JavaScript: introduce getter and setter kinds + computed properties

[masatake]

Partially close #1949.
See https://www.ecma-international.org/ecma-262/5.1/#sec-11.1.5

Signed-off-by: Masatake YAMATO yamato@redhat.com

[coveralls]

Coverage increased (+0.01%) to 84.965% when pulling 9d44448 on masatake:js-getter-and-setter into 75ba24d on universal-ctags:master.

[masatake]

I wonder we can remove signature field for getter kind objects.

[b4n]

Looks pretty good!

However, there's case where we'll have to do something with computed property names that aren't a simple literal: unfortunately, the following code is perfectly valid JS:

var o = {
	[(1+2)*3]: {
		subnum: function (){}
	},
};

Currently this will cause trouble ("warning: null tag" for the property, and subnum has an invalid scope).
I'm not sure if it's worth doing much here, but maybe we should record the expression and just use this literally? Not sure what makes sense, as we won't ever do this perfectly correctly though -- here the property name is actually 9, but could be an arbitrary computation using dynamic variables and such.

In any case, I think this can be dealt with later, as it's covering the cases where we really have something correct to give.

[masatake]

I introduced is_dynamic_prop local variable to skip (1+2)*3. If you did not know this, look at
9d44448.
However, I didn' handle the nested case. For capturing subnum and fill the scope field for subnum, we have to put an anonymous tag as the proxy(?) for (1+2)*3.

Anyway I guess we have more things to do about #1949. So I would like to merge this PR.
Is your approve for my changes including 9d44448 ?

[b4n]

@masatake yes I've seen it, but as mentioned it's not checked in case of recursing into a subclass; and there we need something to provide scope for the members. Another way could be to ignore the subclass methods in this case.

And yes, my review applies at the time I made it, to the 3 commits.

[masatake]

@b4n, thank you!