Closure iterators are not supported by VM

[ringabout]

I can raise error messages in JS backend too.
Are there some plans or ideas to implement closure iterators in VM or JS backend.

[timotheecour]

Are there some plans or ideas to implement closure iterators in VM or JS backend.

• I've shown in experiment showing how iterators can (almost) be implemented with a library solution #15655 that you can implement the same thing as a closure iterator using a pure library solution, and it works with all backends (c, js, vm etc) without special support needed.

At call site, the only difference with you need to add an explicit iterate:

for k,v in myIter2(2,5):
  echo k, v
  if k == 0: continue
  if k == 1: break
  
=>
for k,v in iterate myIter2(2,5):
  echo k, v
  if k == 0: continue
  if k == 1: break

At declaration site, you use a regular proc instead of an iterator to implement it, see details in #15655

This is already useful today, for code that needs functionality equivalent to closure iterators in VM or js.
And with some syntax sugar, the compiler could maybe map the existing syntax (iterator myIter() ...) to this, and not have to implement those for all backends, thus simplifying the language. Note that with alias PR (#11992),you'll also be able to define inline iterators the same way.

[Araq]

I've shown in #15655 that you can implement the same thing as a closure iterator using a pure library solution, and it works with all backends (c, js, vm etc) without special support needed.

Yes but "moved feature X from the compiler to the libraries" always causes lots of work for anybody involved and plenty of subtle regressions. We should simply fix the compiler so that closure iterators are supported in the VM etc. Let's fix what we have.