Re: [boost] Boost.URL review

For this case, I definitely prefer ref to view. Zach On Wed, Aug 24, 2022 at 10:37 AM Vinnie Falco wrote:

On Wed, Aug 24, 2022 at 7:59 AM Zach Laine via Boost wrote:

...

This seems to be the heart of the matter. I expect params to be a real flat_map or map or sorted vec, and it's not.

Yeah, I see what you mean. "params" sounds like its a first-class object, but it isn't.

...

If my expectation is wrong because of the lack of a _view suffix in the name, that's fine -- but please add that to make things clear.

There are 8 containers which need names:

segments segments_view segments_encoded segments_encoded_view params params_view params_encoded params_encoded_view

We prefix the type name with the type of element "segment" or "param", so that it sorts nicely. We could have went with "encoded_segments" which also makes sense but then they don't group together as well and it hurts discoverability in the docs (Ramey Rule).

The _encoded suffix is self-explanatory. We chose the suffix _view to distinguish from the read-only containers from the mutating ones. Someone suggested putting "const" and "mutable" in the name but in my opinion that is ugly:

const_segments_view mutable_segments_view const_encoded_segments_view mutable_encoded_segments_view const_params_view mutable_params_view const_params_view mutable_params_view

...

However, the _view suffix in std C++ is used exclusively for immutable types, so maybe something else is indicated, like _facade or some variation of span.

As if we didn't have a hard enough time coming up with 8 names that have synergy and aesthetics :) How about the suffix _ref?

segments_ref segments_cref segments_encoded_ref segments_encoded_cref params_ref params_cref params_encoded_ref params_encoded_cref

Or we could mix _ref and _view?

segments_ref segments_view segments_encoded_ref segments_encoded_view params_ref params_view params_encoded_ref params_encoded_view

We did the best we could with the current names and I know they aren't perfect. We're more than happy to hear feedback on whether any of these alternatives are resonating with folks, or if someone wants to propose a new set of 8 identifier names.

Thanks