patching an entity will not set 'PATCH' as method

[fiws]

but patch (lowercase)
I don't think that's how it should be.

[opengeek]

This is a big problem because of JakeChampion/fetch#37 ...

[opengeek]

Looks like this has to be fixed outside of fetch, as they are not going to normalize patch to uppercase.

[fiws]

restful.js should not rely on fetch normalizing the method names anyway IMO. Just supply the right method names..

[opengeek]

Agreed — I will try and submit a "patch" for this issue today that uppercases all the method names for consistency.

[fiws]

👍 (that pun)

could just uppercase it here https://github.com/marmelab/restful.js/blob/master/src/model/endpoint.js#L16
not sure if that will break anything though

[opengeek]

What do you think of the approach I've taken in #83 ? Instead of forcing all to uppercase, I thought it would be best to just define the methods of the endpoint as uppercase to begin with.

[fiws]

looks good 👍
waiting for merge.. :\

[RobinBressan]

Should be fixed by 0.9.6 thanks to @opengeek.

[fiws]

actually a breaking change – isn't it? Does restful.js not follow semver? :\

[RobinBressan]

Yes it does but I don't think this is a BC. The restful.js api was not updated.

[fiws]

What if projects relied on the "broken" patch method? Those will now eventually hit a roadblock.

[opengeek]

I don't believe it's possible to rely on the bug here, so this is not a BC break. AFAIK, patch wasn't working at all since the switch to using fetch and this simply fixes that bug and consistently addresses the source of the issue, which was the definition of the methods as lowercase and relying on the normalization of those methods to uppercase by fetch.

[fiws]

patch was "working" – it was just sending patch instead of PATCH to the server. The request reached my server.
It's ok for my if this is treated "non-breaking", just my thoughts.

[opengeek]

Well, it broke all of my projects using restful.js when the project moved to fetch because "patch" is not a valid method and the browser will not allow it due to CORS pre-flight checks. But perhaps this should increment the minor pre-1.0 version number to indicate the potential additional BC break on this method. I'll leave that to the maintainers to decide.