sync: improve the docs for recv_many

[SwishSwushPow]

Trying to explain a bit better when recv_many will return and why in some cases the returned number of received messages is below the input parameter limit.

Motivation

I stumbled upon recv_many for the first time ever in a bit of code while reviewing a PR. I took some issue with how the function was called (inside a while loop) and had a look at the docs to understand why that would be necessary. When reading through the docs it seemed to me as if there was seemingly no need for a loop because recv_many would apparently wait for new messages if none were available, only returning when the limit would be reached or the channel would be closed. I gave this a try in a smaller code example and had to realize that this was not the case. Looking at the source I found that recv_many would only wait for the first message, but would return early if the channel queue had no messages available after that, even when not being closed. At this point I felt that the docs were a bit unclear in that regard.

Solution

By describing the behavior a bit more explicitly I am hoping to improve the docs and avoid misunderstandings like mine in the future.

[Darksonn]

If I'm skimming this, I'm going to see "return when the number of received messages reaches limit" and I will get the wrong impression. I agree the current docs are bad, but I don't think this wording is good either.

[SwishSwushPow]

Could you please give me a hint in which way the current wording isn't up to par? Otherwise I'm not sure in which direction to go with this. 😅

[Darksonn]

How about something like this?

Receives between 1 and `limit` messages into `buffer`.

This method is intended as a performance optimization for cases where
you wish to process messages in batches. The method may return less than
`limit` messages, so `recv_many` can *not* be used to receive a fixed
number of messages in one call.

This method waits until at least one message is available, and then
receives as many messages as it can into `buffer` efficiently. The
number of messages added to `buffer` is returned, and this number will
often be less than `limit` even if the channel is still open.

For non-zero values of `limit`, this method will never return `0` unless
the channel has been closed and there are no remaining messages in the
channel's queue. This indicates that no further values can ever be
received from this `Receiver`. The channel is closed when all senders
have been dropped, or when [`close`] is called.

Note that if [`close`] is called, but there are still outstanding
[`Permits`] from before it was closed, the channel is not considered
closed by `recv_many` until the permits are used or dropped.

The capacity of `buffer` is increased as needed.

[SwishSwushPow]

Yup that looks also good. 👍 Just two things:

Does this

Receives between 1 and limit messages into buffer.

contradict this?

For non-zero values of limit, this method will never return 0 unless the channel has been closed and there are no remaining messages in the channel's queue.

Maybe it makes sense to word that first sentence differently and say

Receives between 0 and limit messages into buffer.

---

And my second point: I would like to be a bit more precise here.

..., and then receives as many messages as it can ...

If we explain here in which case no further messages are loaded into the buffer, we can also change this

The number of messages added to buffer is returned, and this number will often be less than limit even if the channel is still open.

to remove the "often" part as we've explained in which situation the function is returning before the limit of messages is reached. I think that would be an important addition to the docs.

My suggestion for that part would be:

This method waits until at least one message is available, and then receives as many messages as possible into buffer (either the number of received messages reaches limit or no more messages are available at the time). The number of messages added to buffer is returned, and this number can be less than limit even if the channel is still open.

---

Now that I formulated my suggestions I am still wondering how my original proposal was not up to par. 😅 I feel like I've tried more to explain what happens in which situation compared to your proposal. In turn you've added the more obvious disclaimer that this function should not be expected to receive a limit amount of messages. At the end of the day I'm happy with any improvement of the docs here, but I'm still curious. 😉

Also, if we go with your proposal, please let me know if I should incorporate your suggestions (and give credit in the commit message I guess?) or if you will take over the contribution. :)

[SwishSwushPow]

Little bump here. Anything I can do to improve the state of this PR? :)