self: in selector

[designerno1]

TestCase

.test {
    &:last-child:not(+ &:first-child) {
      padding-left: 10px;
    }
}

output sass compiler

.test:last-child:not( + :first-child) {
  padding-left: 10px;
}

output scssphp

.test self:last-child:not(+ &:first-child) {
  padding-left: 10px;
}

it's also inside a mixing the same

[designerno1]

SassMeister.com output is

.test:last-child:not(+ :first-child) {
  padding-left: 10px;
}

both LibSass 3.2.5 and Ruby 3.4.14

[robocoder]

Ok.

Reference: https://drafts.csswg.org/selectors-4/#negation

[Cerdic]

There is 2 differents bugs here:

• the first & is well parsed and wrong compiled,
• whereas for the second the parser is not seeing it as a self and rendering it as a string &

[Cerdic]

The first part is easy to solve with #695

For the second we are facing a bigger challenge.

At the moment, when parsing the selectors and enterring in a :not() (or :is() or :has() or :where()) the intern part of parenthesis is parsed as a string, where you can have interpolation but not self selector.

The real good fix would be to parse the intern part of parenthesis as a new selector, as this is what it is and you can meet here any selector syntax.

But then comes some impact on the compiler wich doesn't expect sub selectors in selectors part…

The other fix, easyer but less efficient as a long term solution, is to optionnaly add the & as a possibility in parsing a string

[Cerdic]

with #696 it now compiles as

.test:last-child:not(+ .test:first-child) {
  padding-left: 10px; }

Which is not fully the same as saas compiler, but at least seems correct.
(In fact I'm not sure how/on which rule the & should/need to be optimised and removed in the :not() part)