[css-nesting] Inserting new sub-rules with .style

[sesse]

During test writing, this came to mind:

<style>
  .a {
    color: red;
    &.c { color: hotpink; }
  }
</style>
<script>
document.styleSheets[0].rules[0].style = 'color: navy; &.b{color: green;}';
</script>

What should happen?

1. The change is silently ignored.
2. “color: red” is changed to “color: navy” and no other changes take place.
3. The same, but an additional rule “&.b { color: green; }” is added, above &.c.
4. The same, but it's instead added after &.c.
5. The same, but &.c is removed.

I'd say the sane thing to do is 2 since we parse a list of declarations (and already ignore @media etc. when setting .style), but this needs to be explicit in the spec, I think.

[LeaVerou]

Setting .style should replace anything inside the rule so: “color: red” is changed to “color: navy”, &.c is removed, and &.b is added (not sure what number that is in your list).

[sesse]

Setting .style should replace anything inside the rule so: “color: red” is changed to “color: navy”, &.c is removed, and &.b is added (not sure what number that is in your list).

That would be 5. But 5 is inconsistent with current behavior for e.g. @media, and it's also inconsistent with the current definition of .style in CSSOM (which says to forward to .cssText, which says to parse a CSS declaration block, which cannot contain other rules). https://www.w3.org/TR/cssom-1/#dom-cssstyledeclaration-csstext

[Loirooriol]

If a CSSStyleRule has CSSStyleDeclaration style and CSSRuleList cssRules as different members, it seems strange that assigning to one of them will change the other one...

This seems a good argument for @FremyCompany's proposal that keeps each thing separate.

[LeaVerou]

This seems a good argument for @FremyCompany's proposal that keeps each thing separate.

Or a good argument to hang this off of CSSStyleDeclaration, paving the way for nested inline styles too.

[sesse]

paving the way for nested inline styles too.

I hate to repeat myself, but we are unlikely to implement this, at least in a form that is not strongly restricted. (I cannot speak for Gecko and WebKit, but both have already signaled worries about CSSOM complexity as I understand it, so increasing it further might not be the best course of action.)

[LeaVerou]

paving the way for nested inline styles too.

I hate to repeat myself, but we are unlikely to implement this, at least in a form that is not strongly restricted. (I cannot speak for Gecko and WebKit, but both have already signaled worries about CSSOM complexity as I understand it, so increasing it further might not be the best course of action.)

Given that many things that were previously hard to implement are currently being implemented, I'd be wary of choosing a CSSOM design that excludes this possibility. That would mean that once we do have nested inline styles, we'd end up with an inconsistent CSSOM.

[sesse]

That would mean that once we do have nested inline styles, we'd end up with an inconsistent CSSOM.

It already is inconsistent in exactly this fashion. Reading .cssText includes conditional group rules (such as @media or @container), but you cannot set them with .style (which forwards to .cssText). Making setting .style erase existing nested rules would just be even more inconsistent. (Also, “once we do have nested inline styles” may be construed as somewhat assumptious.)