fix(polar): angleAxis extent when boundaryGap is false

[Ovilia]

Brief Information

This pull request is in the type of:

• bug fixing
• new feature
• others

What does this PR do?

When angleAxis in the polar system has boundaryGap: false, the extent is adjusted since this commit so that polar bars don't overlap each other. The main idea is to make extra space according to data count. But it may overkill in some situations.

Fixed issues

NA.

This PR is not a fix to #20172 , which is not a bug.

Details

Before: What was the problem?

Consider the case when startAngle: 90, endAngle: -90, before this PR, it has the result of:

This is not as expected because a developer set startAngle: 90, endAngle: -90 would expect the range to be half a circle.

After: How does it behave after the fixing?

The extent of angleAxis with boundaryGap: false should only be adjusted if it has the potential to overlap, that is to say, startAngle: 90, endAngle: -90 should still get half a circle:

while startAngle: 90, endAngle: -270 should get 3/4 of a circle:

Please also not that if the range of startAngle and endAngle itself is over 360 degrees, it won't adjust to be smaller than a circle because it is not expected to.

Document Info

One of the following should be checked.

• This PR doesn't relate to document changes
• The document should be updated later
• The document changes have been made in apache/echarts-doc#xxx

Misc

ZRender Changes

• This PR depends on ZRender changes (ecomfe/zrender#xxx).

Related test cases or examples to use the new APIs

I added a test case under test/bar-polar-basic-radial.html.

I've run with npm run test:visual and it passes all but the above case.

Others

Merging options

• Please squash the commits into a single one when merging.

Other information

[echarts-bot]

Thanks for your contribution!
The community will review it ASAP. In the meanwhile, please checkout the coding standard and Wiki about How to make a pull request.

The pull request is marked to be PR: author is committer because you are a committer of this project.

[Ovilia]

We can see that the old code takes 360 as the range of endAngle to startAngle, which is not always the case.

[github-actions]

The changes brought by this PR can be previewed at: https://echarts.apache.org/examples/editor?version=PR-20184@1c08b2f

[100pah]

In the current modified approach, this cases are not reasonable enougth:

(already have a overlap between bar C and bar D) (endAngle changes a little (from -195 to -198) but the appearance changes significantly.)

I think the "adjusting the extent" is a compromise. It's counter intuitive (the adjusted endAngle is not what use specified) but with out that user have avoid overlap manually.

I just try to modify it: make it keep at a "max span" when there is no enough space.

    // Fix extent of category angle axis
    if (angleAxis.type === 'category' && !angleAxis.onBand) {
        const extent = angleAxis.getExtent();
        let spanAngle = normalizeAngle((normalizeAngle(extent[1]) + 360 - normalizeAngle(extent[0])));
        if (angleAxis.inverse) {
            spanAngle = 360 - spanAngle;
        }
        const spanLimit = 360 - 360 / (angleAxis.scale as OrdinalScale).count();
        if (spanAngle >= spanLimit) {
            extent[1] = extent[0] + (angleAxis.inverse ? -1 : 1) * spanLimit;
            angleAxis.setExtent(extent[0], extent[1]);
        }
    }
    // FIXME: this kind of function should be placed in some common util file? (or similar one already exists?)
    // Normalize an angle value to `[0, 360)`.
    function normalizeAngle(val: number) {
        val = val % 360;
        val < 0 && (val += 360);
        return val;
    }

(it's just a illustrative code to show my current understanding.)

But about clockwise (inverse). I haven't understood the logic yet.
The current behavior (before this PR modified) seems weird:

[Ovilia]

I'm not quite sure if we should use normalizeAngle because it erases some information. For example, with extent: [90, -450], it's more intuitive to understand this as a circle and a half. With normalizeAngle((normalizeAngle(extent[1]) + 360 - normalizeAngle(extent[0]))); we get => normalizeAngle(90 + 360 - 270) => normalizeAngle(180) => 180, which is a half circle. We could say we don't support a circle and a half as spanAngle so that one circle is used, but a half circle? I don't think this is expected with extent: [90, -450].

So I implemented this with a simpler logic: let spanAngle = (extent[1] - extent[0]) * (angleAxis.inverse ? -1 : 1);, which I tested to be right. Correct me if I'm wrong.

As for clockwise: false, I think this should be regarded as another bug because it's not introduced in this PR so I suggest fixing this in future PRs.