Block styles: reflect inherited global styles values in block appearance controls

[ramonjd]

The tracking issue logs tasks associated primarily with the following issue:

• Design tools: Better handling of unset / inherited values #43082

Blocks can "inherit" styles from theme.json and user global styles. These styles are loaded into the Block Editor as global styles CSS, meaning that block will display its inherited styles in the editor canvas.

A block's design tool controls however only reflect applied block-level styles and not any global styles values.

Consequently, even though a block may appear to have, for example, red text in the Block Editor canvas (assuming that style is defined in theme.json), the block's color controls will feature empty values.

To avoid confusion to users, a block's style controls should reflect, as much as possible, the styles that a block inherits.

Constraints

The tracking issue is deliberately limited in scope to displaying values from the merged theme.json/global styles object.

Enqueued stylesheets and theme.json custom CSS for example will affect a block's appearance in the editor canvas but they cannot be reliably parsed. Similarly, other styles inherited via the CSS cascade such as a parent block's inner element styles, or a parent with block style variation applied's inner block type styles, would likely require querying the browser element itself.

A possible "Phase 2" could be to test the performance of extracting values using getComputedStyle().

Tasks (WIP)

• Design a global styles inheritance hierarchy map, e.g., Root Global > Root Element > Block > Block Element > Block style variations
• Create a hook/function to parse and structure merged global styles config values (base + theme + user) according to an inheritance map.
• Users should have "read" access to global styles endpoints
• Ensure that preset values such as colors and font sizes are correctly parsed and reflected in the controls.
• Ensure that dynamic values such as theme.json "ref" pointers or relative paths are resolved. E.g., see this comment.
• Migrate block controls to display inherited values
  • Color
  • Spacing
  • Typography
  • Background
  • Border
  • Shadow
  • Layout
  • ...

A corollary, and follow-up issue, is to enable "unsetting" global styles at the block level:

• Design tools: Better handling of unset / inherited values #43082

Related

• Local block appearance controls should reflect values from global styles / theme.json #37752
• Design tools: show "computed" style values #57296
• Design tools: Better handling of unset / inherited values #43082
• Global Styles and Theme.json: Next Steps #62686
• Explore improvements to Styles panel in site editor #53483

[ramonjd]

Just jotting down some notes...

I've been playing around with a simple merge function (using deepmerge.all) to get an idea of the various inheritances.

1. mergedBaseThemeUserStyles[styleProperty] 
2. mergedBaseThemeUserStyles.elements[elementName][styleProperty]
3. mergedBaseThemeUserStyles.blocks[blockName][styleProperty]
4. mergedBaseThemeUserStyles.blocks[blockName].variations[variationName][elementName][styleProperty]
5. mergedBaseThemeUserStyles.blocks[blockName].variations[variationName][styleProperty]

styleProperty is an optional key as I think we pass around the full style object for blocks.

Without taking into account any special cases, it kinda works in the simple unit test I wrote, with some questions/caveats:

• The hook needs to get block variation name using the block name. See packages/block-editor/src/hooks/block-style-variation.js
• Should it be smart enough to map block names to corresponding element styles. Is there an existing map?
• Should we resolve dynamic values here? E.g., "ref" and relative paths.
• To display preset values in block style controls, we need to decode them. Probably at the individual hook level?

Missing from the above hierarchy, is the following case, even if it is one:

1. We want the inherited styles for block 1
2. block 1 lives inside block 2
3. block 2 has style variation X applied to it
4. style variation X contains styles for ['block 1']

Is that what we generally name style variation overrides?

[aaronrobertshaw]

Some good questions there, I'm still in the process of getting my head around all the permutations here.

Missing from the above hierarchy, is the following case, even if it is one:

In addition to what you coined as "style variation overrides", there's a similar case with a parent block instance that defines element styles.

In both cases, we'd need to be able to determine that some ancestor had applied either a block style variation or instance element styles. Different variations or element styles could be applied at different levels within the ancestor tree requiring multiple merges to arrive at something close to the final set that is seen in the editor canvas.

This sounds rather complex and resource-intensive. Further complicating this scenario, block style variations can define element styles as well as element styles within block types, within that variation.

As noted in this issue's constraints section, this is probably best delayed until we can validate approaches for the simpler use cases outlined via the above merge strategy.

[jasmussen]

Thanks for looking at this. As you start to dive in, I'd encourage looking for quick wins we can ship. Perfection might be good eventually, but we can also start with less, potentially. So even if one or other item on this list is addressed, it'll leave things in a better place.

[ramonjd]

As you start to dive in, I'd encourage looking for quick wins we can ship. Perfection might be good eventually, but we can also start with less, potentially.

Agree 100%

It's a complex area this one, and it looks like there'll be compromises to contend with at most corners, e.g., how to represent inherited element styles such as h1, so the smaller the steps, the better! 😄

[aaronrobertshaw]

I also agree with the "small steps and quick wins" approach.

My only concern is getting into a situation where the values in the block inspector controls appear even more hit-and-miss than they do now. This isn't to suggest we shouldn't proceed as planned but more that I'd like to keep in mind the problem we're trying to solve for the end user.