Noissue - fix generation of a new instance creation of a class without subclass differentiator

If a class has no subclass differentiator then instead of creation of a new wrapping JS object, namely calling of `new exports.MyClass(pointer)`, there was the call `exports.MyClass(pointer)`.

Review: https://codereview.adblockplus.org/29543756/

[sergei, 2017-09-13 17:09:10]

https://codereview.adblockplus.org/29543756/diff/29543757/compiled/bindings/g...
File compiled/bindings/generator.cpp (right):

https://codereview.adblockplus.org/29543756/diff/29543757/compiled/bindings/g...
compiled/bindings/generator.cpp:260: result += "    result = exports." +
cls->name + ".fromPointer(result);\n";
Another option here could be to keep the if and for the branch without the
differentiator put ("new " added):
`result += "    result = new exports." + cls->name + "(result);\n";`
and remove the generation of implementation of fromPointer for such classes.
However, I think it can be convenient for debugging and maybe even for some
future changes to call constructors only from "fromPointer".
What is your opinion?

[Wladimir Palant, 2017-09-19 08:51:48]

https://codereview.adblockplus.org/29543756/diff/29543757/compiled/bindings/g...
File compiled/bindings/generator.cpp (right):

https://codereview.adblockplus.org/29543756/diff/29543757/compiled/bindings/g...
compiled/bindings/generator.cpp:260: result += "    result = exports." +
cls->name + ".fromPointer(result);\n";
On 2017/09/13 17:09:10, sergei wrote:
> Another option here could be to keep the if and for the branch without the
> differentiator put ("new " added):
> `result += "    result = new exports." + cls->name + "(result);\n";`
> and remove the generation of implementation of fromPointer for such classes.
> However, I think it can be convenient for debugging and maybe even for some
> future changes to call constructors only from "fromPointer".
> What is your opinion?

Trouble is, bindings overhead is very much non-negligible - that's the reason
why I am micro-optimizing here. So I'd rather keep the overhead as low as
possible and not call fromPointer unless necessary. In fact, fromPointer was
inlined here originally and I am still not sure whether turning it into a method
call was the right decision to make.

[sergei, 2017-09-19 09:14:35]

https://codereview.adblockplus.org/29543756/diff/29543757/compiled/bindings/g...
File compiled/bindings/generator.cpp (right):

https://codereview.adblockplus.org/29543756/diff/29543757/compiled/bindings/g...
compiled/bindings/generator.cpp:260: result += "    result = exports." +
cls->name + ".fromPointer(result);\n";
On 2017/09/19 08:51:48, Wladimir Palant wrote:
> On 2017/09/13 17:09:10, sergei wrote:
> > Another option here could be to keep the if and for the branch without the
> > differentiator put ("new " added):
> > `result += "    result = new exports." + cls->name + "(result);\n";`
> > and remove the generation of implementation of fromPointer for such classes.
> > However, I think it can be convenient for debugging and maybe even for some
> > future changes to call constructors only from "fromPointer".
> > What is your opinion?
> 
> Trouble is, bindings overhead is very much non-negligible - that's the reason
> why I am micro-optimizing here. So I'd rather keep the overhead as low as
> possible and not call fromPointer unless necessary. In fact, fromPointer was
> inlined here originally and I am still not sure whether turning it into a
method
> call was the right decision to make.

I have uploaded the version without fromPointer for classes without the subclass
differentiator.

[Wladimir Palant, 2017-09-19 09:56:30]

LGTM

https://codereview.adblockplus.org/29543756/diff/29549577/compiled/bindings/g...
File compiled/bindings/generator.cpp (left):

https://codereview.adblockplus.org/29543756/diff/29549577/compiled/bindings/g...
compiled/bindings/generator.cpp:458: }
Having this method here for consistency won't really hurt, but removing it is
fine with me - it's not really used anywhere.