Closed dbu closed 8 years ago
Yeah, probably this point of view makes the most sense. Even if the logic of package organization might not be clear for everyone at first, it is easy to understand, so let's follow this.
Client-helper maybe?
More than implementation-helper or integration-helper (like the adapter-integration-test package name)
(Client helper is confusing IMO)
It can be closed i think, as all classes here now are user utilities.
right, thanks!
Reopening this as I think there is room for some more separation here. PR is on the way, after that we can decide about any further separation.
i think the Batch stuff we talked about is now fine here in utils, as there is no more interface. but if there are other implementor-only things i agree to separate them.
Yeah, that's why I changed the title.
I think that partly the Fulfilled/Rejected Promise thing can also be partly implementor things, but I am in the middle of planning.
In the meantime: naming question is still open. Utils? How should we call the two packages?
i think utils is a good name for user utilities. the other could be implementor-tools maybe?
From the purpose of point of view, it is a good name.
From a domain point of view, it is not really.
Pushed changes in the implementation_separation branch.
I would use these tools for example in the guzzle6 adapter where the sync behavior is currently emulated.
From the purpose of point of view, it is a good name.
From a domain point of view, it is not really.
which one? utils? i would stick with utils as purpose is more relevant here than domain - its a collection of stuff really...
for the implementor tools, i am open for better ideas, just tried to come up with a name :-)
which one? utils?
It is ok.
for the implementor tools
Yeah, this one. I would drop the implementor part if possible. tools? httplug-tools? client-tools? integration-tools as @joelwurtz suggested?
I don't think we will find a name which describes the package correctly, so we definitely need to describe it anyway.
client-tools seems not bad. it goes well with the virtual client-implementation package
Solved in #20
We discussed the role of php-http/utils versus php-http/helper. One idea was to separate things that HTTP client users would need versus HTTP client providers.
BatchRequest and BatchResponse are only relevant for implementors, not for users of a client. The users have to use the BatchResponse interface, and the BatchRequest trait is only relevant for adapters for non-parallel clients. Imho we should move those to helper (and maybe think of a better name for that repository, like client-implementation-helper)