W3C home > Mailing lists > Public > www-style@w3.org > April 2012

[css-variables] Should var properties respect !important?

From: Tab Atkins Jr. <jackalmage@gmail.com>
Date: Mon, 2 Apr 2012 18:31:47 -0700
Message-ID: <CAAWBYDAP24rng184yDSNqcyT+hjx7gbbtTjjZv9cPi5w839Vxg@mail.gmail.com>
To: www-style list <www-style@w3.org>
While implementing the parsing side of Variables recently, Luke
realized that it's unclear whether an !important token at the end of a
variable's value actually makes it important, or if it's just treated
as part of the variable's value.  I'm not sure which option is the
best, but here are some arguments for both.

1) Treat !important as part of the variable's value

This option reduces the amount of interpretation on the var property's
value, which is currently *almost* entirely uninterpreted (only var()
references are expanded within it).

Within (1), there are three sub-options:

1a) If variable expansion places an !important at the end of a
property value, pay attention to it

This makes sense because it's consistent with "expand variables then
parse like normal", but I don't think it's workable.  Expansion
doesn't occur until computed-value time, long after the cascade is
run, and !important affects the cascade.  Thus, this appears to be a
major violation of the value computation layering.

1b) If variable expansion produces an !important at the end of a
property value, ignore it.  (!important appearing anywhere else is a
syntax error, like usual)

This is the minimal change - you expand variables, parse like normal,
and then simply ignore a trailing !important because it's too late to
act on.

1c) If variable expansion produces an !important token anywhere, ignore it.

This is somewhat inconsistent with "expand variables then parse like
normal", but it has the benefit of being stable across small edits -
adding something to a property such that the "!important"-carrying var
is no longer the last component won't cause a sudden mysterious syntax

2) Treat !important like normal, and have it effect the cascade (and
not be part of the variable's value)

The benefit of this is that you have the normal !important behavior,
which may be less surprising.  Though using !important is *usually* a
bad idea, on occasion it's worthwhile, and this preserves that ability
for those cases.

This also has sub-options:

2a) If a non-trailing !important appears in a var property, it's a syntax error.

This prevents us from getting into the (2b) and (2c) situations
entirely, but it means more preprocessing of variables.  Every
restriction we place on a variable's value reduces its utility as a
generic cascade-using information carrier, which I consider an
important use-case (that is, using var properties to attach data to
the DOM for JS purposes, rather than using them in the var()

2b) and 2c) Non-trailing !important in the var property isn't a syntax
error; it's treated like (1b) or (1c) when expanded

Like I said, I'm not strongly in favor of any option.  Thoughts?

Received on Tuesday, 3 April 2012 01:32:35 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 17:20:52 GMT