[WIP] Refactor interpolation

[mgreter]

Preview (WIP) of interpolation refactoring. Got rid of most of the escape vodoo.

Currently still fails in two main corners:

• missing leading white-space in argument lists (nulls in arglists)
• another white-space related bug with selector lists
• libsass issue Dynamic division with interpolation not compatible with libsass #948 (we seem to parse it wrong, see below)

Parsed AST for issue 948:

Binary_Expression 0x26933f0 [interpolant: 0]  [delayed: 0]  (0@[0:11]-[0:13]) [mul]
 left:  Textual  [NUMBER]0x2693240 [10]
 right: String_Schema 0x2b15f30 4 [has_interpolants] <>
 right:  Textual  [NUMBER]0x26932d0 [5] [delayed]
 right:  String_Constant 0x2693360 4 (0@[0:19]-[0:21]) [px] [interpolant] <>

I guess we had a "hotfix" in place for this before. But we may just don't parse it correctly.
Edit: Seems to be related to how we handle binary_expressions, see test code below:

foo {
  bar: 10 * 5#{px};
  bar: 10 / 5#{px};
  bar: 10 + 5#{px};
  bar: 10 - 5#{px};
}

[mgreter]

Spec test I'm currently using to test binary expressions with interpolations:

foo {
  bar: 10 * 5;
  bar: 10 / 5;
  bar: 10 + 5;
  bar: 10 - 5;
}

foo {
  bar: 10 * #{5};
  bar: 10  /  #{5};
  bar: 10 + #{5};
  bar: 10 - #{5};
}

foo {
  bar: 10 * 5#{px};
  bar: 10 / 5#{px};
  bar: 10 + 5#{px};
  bar: 10 - 5#{px};
}

baz {
  baz: #{5} * 1;
  baz: #{5} / 1;
  baz: #{5} + 1;
  baz: #{5} - 1;
}

baz {
  baz: #{5px} * 1;
  baz: #{5px} / 1;
  baz: #{5px} + 1;
  baz: #{5px} - 1;
}

bar {
  width: #{1+2}+4;
  width: #{1/2}/4;
  width: #{1+2}+4;
  width: #{1-2}-4;
}

foo {
  bar: 10/5;
  bar: 10/ 5;
  bar: 10 /5;
  bar: 10 / 5;
  bar: 10  /  5;
}

div {
  baz: #{4/5}/1;
  baz: #{4/  5}/  1;
  baz: #{4  /5}  /1;
  baz: #{4  /  5}  /  1;
}

mul {
  baz: #{4*5}*1;
  baz: #{4*  5}*  1;
  baz: #{4  *5}  *1;
  baz: #{4  *  5}  *  1;
}
add {
  baz: #{4+5}+1;
  baz: #{4+  5}+  1;
  baz: #{4  +5}  +1;
  baz: #{4  +  5}  +  1;
}

sub {
  baz: #{4-5}-1;
  baz: #{4-  5}-  1;
  baz: #{4  -5}  -1;
  baz: #{4  -  5}  -  1;
}

Expected result:

foo {
  bar: 50;
  bar: 10 / 5;
  bar: 15;
  bar: 5; }

foo {
  bar: 10 * 5;
  bar: 10 / 5;
  bar: 10 + 5;
  bar: 10 - 5; }

foo {
  bar: 50px;
  bar: 10/5px;
  bar: 15px;
  bar: 5px; }

baz {
  baz: 5 * 1;
  baz: 5 / 1;
  baz: 5 + 1;
  baz: 5 - 1; }

baz {
  baz: 5px * 1;
  baz: 5px / 1;
  baz: 5px + 1;
  baz: 5px - 1; }

bar {
  width: 3+4;
  width: 1/2/4;
  width: 3+4;
  width: -1-4; }

foo {
  bar: 10/5;
  bar: 10/ 5;
  bar: 10 /5;
  bar: 10 / 5;
  bar: 10  /  5; }

div {
  baz: 4/5/1;
  baz: 4/5/ 1;
  baz: 4/5 /1;
  baz: 4/5 / 1; }

mul {
  baz: 20*1;
  baz: 20* 1;
  baz: 20 *1;
  baz: 20 * 1; }

add {
  baz: 9+1;
  baz: 9+ 1;
  baz: 9 +1;
  baz: 9 + 1; }

sub {
  baz: -1-1;
  baz: -1- 1;
  baz: 4 -5 -1;
  baz: -1 - 1; }

What this WIP currently produces:

foo {
  bar: 50;
  bar: 10 / 5;
  bar: 15;
  bar: 5; }

foo {
  bar: 10 * 5;
  bar: 10 / 5;
  bar: 10 + 5;
  bar: 10 - 5; }

foo {
  bar: 50px;
  bar: 10/5px;
  bar: 15px;
  bar: 5px; }

baz {
  baz: 5 * 1;
  baz: 5 / 1;
  baz: 5 + 1;
  baz: 5 - 1; }

baz {
  baz: 5px * 1;
  baz: 5px / 1;
  baz: 5px + 1;
  baz: 5px - 1; }

bar {
  width: 3+4;
  width: 1/2 / 4;
  width: 3+4;
  width: -1-4; }

foo {
  bar: 10/5;
  bar: 10/ 5;
  bar: 10 /5;
  bar: 10 / 5;
  bar: 10  /  5; }

div {
  baz: 4/5 / 1;
  baz: 4/5 / 1;
  baz: 4/5 / 1;
  baz: 4/5 / 1; }

mul {
  baz: 20 * 1;
  baz: 20 * 1;
  baz: 20 * 1;
  baz: 20 * 1; }

add {
  baz: 9+1;
  baz: 9 1;
  baz: 9 + 1;
  baz: 9 + 1; }

sub {
  baz: -1-1;
  baz: -1- 1;
  baz: 4 -5 -1;
  baz: -1 - 1; }

As far as I can see this now would only differ in whitespace, but I had to do some very crude checks to get this behavior. @xzyfer IMO you said you want to takle this too? It seems that binary expressions somehow remember the whitespace around operators. IMO this needs to be addressed first before I could clean up the binary expression part of this refactoring.

[saper]

Libsass is also not accepting asterisk glued to numbers and interpolations, maybe this is related?

[xzyfer]

Part of this code will be addressed with my static value refactor which
will introduce a basic sass script parser. Doing so will reduce the scope
of string schema which is at the root of our interpolation issues.
On 15 Nov 2015 5:21 am, "Marcin Cieślak" [email protected] wrote:

Libsass is also not accepting asterisk glued to numbers and
interpolations, maybe this is related?

—
Reply to this email directly or view it on GitHub
#1738 (comment).

[mgreter]

@xzyfer so will your refactor solve all the issues in the sample I posted above?
Also is there a WIP PR where I can check in what direction you're going with that?

[xzyfer]

I'm not planning to directly address them. My work is focusing on improving
how we parse declaration values. This means some cases where we currently
use string schema will go away potentially improving these issues as a
consequence.

As far as my approach I'm simply changing our parser to better match the
Ruby implementation.
On 15 Nov 2015 9:40 am, "Marcel Greter" [email protected] wrote:

@xzyfer https:/xzyfer so will your refactor solve all the
issues in the sample I posted above?
Also is there a WIP PR where I can check in what direction you're going
with that?

—
Reply to this email directly or view it on GitHub
#1738 (comment).