servant-client: Run ClientEnv's makeClientRequest in IO #1595
Conversation
|
@Minnozz Hi! Has there been a discussion about this PR? :) |
|
@tchoutri Hi! This change was suggested by |
|
Great, then I'll leave @alpmestan review your PR. :) |
There was a problem hiding this comment.
This looks fine to me. We'll want to add an entry in the relevant changelog. Also, this will force a major version bump.
Let's perhaps leave a few days to give others a chance to give feedback on this, and for you to add a changelog entry (see how other PRs do it, with entries in the changelog.d directory).
|
@alpmestan I added a changelog entry, do you think it needs more explanation? |
|
@Minnozz Should be fine, I think people using servant are familiar enough with If you ever feel like writing down a little something that demonstrates how you're using this to your advantage (cookbook recipe, blog post, whatever), then it can serve as a reference for users who might need this in the future. But that's more work that adding a few lines in the changelog :-) |
|
Considering the approval of this PR, I will merge, but a cookbook or a blog post would be very appreciated, @Minnozz :) |
|
Please have a look at #1787 trying to read this change through I can’t say why this change is necessary. Why was it worth to break the API? Why does it have to be exactly |
|
@unclechu The change was necessary for my project to be able to call applyDigestAuth, as is noted in the description of this PR. As for requiring precisely IO: as a Haskell novice I did not understand the difference between IO and MonadIO when I made this PR; feel free to propose a PR with a better API :) |
|
+1, submit a PR to depend on MonadIO, it would be an improvement. |
I don’t think it was necessary, you could always call
I’ll make a pull request. |
|
It's not about |
|
I would just mean the only change required is for |
The `IO` wrap for the return type of this function was never necessary. Inside the function there’s only `return` call that wraps a _pure_ value into `IO`. There is nothing else that would be related to `IO`. The `IO` wrap was introduced in 9df16f9 In order to not break the API again the wrap was preserved but this change makes it to depend on any `Applicative` instead of `IO`. So it still works in `IO`/`MonadIO` context but allows to use something like `Data.Functor.Identity` to work with _pure_ values. See for more context: haskell-servant#1595
|
The change for the |
The `IO` wrap for the return type of this function was never necessary. Inside the function there’s only `return` call that wraps a _pure_ value into `IO`. There is nothing else that would be related to `IO`. The `IO` wrap was introduced in 9df16f9 In order to not break the API again the wrap was preserved but this change makes it to depend on any `Applicative` instead of `IO`. So it still works in `IO`/`MonadIO` context but allows to use something like `Data.Functor.Identity` to work with _pure_ values. See for more context: #1595
The `IO` wrap for the return type of this function was never necessary. Inside the function there’s only `return` call that wraps a _pure_ value into `IO`. There is nothing else that would be related to `IO`. The `IO` wrap was introduced in 9df16f9 In order to not break the API again the wrap was preserved but this change makes it to depend on any `Applicative` instead of `IO`. So it still works in `IO`/`MonadIO` context but allows to use something like `Data.Functor.Identity` to work with _pure_ values. See for more context: haskell-servant#1595
Use cases: