feat(sampler): add min-max sampler function

[snukhulov]

Brief Information

This pull request is in the type of:

• bug fixing
• new feature
• others

What does this PR do?

Add min-max sampler function

Details

Before: What was the problem?

In some cases min-max sampling algorithm fits line data more naturally than the existing ones. For example, for a sine with a lot of data points min-max sampler is the best choice.

Example.
The original sine without sampling (200 000 data points).

Using 'average' sampler

Using 'lttb' sampler

Using 'max' sampler

Using 'min' sampler

After: How does it behave after the fixing?

Using new 'minmax' sampler

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 feat(sampler): add min-max sampler function echarts-doc#374

Misc

ZRender Changes

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

Related test cases or examples to use the new APIs

N.A.

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.

Document changes are required in this PR. Please also make a PR to apache/echarts-doc for document changes and update the issue id in the PR description. When the doc PR is merged, the maintainers will remove the PR: awaiting doc label.

[github-actions]

The changes brought by this PR can be previewed at: https://echarts.apache.org/examples/editor?version=PR-19279@4a63340

[Ovilia]

This seems to be a great idea. Can you explain what min-max mean first?

[snukhulov]

This seems to be a great idea. Can you explain what min-max mean first?

Thank you.
Like 'min' sampler that highlights an extremum (turning-point) at a local minimum and like 'max' sampler that highlights an extremum at a local maximum, 'minmax' sampler highlights the larger extremum (minimum or maximum extremum in absolute comparison) at a frame of data points.
'minmax' sampler is suitable for sine or other periodic data

[Ovilia]

Thanks for your explanation. Can you provide a link or something for further understanding? Is this a concept well-known or you invented it?

[snukhulov]

Thanks for your explanation. Can you provide a link or something for further understanding? Is this a concept well-known or you invented it?

It's my own implementation of the known min-max downsampling algorithm

[Ovilia]

Thanks! Please also add test cases in test/line-sampling.html. Please pay special attention both to the cases where there are negative values or not.

[snukhulov]

Thanks! Please also add test cases in test/line-sampling.html. Please pay special attention both to the cases where there are negative values or not.

I added the described test cases

[Ovilia]

Thanks for your contribution! This will be released in the next version.

[Ovilia]

Updated demo here

[echarts-bot]

Congratulations! Your PR has been merged. Thanks for your contribution! 👍

[sulivanganter]

Hi, @snukhulov

I think there might be something wrong. If you change the number of samples in your example to 100,000, updated exemple ,you can see the problem with returning just the absolute max value.

This is just a suggestion and has not been tested:

You need two values for each frame for min and max, something like this:

 minmax: function (frame) {
    let turningPointMaxValue = -Infinity;
    let turningPointMinValue = Infinity;

    for (let i = 0; i < frame.length; i++) {
        const originalValue = frame[i];

        if (originalValue > turningPointMaxValue) {
            turningPointMaxValue = originalValue;
        }
        
        if (originalValue < turningPointMinValue) {
            turningPointMinValue = originalValue;
        }
    }

    // Return an array with the min and max values
    return [turningPointMinValue, turningPointMaxValue];
}

[snukhulov]

Hi, @snukhulov

I think there might be something wrong. If you change the number of samples in your example to 100,000, updated exemple ,you can see the problem with returning just the absolute max value.

This is just a suggestion and has not been tested:

You need two values for each frame for min and max, something like this:

 minmax: function (frame) {
    let turningPointMaxValue = -Infinity;
    let turningPointMinValue = Infinity;

    for (let i = 0; i < frame.length; i++) {
        const originalValue = frame[i];

        if (originalValue > turningPointMaxValue) {
            turningPointMaxValue = originalValue;
        }
        
        if (originalValue < turningPointMinValue) {
            turningPointMinValue = originalValue;
        }
    }

    // Return an array with the min and max values
    return [turningPointMinValue, turningPointMaxValue];
}

Hi! Thank you for the research and the suggestion!

You're right about there are some problems with the proposed min-max sampler function. It's designed for a sine periodic data like y(x)=A*sin(x+n), but there are some sampling problems with aperiodic component of a signal (like y(x)=A*sin(x+n)+k).

Your suggestion looks great, but I have some doubts about returning type of your function. I think it only could be number, but couldn't be array of numbers.

I thought about improvement of the proposed min-max sampler, but there is motivation to keep it simple like other existing sampling functions.

But I'am still looking for an improvement of min-max sampler function that could solve sampling problem of data with aperiodic component like y(x)=A*sin(x+n)+k and could be simple at the same time.