might be there bundler configuration or life reload that messes it up. IF you do some tests, you notice on life reload its rendered twice.

codesandbox is flaky there, hence your odd numbers. Try it locally, it should work

I'll try, thank you!

For anyone landing here using Typescript, it is generally advised to avoid the use of null. It simplifies types, and once you try to stop using it you realize you really don't need it 95% of the time because the JS already gives you undefined to work with.

So here's a slight adaptation:

let uniqueId = 0;
const getUniqueId = () => uniqueId++;

export function useComponentId() {
  const idRef = useRef<number>();
  if (idRef.current === undefined) {
    idRef.current = getUniqueId();
  }
  return idRef.current;
}

Using this with Strict Mode will give you a different value on the 2nd render. I assume this is okay to use otherwise, and it's just some quirk of Strict Mode that produces differing values? I've used it in production w/o apparent issue...

Edit: looks like it's the result of having a harmless side effect: facebook/react#20826

Trying to move away from this Hook in order to be StrictMode compatible (despite it working 100% fine otherwise)... wish there was a React.useComponentName Hook...

The useRef version works in my testing, but why not just set the initial value rather than checking undefined/null and assigning? Am I missing something here?

export function useComponentId() {
  const id = useRef(getUniqueId());
  return id.current;
}

(edited, forgot .current in the return)

@notthatnathan it has actually been explained here.

Oops, missed it. Thanks.

FWIW, after generating the id (I used shortid()) as the initial value (as mentioned above), it doesn't change in my testing. I don't believe the undefined check and reassignment is necessary with refs.

Also, for test snapshot support, you'll want a predictable id, maybe using a different prop from the instance.

const id = useRef(process.env.NODE_ENV === 'test' ? `test-${title}` : shortid());

FWIW, after generating the id (I used shortid()) as the initial value (as mentioned above), it doesn't change in my testing. I don't believe the undefined check and reassignment is necessary with refs.

Why is it not needed and how do the two things relate (I don't know your shortid implementation).

Also, for test snapshot support, you'll want a predictable id, maybe using a different prop from the instance.

const id = useRef(process.env.NODE_ENV === 'test' ? `test-${title}` : shortid());

You add test-related code in a component? And also you change behavior depending on how id is used the outcome might be different in test than prod.

You add test-related code in a component? And also you change behavior depending on how id is used the outcome might be different in test than prod.

Doesn't change behavior, just changes the id. Otherwise every run of your test, you get snapshot updates (new id on mount), which defeats the purpose of snapshots.

Why is it not needed and how do the two things relate (I don't know your shortid implementation).

I'm not sure I understand the question. I'm just saying that this:

  const idRef = useRef(null);
  if (idRef.current === null) {
    idRef.current = getUniqueId()
  }
  return idRef.current;

and this

  const idRef = useRef(getUniqueId());
  return idRef.current;

both return a unique ID that doesn't change on re-render. Which makes sense, refs wouldn't be useful if the initial value changed. Log from the calling component to see what I mean.

(shortid is just an id-generating package my org uses)

updated, typo in the second code example

You add test-related code in a component? And also you change behavior depending on how id is used the outcome might be different in test than prod.

Doesn't change behavior, just changes the id. Otherwise every run of your test, you get snapshot updates (new id on mount), which defeats the purpose of snapshots.

It does. You use different ID generation methods. If one of the different methods generate non-unique ID, it will work in one, but fail in the other or lead to different results, depending on how you use the generated ID.

Sorry. I don't get what you are talking about. You've just changed things replying to my initial question. Why is it null instead of undefined? Also you did not answer why the undefined check is not needed. I don't understand the problem.

In cases where I don't need globally unique IDs, but rather IDs unique per component I'm using this hook:

export function useIdGenerator(): () => number {
  const ref = useRef(0);

  function getId() {
    ref.current += 1;
    return ref.current;
  }

  return getId;
}

On a side note, nanoId is considerably faster than shortId or uniqueId according to their benchmarks.

There is an open issue for this on the react repo.

Here is a more robust implementation that also supports SSR.

Also, for test snapshot support, you'll want a predictable id, maybe using a different prop from the instance.

You shouldn't be using snapshots at all. write proper tests.

For anyone stumbling upon this in 2023 there's React.useId for this very purpose since React v18.