There are differences between setInterval
and setTimeout
that you may not want to lose by always restarting your timer when the component re-renders. This fiddle shows the difference in drift between the two when other code is also running. (On older browsers/machines—like from when I originally answered this question—you don't even need to simulate a large calculation to see a significant drift begin to occur after only a few seconds.)
Referring now to your answer, Marco, the use of setInterval
is totally lost because effects without conditions dispose and re-run every time the component re-renders. So in your first example, the use of the current
dependency causes that effect to dispose and re-run every time the current
changes (every time the interval runs). The second one does the same thing, but actually every time any state changes (causing a re-render), which could lead to some unexpected behavior. The only reason that one works is because next()
causes a state change.
Considering the fact that you are probably not concerned with exact timing, is is cleanest to use setTimeout
in a simple fashion, using the current
and auto
vars as dependencies. So to re-state part of your answer, do this:
useEffect(
() => {
if (!auto) return;
const interval = setTimeout(_ => {
next();
}, autoInterval);
return _ => clearTimeout(interval);
},
[auto, current]
);
Generically, for those just reading this answer and want a way to do a simple timer, here is a version that doesn't take into account the OP's original code, nor their need for a way to start and stop the timer independently:
const [counter, setCounter] = useState(0);
useEffect(
() => {
const id= setTimeout(() => {
setCounter(counter + 1);
// You could also do `setCounter((count) => count + 1)` instead.
// If you did that, then you wouldn't need the dependency
// array argument to this `useEffect` call.
}, 1000);
return () => {
clearTimeout(id);
};
},
[counter],
);
However, you may be wondering how to use a more exact interval, given the fact that setTimeout
can drift more than setInterval
. Here is one method, again, generic without using the OP's code:
// Using refs:
const [counter, setCounter] = useState(30);
const r = useRef(null);
r.current = { counter, setCounter };
useEffect(
() => {
const id = setInterval(() => {
r.current.setCounter(r.current.counter + 1);
}, 1000);
return () => {
clearInterval(id);
};
},
[] // empty dependency array
);
// Using the function version of `setCounter` is cleaner:
const [counter, setCounter] = useState(30);
useEffect(
() => {
const id = setInterval(() => {
setCounter((count) => count + 1);
}, 1000);
return () => {
clearInterval(id);
};
},
[] // empty dependency array
);
Here is what is going on above:
(first example, using refs): To get setInterval
's callback to always refer to the currently acceptable version of setCounter
we need some mutable state. React gives us this with useRef
. The useRef
function will return an object that has a current
property. We can then set that property (which will happen every time the component re-renders) to the current versions of counter
and setCounter
.
(second example, using functional setCounter
): Same idea as the first, except that when we use the function version of setCounter
, we will have access to the current version of the count as the first argument to the function. No need to use a ref to keep things up to date.
(both examples, continued): Then, to keep the interval from being disposed of on each render, we add an empty dependency array as the second argument to useEffect
. The interval will still be cleared when the component is unmounted.
Note: I used to like using ["once"]
as my dependency array to indicate that I am forcing this effect to be set up only once. It was nice for readability at the time, but I no longer use it for two reasons. First, hooks are more widely understood these days and we have seen the empty array all over the place. Second, it clashes with the very popular "rule of hooks" linter which is quite strict about what goes in the dependency array.
So applying what we know to the OP's original question, you could use setInterval
for a less-likely-to-drift slideshow like this:
// ... OP's implementation code including `autoInterval`,
// `auto`, and `next` goes above here ...
const r = useRef(null);
r.current = { next };
useEffect(
() => {
if (!auto) return;
const id = setInterval(() => {
r.current.next();
}, autoInterval);
return () => {
clearInterval(id);
};
},
[auto]
);