String interpolation behaves differently in mixin and in ruleset

[SomMeri]

String and escaped value interpolation behavior depends on context:

• If a mixin contains interpolated string, the interpolation iterates while possible.
• If a ruleset contains interpolated string, the interpolation runs only once.

The same value @{box-@{suffix}} with the same scope variables available can end up either as @{box-large} or as 100px. The result depends on whether mixin was called.

The issue #1387 suggests that the interpolation should iterate.

Full example:

@box-small: 10px;
@box-large: 100px;

.mixin { // both ruleset and mixin
    width: ~"@{box-@{suffix}}";
    weird: ~"@{box}-@{suffix}}";
    @box: ~"@{box";
        @suffix: large;
}
.interpolation-mixin {
    .mixin(); //call the above as mixin
}

compiles into:

.mixin {
  width: @{box-large};
  weird: @{box-large};
}
.interpolation-mixin {
  width: 100px;
  weird: 100px;
}

Expected result:

.mixin {
  width: 100px;
  weird: 100px;
}
.interpolation-mixin {
  width: 100px;
  weird: 100px;
}

[seven-phases-max]

Yes, currently the number of nested mixins calls -> number of allowed nested string interpolations:

@foobar: #69c;

.z {
    @pre:   foo;
    @post:  bar;
    @p:     p;

               @a: ~"@{@{pre}@{post}}";
  .1()        {@b: ~"@{@{pre}@{post}}"} .1;
  .2()        {@c: ~"@{@{@{p}re}@{post}}"} .2;
  .3() { .4() {@d: ~"@{@{@{p}re}@{post}}"} .4} .3;

    a:  @a;  // @{foobar}
    b:  @b;  // #6699cc
    c:  @c;  // @{foobar}
    d:  @d;  // #6699cc
}

This is because each mixin expansion evaluates mixin variables thus "removing" the top-level @{...}.

[SomMeri]

This was fixed by previous commit.