Known Outstanding Issues Megathread

[connorskees]

This issue serves as a way to keep track of all known compilation issues

Large features

• @use and the module system (MVP)
• @forward rewrite parsing, evaluation, and serialization #67
• indented syntax rewrite parsing, evaluation, and serialization #67
• compressed output
• plain-CSS imports rewrite parsing, evaluation, and serialization #67

Smaller features

These ideally come with minimal reproductions, though the reason why they're failing may be incorrect as they haven't been fully looked into

• comma separated imports
  • @import "hey1.css", "cookie.css", url("hey2.css"), "fudge.css";
• parens surrounding single word in media query
  • @media (color) {a {color: red;}}
• min as arg to min/max
  • min(1, min(2))
• special function as arg to min/max
  • min(1, env(--foo))
• inequality sign inside interpolation in media query
  • @media (#{"100px < width < 500px"}) {a {interpolation: in-parens}}
• invalid simple selector (should error, but panics)
  • simple-selectors(">")
• vendor prefixed known at rules (one of @supports, @keyframes)
  • @-webkit-keyframes {}
• styles can render at the toplevel, but they should not be able to rewrite parsing, evaluation, and serialization #67
  • color: red; should error
• zero divided by zero panics, but should return NaN
  • (0 / 0)
• one divided by zero panics, but should return Infinity (this has been updated to return an error instead of crashing) rewrite parsing, evaluation, and serialization #67
  • (1 / 0)
• division should only occur when either of the operands is calculated or if they operation is in parens
  • 1 / 2 should be emitted as 1/2, but it currently gives 0.5
• CSS variables should include everything, notably comments rewrite parsing, evaluation, and serialization #67
  • --foo: //;
• numbers should use fuzzy matching -- that is, they should be equal if they are the same to 10 decimal points rewrite parsing, evaluation, and serialization #67
  • .9999999999999999999999999999999 == .99999999999999999999999999999998
• we should allow a BOM at the start of files
• all instances of @import should be moved to the top of the file
• rgb(a) functions should allow / rewrite parsing, evaluation, and serialization #67
  • rgba(1, 2, 3 / 4)
• we dont retain empty parens inside url(...) rewrite parsing, evaluation, and serialization #67
  • url((((())))) (this should actually produce an error, but it just emits url()
• @import should include strings verbatim rewrite parsing, evaluation, and serialization #67
  • @import "hux\ bux.css"; this should include the \, but it does not because we use normal string parsing
• subtraction with no space between operands and rhs is interpolated
  • 10-#{10} should give 10 -10 but it gives 10-10
• non-comparable inverse units rewrite parsing, evaluation, and serialization #67
  • (1px / 1em)
• !optional in @extend
• deny content block being passed to @mixin without @content rewrite parsing, evaluation, and serialization #67
• @extend should not be possible between media query boundaries
• @media query merging rewrite parsing, evaluation, and serialization #67
• inspect(...) does not properly preserve parens in some cases
• @charset should only accept a string
  • @charset 1+1; should error
• function names containing interpolation should always be evaluated as unknown functions rewrite parsing, evaluation, and serialization #67
  • color: qu#{o}te(red) current returns "red", but should give quote(red)
• unknown at rule with query but no curly braces
  • @foo "bar"; (we do parse this correctly, but we emit it with two semicolons)
• unicode ranges
  • e.g. U+A2??
• allow varargs to contain named arguments and support builtin fn keywords rewrite parsing, evaluation, and serialization #67
• feature complete parsing of @supports rewrite parsing, evaluation, and serialization #67
• support angle units other than deg to builtin fns, math.cos(1grad) rewrite parsing, evaluation, and serialization #67

[pickfire]

@connorskees I am interested to help out on compressed output, are you working on that? If possible, maybe you can write out some mentor instructions?

[Keats]

It looks like it's just missing compressed output to be a good replacement to sass-rs? Would you consider this library ready @connorskees ?

[connorskees]

@Keats I think it is close to being ready! There's basically the same underlying issue preventing a number of features that I would like to resolve before I am willing to call the library ready. I'd like to get this library over the finish line in the first quarter of 2021, assuming I have enough free weekends.

I think, though, if given the explicit warning that the library is experimental and may not work for all features of the language, it may be fine to rely on grass in an opt-in fashion. The known features remaining are admittedly quite niche (as far as I can tell) -- specifically I am concerned about @media query merging, @extend inside @media, complex uses of @at-root, invalid syntax inside plain-CSS variables, and certain arguments to min/max.

There shouldn't be any more breaking API changes for quite some time.

I think if you check back in 2-3 weeks, it should be in a good enough state to use experimentally. And then in another 2-3 months it should hopefully be in a good enough state to use seriously.

[Keats]

Nice! I'll make it an opt-in feature of the next version of Zola. The listed features are pretty niche I think so it shouldn't cause issues for most users I think and I can finally drop compass soon-ish.

[pickfire]

cc @kdy1 who may be interested for swc-project/swc#950

[Keats]

@connorskees is this project still active?

[connorskees]

hey @Keats, yes this project has recently come back to life. I had to take a hiatus at the beginning of the year, but I am hoping to wrap this project up. I have made a number of improvements that should resolve many of the concerns I had earlier.

This project should be fine to use in a production setting. I have verified that grass' output is byte-for-byte the same with dart-sass for the last 2,500 commits to bootstrap, which includes all of bootstrap 4 and 5.

[pickfire]

@connorskees I think compressed output is not as good as the original one yet. There are quite some optimizations that we could do. Also, I think we can do better maybe. So not sure if you want to mark that as checked.

[connorskees]

@pickfire We actually have a pretty good amount of compressed output-parity. I've just pushed up a commit adding support for compressed number and list values. I think the number of bytes we could further remove is quite small and will not affect most users.

For reference, the features of dart-sass's compressed output are enumerated here.