Dynamically generated mixins suddenly can be used as mixins

[SomMeri]

E.g. this:

.@{fa-css-prefix} { display: inline-block; }
.@{fa-css-prefix}-2x { font-size: 2em; }

.my-class {
  .my-icon;
  .my-icon-2x;
}

@fa-css-prefix: my-icon;

compiles into:

.my-icon {
  display: inline-block;
}
.my-icon-2x {
  font-size: 2em;
}
.my-class {
  display: inline-block;
  font-size: 2em;
}

Was this intentional and do we know how variables scoping works? This works now and people seem to be discovering it. I did not dug into how exactly the feature works, is there anyone who knows?

Related to #1399 and #1836 .

[seven-phases-max]

Not quite suddenly :) It was in #1624 PR.

do we know how variables scoping works?

Yes. It's all is almost as usual except the #1399 (comment) first example (note that the third example "experiment" was not included in the PR). In summary: mixins are evaluated before variables so "dynamically generated class" is not available as mixin in the scope of its definition (hence it can't re-define itself by redefining an interpolated variable).

I'm not sure, maybe we could close #1399 and open a new issue dedicated to further improvements? (#1399 has too many examples to comment: some of them are implemented in #1624, some do not apply at all, some are more about "selector interpolation" rather than using those as mixins and others need more decisions/discussions).

[SomMeri]

@seven-phases-max So, the order is somewhat like this:
1.) match mixins calls with mixins,
2.) replace variables by their values,
3.) evaluate guards and replace mixins?

Another question: Dynamic mixin has to be defined before being used. Was that intentional? Following throws compile error:

.my-class {
  .my-icon;
}

.@{fa-css-prefix} { 
  display: inline-block; 
}

@fa-css-prefix: my-icon;

[seven-phases-max]

So, the order is somewhat like this:

It's even more nuclear. Variables are replaced with their values after mixins of the same scope are expanded/replaced. It's not in #1624 code, it's just how mixin expansion is made in general, e.g.:

@var: 1;

.test {
    var: @var;
    .mixin() {@var: 2;}
    .mixin();
}

Dynamic mixin has to be defined before being used. Was that intentional?

No. The #1624 actually does not change any mixin/variable evaluation/scope stuff, it's just expands @{var} to its value, that's it (but the value itself is set elsewhere so if variable is not evaluated yet the mixin match fails because @{var}'s value could be just "@var" yet). I think the PR diff. can put some light on how it works: lines 37-39 is just what it does, nothing else.

[SomMeri]

@seven-phases-max Thanks for the answer. I will play with the feature once I will have some free time. I have none right now :(