` didn't implement `ExactSizeIterator` even when `T1` and `T1` both implements it.
In theory though, if `T1` and `T2` all have exact sizes, the chained iterator should also have exact size.
Is this a missing piece of the standard library? I am looking forward to write a generic macro for those `take_exact_n` functions.
---
### Author:  [@cuviper](https://internals.rust-lang.org/u/cuviper)
#### Post date: [June 15, 2020, 2:22am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/2 "2020-06-15T02:22:53Z")
I believe the main reason it doesn't have this already is that adding two lengths may overflow.
---
### Author:  [@earthengine](https://internals.rust-lang.org/u/earthengine)
#### Post date: [June 15, 2020, 2:34am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/3 "2020-06-15T02:34:19Z")
It is a bit weird since we can create `ExactSizeIterator` with 0 or 1 items, but we couldn't even create one that have exactly 2/3 items unless defining a newtype.
In general though, rust does not prevent this kind of errors: if you access an `Vec` with out-of-bound index you panic. So I don't know why we couldn't do the same here - if the user access the `len` method and it is overflowed, just let it panic.
---
### Author:  [@197g](https://internals.rust-lang.org/u/197g)
#### Post date: [June 15, 2020, 4:35am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/4 "2020-06-15T04:35:43Z")
Interestingly, `a.chain(b).take(usize::MAX)` would be able to be `ExactSizeIterator` but that would require all kinds of specialization. I wonder if the precedent here is enough to introduce this as its own concept?
---
### Author:  [@steffahn](https://internals.rust-lang.org/u/steffahn)
#### Post date: [June 15, 2020, 7:12am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/5 "2020-06-15T07:12:55Z")
You can go a lot further with all kinds of logic like this.
For example this one is always of length 10, no matter what kind of iterator `i` is:
```rust
once(x).chain(i).cycle().take(10)
```
Edit: This example would be particularly hard to archieve based on the type system alone since the _type_ of the iterator that `once` produces also allows for the iterator to be empty (after `next` was called).
---
### Author:  [@scottmcm](https://internals.rust-lang.org/u/scottmcm)
#### Post date: [June 15, 2020, 7:56pm UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/6 "2020-06-15T19:56:13Z")
> [@earthengine](#):
>
> In theory though, if `T1` and `T2` all have exact sizes, the chained iterator should also have exact size.
That's true in theory, where we have infinite-precision integers. But it's not true in practice where we only have `usize`.
This is the very nuanced difference between the rules for `ExactSizeIterator` -- where the len must be exact -- and `TrustedLen` -- where the size hint is allowed to be infinite -- which is why their rules for various adapters are different.
---
### Author:  [@kornel](https://internals.rust-lang.org/u/kornel)
#### Post date: [June 15, 2020, 8:47pm UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/7 "2020-06-15T20:47:27Z")
Is panicking acceptable in such case? Would presence of panic negatively affect optimizations?
Maybe it could saturate? Behaving like `.take(usize::MAX)` shouldn't be limiting for practical applications, since it's already problematic to process this many items on 32-bit machines, and outright impossible on 64-bit.
And it can never overflow with any collections that are actually in memory, because by definition there's not enough address space for two collections to be that big.
---
### Author:  [@cuviper](https://internals.rust-lang.org/u/cuviper)
#### Post date: [June 15, 2020, 8:53pm UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/8 "2020-06-15T20:53:04Z")
I'm going to lazily reply with links... 😅
>
>
> When both iterators of the chain implement ExactSizeIterator, the chain itself should implement that trait as well.
>
---
### Author:  [@mjw](https://internals.rust-lang.org/u/mjw)
#### Post date: [June 16, 2020, 7:38am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/9 "2020-06-16T07:38:19Z")
This response makes me think that Rust would benefit from having a technical FAQ.
---
### Author:  [@scottmcm](https://internals.rust-lang.org/u/scottmcm)
#### Post date: [June 19, 2020, 12:23am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/10 "2020-06-19T00:23:46Z")
Well, I bet a PR to describe this in `Chain`'s docs would be welcome.
---
### Author:  [@system](https://internals.rust-lang.org/u/system)
#### Post date: [September 17, 2020, 12:23am UTC](https://internals.rust-lang.org/t/should-we-implement-exactsizeiterator-for-chain/12556/11 "2020-09-17T00:23:58Z")
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.