[css-shape] Logical values for the `shape()` function

[stof]

The new shape() function has been defined only with physical commands. The current design solves the need for responsiveness compared to path() but still makes it hard to support LTR and RTL on the same website as it needs to redefine the whole shape for RTL to invert the physical dimensions (and I'm not even talking about vertical writing modes).

Would it be possible to define logical values for the commands of the shape ?

[noamr]

You can use logical values in shape() to * functions, e.g. hline to inline-end. Can you be more specific about what's missing?

[stof]

The documentation on https://developer.mozilla.org/en-US/docs/Web/CSS/basic-shape/shape does not show hline to inline-end as a supported value (neither in prose nor in the formal syntax which is derived from the spec).

And this would only works for cases where I need a line going until the end of the shape, which is quite restrictive. It also does not match syntax for anything else than vline and hline. All other commands work with coordinate pairs, which are all physical coordinates.

What I need is being able to define a shape which is reversed horizontally in a RTL context (basically being equivalent to having an X axis going from right to left). For that, it would require allowing to give a coordinate along the inline axis instead of along the x axis.
Note that my description here focuses on the RTL use case (where the inline axis is reversed and block axis stays the same), but full logical support should also support vertical writing modes (where inline and block axis are swapped). In such context, even the name hline does not make sense (as a line following the inline axis would become vertical).

[noamr]

The documentation on https://developer.mozilla.org/en-US/docs/Web/CSS/basic-shape/shape does not show hline to inline-end as a supported value (neither in prose nor in the formal syntax which is derived from the spec).

Sorry, for hline it's x-end and x-start.
You can use inline-end etc in the 2d commands.
This is in the spec: https://www.w3.org/TR/css-shapes/#typedef-shape-command
The "to" commands accept a "position".

Agreed that it can be documented better in mdn.

What I need is being able to define a shape which is reversed horizontally in a RTL context (basically being equivalent to having an X axis going from right to left). For that, it would require allowing to give a coordinate along the inline axis instead of along the x axis. Note that my description here focuses on the RTL use case (where the inline axis is reversed and block axis stays the same), but full logical support should also support vertical writing modes (where inline and block axis are swapped). In such context, even the name hline does not make sense (as a line following the inline axis would become vertical).

For this use case, would you expect the entire shape to be aligned with the direction, or be able to mix and match between logical and physical coordinates within the shape?

[stof]

My use case involves the entire shape to be aligned with the direction.

I fail to imagine a case where half of the commands would depend on the direction while the other half would be logical, as it means the LTR and RTL shapes would look totally different.

The "to" commands accept a "position".

this still only works when you need to draw until the inline-end, not when you need to draw until 58% of the width alongside the inline axis (at least as far as I can tell).

[noamr]

My use case involves the entire shape to be aligned with the direction.

I fail to imagine a case where half of the commands would depend on the direction while the other half would be logical, as it means the LTR and RTL shapes would look totally different.

The "to" commands accept a "position".

this still only works when you need to draw until the inline-end, not when you need to draw until 58% of the width alongside the inline axis (at least as far as I can tell).

Yes, agreed.
I wonder if we can define this as a value for <<geometry-box>>