Regression from v0.1.10 - 'nested' vars not being processed

[Paul Welsh]

In v0.1.10, you could use nested vars that would be parsed before being set to a mixin, e.g:

:root {
  --original-var: 4px;
  --nested-var: calc(2 * var(--original-var));
}
.class-name {
  @mixin mixinName var(--nested-var);
}

nested var would be resolved in the mixin as calc(2 * 4px)

As of v2, this no longer works and is instead simply passed through as calc(2 * var(--original-var)) (i.e not resolving the original var)

[Ivan Demidov]

@spacedawwwg Thank, i'll check if it's a good catch.

[Ivan Demidov]

@spacedawwwg No, this is unfortunately not a catch, I created an example and it passes the tests. You can make adjustments to the test so that it fails, or create an example that clearly falls with your result.

https://github.com/Scrum/postcss-at-rules-variables/pull/210/files

[Paul Welsh]

Oh, Ok - I only popped the example together based on my scenario but this is very weird then as I've had to revert back to back to v0.1.10 to get this working in my project. I wonder if it's 1) a config issue or 2) some kind of clash with another plugin

I'll do some more investigating and report back

[Paul Welsh]

I found the issue - It requires that any CSS vars used need to be included before they used now.

I wonder if this is a change in how PostCSS 8 works now?

[Ivan Demidov]

I wonder if this is a change in how PostCSS 8 works now?

I don't think my plugin and test uses PostCSS 8

[Paul Welsh]

@Scrum Sorry, I should have been more specific "included before they used" using the postcss-import plugin

Before hand the plugin used variables defined anywhere in the CSS (as if all the CSS had been previously parsed)

[Ivan Demidov]

I have created an example, please supplement it to reproduce your situation.

[Paul Welsh]

@Scrum popped it up here https://github.com/spacedawwwg/postcss-at-rules-demo

[Ivan Demidov]

This is due to the fact that the ad comes after the definition. Postscss makes only one iteration from top to bottom, in your case you will have to do multiple iterations when declaring a new value.

It seems to me that it is not entirely correct to make definitions after the declaration.

[Ivan Demidov]

now you can use the option declarationbywalk to find all declared variable declarations