Completion metadata

[cichli]

cc @bbatsov @gtrak

As discussed in #12. This is a breaking change but it would be easy to keep the old behaviour and add a new function for this.

See the tests for more examples but completions now returns a sequence that looks like this:

({:candidate "*1" :ns cljs.core :type :var}
 {:candidate "-add-watch" :ns cljs.core :type :protocol-fn}
 {:candidate "aget" :ns cljs.core :type :fn}
 {:candidate "doseq" :ns cljs.core :type :macro}
 {:candidate "IdGenerator" :ns goog.ui.IdGenerator :type :import}
 {:candidate "TestRecord" :ns cljs-tooling.test-ns :type :record}
 {:candidate "TransientArrayMap" :ns cljs.core :type :type})

The possible values of :type are:

:special-form
:ns
:fn
:protocol-fn
:record
:type
:protocol
:var
:macro
:import

If there are multiple candidates with the same value of :candidate (i.e. any of the fns in cljs.core that wrap macros with the same name), then only one will be used - the one that would be returned by an info call for the value of :candidate (i.e. order of preference is the same as the list above).

Not sure if we should force the client to account for all of the possible types or collapse them down a bit?

[bbatsov]

Great work! Ideally @alexander-yakushev would add the same behaviour to compliment soon as well!

[alexander-yakushev]

Can anyone tell me the different between :candidate and :name?

[cichli]

Sorry - didn't include enough examples to make it clear :-). :candidate can contextually be fully-qualified, e.g. if prefix is cljs.core.async.macros/g then completions will look like:

({:candidate "cljs.core.async.macros/go" :ns cljs.core.async.macros :name cljs.core.async.macros/go :type :macro}
 {:candidate "cljs.core.async.macros/go-loop" :ns cljs.core.async.macros :name cljs.core.async.macros/go-loop :type :macro})

and for prefix dose:

({:candidate "doseq" :ns cljs.core :name cljs.core/doseq :type :macro})

So basically the values of :candidate are what the editor will actually use for completion, whereas :name is just the non-namespace-qualified namespace-qualified name.

[cichli]

Happy to remove :name if it's not that useful btw.

[bbatsov]

Seems your original examples was wrong - as :name and :candidate where reversed there, which looked wrong (but I didn't notice until @alexander-yakushev mentioned it). I think it the current format is good.

[alexander-yakushev]

Don't worry, I'm asking this for myself because @bbatsov won't give me a break :D. So Bozhidar is this something you want from Compliment too?

[bbatsov]

@alexander-yakushev Yes. I think I created a ticket for this half a year ago. It will make it possible for us to have richer completion metadata (e.g. in company, ac, etc). See alexander-yakushev/compliment#16

[cichli]

Yep, well spotted - I lied about :name, it's not consistent at all right now 🤦. I think it should always be namespace-qualified (as it is currently for macros only). Will sort it tonight.

[alexander-yakushev]

@bbatsov I specifically mean :name and :candidate distinction.

[cichli]

We can probably just omit :name entirely unless the client intends to actually use it. info ops already work fine on completion candidates regardless of whether they're ns-qualified or not.

[cichli]

Okay, patch updated with :name removed and some other bits:

• test case for fns that shadow macros in the same ns
• :special type renamed to :special-form to match compliment
• some docstring tweaks

[bbatsov]

👍 Let's merge this!

P.S. While you're on the subject of small tweaks - you might improve the README as well. :-)

[gtrak]

Will take a look tomorrow. Thanks!

On Fri Feb 27 2015 at 2:28:34 PM Bozhidar Batsov [email protected]
wrote:

[image: 👍] Let's merge this!

P.S. While you're on the subject of small tweaks - you might improve the
README as well. :-)

—
Reply to this email directly or view it on GitHub
https://github.com/gtrak/cljs-tooling/pull/15#issuecomment-76456072.

[gtrak]

Great work!