Question about line stroke gradient direction

[TimeTravelPenguin]

I am not entirely sure if this is a Typst feature or something due to Cetz. In my project, I have lines that are coloured as a gradient from, say, red to blue. My algorithm relies on the idea that a line has a source and destination. Hence, I colour my lines based on the parent colour of each destination.

That is, if "from" is red and "to' is blue, then a line connecting "from" and "to" should be a gradient of red to blue.

The issue, however, is that this is not the behaviour that I see. In my project, I have such lines connecting hexagons on a grid. As an example:

This image shows the expected result. Now, when the algorithm draws a "to" that is leftward of the "from" x-coordinate, then the line is effectively "flipped". Instead of drawing from left to right, the line is conceptually rotated about the "from" position, to be a line from right to left. I would expect that a line from coordinate A to B (from "form" and "to") would account for the directionality of the line. The is, draw.line((0, 0), (1, 1), ...) is at an angle of 45º. In contrast, draw.line((1, 1), (0, 0), ...) is drawn at a 225º angle. I would expect that setting the fill of each to the same red-to-blue gradient would show two lines with gradients in the opposite direction. However, this is not what I have experienced. It appears as if the fill doesn't care for the direction of the line.

I tried a few different things to mitigate this issue, but the only nice result came from me explicitly making some of the gradients blue-to-red.

I can't add any additional images right now, so I hope this description makes sense. I want to know if this is an intended feature.

If you need additional imagery, please let me know and I will get some up when I hop on my machine in the coming day to two.

[johannes-wolf]

There seems to be a bug with gradient+path in Typst 12.0:

#let red-blue = gradient.linear(red, blue, relative: "parent")

#path((0cm, 0cm), (5cm, 0cm), stroke: red-blue + 5pt)

#cetz.canvas({
  import cetz.draw:*
  line((0,0), (5,0), stroke: red-blue + 5pt)
})

Does setting relative: "parent" work?

[johannes-wolf]

With relative: "self" I get two red lines, which is wrong. Using it with line (Typst) works.

[TimeTravelPenguin]

I did initially try using the relative parameter to try and get things working. I cannot recall the exact result, but from memory, changing the default caused it to produce incorrect results. However, that may be due to something else I was working on, so I cannot be completely certain.

I wonder if the bug has been documented. Thanks for your information!