Faraday: Deletes with bodies

A PR will be up shortly for your review.

Hi @Carlbc18,

could you please provide a link to the above changes?
I guess that's only true for HTTP 2.0 specs, so we might need to consider this feature carefully.

https://tools.ietf.org/html/rfc7231#section-4.3.5

FYI: This same block of text applies to GET && HEAD requests as well.

specifically this part of the spec:

A payload within a DELETE request message has no defined semantics;
sending a payload body on a DELETE request might cause some existing
implementations to reject the request.

i have a code change with tests that works (not sure if you want it). how do i create a pr for this project?

I guess that's only true for HTTP 2.0 specs, so we might need to consider this feature carefully.

this is actually part of HTTP 1.1

I would definitely consider it if you want to create a PR.
You can do it as with any other repo, simply fork faraday and apply the changes on your fork, then create a PR to our master branch.

Thanks @iMacTia. I'll put it up. Whats your views on changing the version finally have a major version number with this? This change could be breaking for many. I'd like to follow semver pattern, and version faraday to be 1.0.0 at this point. Let me know your thoughts.

We're definitely trying to stick to semver and we've v1.0 scheduled with some breaking changes.
If this change is gonna break backward compatibility, then it will be released only in version 1.0 (which currently doesn't have a fixed release date).
However, my understanding is that you should now be able to OPTIONALLY add a body to GET, HEAD and DELETE requests, so shouldn't that be backward compatible? Am I missing something?

@iMacTia i versioned to 0.13.0 and the PR is up. We can take any other conversation around this to the PR as i'm sure there will be comments. Thanks again!

This is a follow up on my https://github.com/lostisland/faraday/pull/695#issuecomment-305145499.
First of all we should decide HOW we would like this to work.
Assuming changes should be backward compatible, the below should still work:

conn = Faraday.new(...)
conn.get('/path', {page: 1, per: 10})

# performs "GET /path?page=1&per=10"

Option 1 as stated in my comment would be to make this behaviour configurable:

conn = Faraday.new(..., standard: :http1_1)
conn.get('/path', {page: 1, per: 10})

# performs "GET /path" with body "page=1&per=10" (assuming www_form_url_encoded request)

Option 2 instead would mean something more powerful:

conn = Faraday.new(...) # note no special config here
conn.get('/path', {page: 1, per: 10}, body: {some_key: 'some_value')

# performs "GET /path?page=1&per=10" with body "some_key=some_value"

There's an issue though, option 2 might only work with ruby >= 2 as ruby 1.9 doesn't support keyword params 😞

I think in regards to this not working on ruby 1.9 I feel like that is ok. Ruby 1.9 is many years old. I'd advocate we version bump to at least 2.0 for minimum ruby version, which would go inline with a v1.0 release of Faraday. Also option 2 more closely follows http 1.1 spec in my opinion as you'd have the option to pass parms and or a optional body.

I feel like attempting to maintain backwards compatibility with a http spec feature that does work with ruby 1.9 is going to cause problems. I think this feature would be better suited for V1 of Faraday. I feel like there are too many variables were needing to support for this change (ruby v, params passing). Maybe I'm overthinking....

I think in regards to this not working on ruby 1.9 I feel like that is ok. Ruby 1.9 is many years old. I'd advocate we version bump to at least 2.0 for minimum ruby version, which would go inline with a v1.0 release of Faraday. Also option 2 more closely follows http 1.1 spec in my opinion as you'd have the option to pass parms and or a optional body.

I totally agree with al the above. That's indeed the best course of action, even though this means we'll still have to wait for v1.0

So to define a v1.0 backward compatible API, we could attempt the following:

GET, HEAD, DELETE REQUESTS

conn = Faraday.new(...) # note no special config here
conn.get(path, url_params, body: {...})
# second parameter is URL PARAMS, optionally accept a named parameter for request body
# works the same for head and delete

POST, PUT, PATCH REQUESTS

conn = Faraday.new(...) # note no special config here
conn.post(path, body, url_params: {...})
# second parameter is REQUEST BODY, optionally accept a named parameter for url_params
# works the same for put and patch

I think this way we're covering every possible combination and extending post/put/patch requests as well!

Definitely agree. Do we want to decline my PR and open a new one for this? Is there a V1 branch cut we can work off of?

Happy to close that one and start from scratch as it will be completely different. Sorry for that!

I'll create a V1 branch before merging in so feel free to start from master and I'll redirect the PR :)

Hi @Carlbc18 just wanted to let you know that branch 1.0 is now available for you 😃.

Thank you, thank you! 😀

For anyone still waiting on this, I found the following workaround which seems to work in my case:

conn = Faraday.new
body = {
some_key: 'some value'
}
conn.run_request(:delete, url, body, headers)

I hope this helps someone until delete no longer nils the body by default

So is there any update on when this will be available?

I haven't seen any PR yet, but works on v1 are slowly progressing.
I might be able to work on this myself but I can't give an estimate at the moment.
The updated API on how requests should work is explained on my comment above, so if anyone wants to have a shot at it before me, a PR is very welcome.

For a fully backwards compatible solution, the PR can be branched out of master, and I can make another 0.x release.
If the change is going to break backwards compatibility then it's better to branch out of 1.0 branch

https://github.com/lostisland/faraday/pull/855 is a draft PR to support this.

I'm rejecting this feature. You can use the block form in Faraday to make a request with a body:

conn.delete(url) do |req|
  req.body = { some_key: 'some value' }
end

This idea would have made simple deletes with bodies a single one liner, but the implementation adds more complexity and surprising behavior than I'd like. This block form works great in Faraday already. I prefer modifying properties on a class to adding more arguments to a method:

conn.delete('/foo/1') do |req|
  req.headers["X-Men"] = "marvel"
  req.params[:confirm] = 1
  req.body = { some_key: 'some value' }
end