Skip merging same properties while allowing same selectors to be merged?

[corysimmons]

I want to merge same selectors, but want to keep same properties for fallback reasons.

[jakubpawlowicz]

You may want to use --skip-aggressive-merging and --skip-compacting together. Hope that helps!

[corysimmons]

I already tried --skip-aggressive-merging. It doesn't merge same selectors.

a {
  color: blue;
}
a {
  font-size: 16px;
  font-size: 1rem;
}

turns into

a {
  color: blue;
  font-size: 1rem;
}

when I'm looking for

a {
  color: blue;
  font-size: 16px;
  font-size: 1rem;
}

[corysimmons]

Or is there a way to specify specific properties to ignore?

[jakubpawlowicz]

Unfortunately, there's currently no way to do that. However it sounds like an interesting option to add.

[jakubpawlowicz]

However in this very case you just need to specify a compatibility mode --compatibility ie8, since rem units are not supported in IE<9.

[corysimmons]

I am specifying ie8.

Ah you're right! I was using cssnano which was obliterating it. Thanks!

[jakubpawlowicz]

👍

[jakubpawlowicz]

On master (4.0 soon):

cleancss -O2 --skip-aggressive-merging --compatibility ie8 ...

Note --skip-aggressive-merging will be gone in 4.0 too, so you can skip this option when #290 gets fixed.

[alexlamsl]

@jakubpawlowicz so with your example, all the level 1 options are still active?

On that note, given the recent proliferation of options at level 1, would it need something like the all shorthand in level 2 for convenience? Having said that, I realise not all of them are even boolean to begin with...

[jakubpawlowicz]

Fair point. The only two that are not boolean are roundingPrecision and specialComments. Could all perhaps apply to boolean options only? It's also interesting when you take the opposite case, e.g. all: 5, and try applying it to boolean options.

[alexlamsl]

The only useful case I can think of for all at level 1 would be to turn on only specific optimisation at level 2. Would level:{1:false,2:{duplicateRulesRemoving:true}} do that? If so, I guess it's good enough.

[jakubpawlowicz]

Yes, that's the use case. Alternatively it could be level:{1:{all:false},2:{duplicateRulesRemoving:true}} to keep both options as hashes.

Actually with the 2nd way you could still selectively turn on level 1 optimizations if needed too, e.g. level:{1:{all:false,tidySelectors:true},2:{duplicateRulesRemoving:true}}

[alexlamsl]

@jakubpawlowicz indeed that would be very powerful. Perhaps with #853 you can invert some of those boolean options in level 1 and change specialComments so they all accept false as meaning "leave the source alone"?

[jakubpawlowicz]

You made a good point about inverting options, because saying all: false on level 1 should mean "disable all" which is currently not the case.

[jakubpawlowicz]

The updated code on master should do the trick. Thanks @alexlamsl.

[corysimmons]

👏