Fix list spacing and size (again)

[v-rob]

Fixes #10866 and fixes #10862 as well as making old coordinate system custom spacing more sane and making rounding behavior more consistent with the rest of formspec elements.

Devtest should look like this:

[kilbith]

Tested. I'm not exactly sure whether your PR changes the size or the spacing but the behavior has changed between upstream/HEAD and your branch.

Before

After

[kilbith]

Ok confirmed, this PR changes the slot size to make it bigger than currently.

[v-rob]

Sorry, still haven't had time to do all the testing I want to. I will look into that as soon as I can.

[v-rob]

I found the problem; i3 is using inconsistent units. Note how here, size and spacing are set to 0.96 and 0.15 respectively:

https://github.com/minetest-mods/i3/blob/89c2249bcba1145725042f80fc3f51c6ef14e4a2/init.lua#L2236

On the other hand, the gradient slot boxes use size and spacing of 1 and 0.1, respectively:

https://github.com/minetest-mods/i3/blob/89c2249bcba1145725042f80fc3f51c6ef14e4a2/init.lua#L2242

Changing the original line to have 1 and 0.1 for size and spacing fixes it. I guess that this was using the incorrect behaviour from pre-#10861, so it appeared incorrect in this PR.

However, your mod exposed one problem, which is that this PR uses round which gives better behaviour in some circumstances while worse in others. I will change it back to floor behaviour to keep it consistent with everything else. It might be possible to make the rounding behaviour better than either, but rounding is a problem shared universally by formspecs, so I'll leave it at floor for now.

[kilbith]

On the other hand, the gradient slot boxes use size and spacing of 1 and 0.1, respectively:

https://github.com/minetest-mods/i3/blob/89c2249bcba1145725042f80fc3f51c6ef14e4a2/init.lua#L2242

Nope, it has size as size which is 0.96, not 1.

[v-rob]

My mistake on that specific part, but my point still stands. The width and height of the slots and gradients are the same, but the spacing and positions are not. If you change the size and spacing to 1 and 0.1, they will align. Or, you could use 0.96 and make sure the position of the gradients also uses this value. Either way, the PR is not in error here.

[kilbith]

If you change the size and spacing to 1 and 0.1, they will align

I confirm that. But I don't get why the spacing has to be different between slot and box when size is <1 for slot + box, and why it can be kept the same when it's size 1.

[v-rob]

I confirm that. But I don't get why the spacing has to be different between slot and box when size is <1 for slot + box, and why it can be kept the same when it's size 1.

It doesn't; in the original PR, this was true (because it had buggy imgsize handling), but that was fixed in #10861. Both slot size and slot spacing are fractions of the coordinate and not dependent on each other. Here's an image of the current behaviour:

So, to fix i3's slot backgrounds to use any size/spacing, change line 2242:

-- From
fs(fmt("box", i + inv_x + (i * 0.1), inv_y, size, size, ""))
-- To
fs(fmt("box", i * size + inv_x + (i * spacing), inv_y, size, size, "")) -- Notice the added `* size` to fix positioning and using `spacing` instead of `0.1`

And then any value for size or spacing will work for both list and boxes. No change needs to be made for the list because that's not the issue.

(BTW, i3 is a really nice inventory, and it really uses the new formspec features to their potential.)

[v-rob]

Fixed rounding behaviour and did more thorough testing. Everything looks good to me.

[kilbith]

Tested the last commit and it does indeed fix some rounding issues I had when adjusting the size with floating values.

So I guess I can give a 👍 to this PR now.

[sfan5]

Looks good.