Re: Binning of timeseries data, urgent question

[Ovilia <ov...@gmail.com>]

Thanks Ehsan,

The most essential part for these issues is the ability to aggregate data
(or named grouping/binning), which is scheduled in v5.4.

What you mentioned as "stacking" doesn't seem to behave like current
"stacking" [1] with Apache ECharts where each stacking element belongs to
different series and in most cases have different colors. As for the time
series, if data should be grouped into monthly, each month should have only
one bar with the sum of the data within the month, rather than stacking
bars each of which represents a single data.

If you have different ideas about this, please let me know and let's see
what we can provide for most reasonable situations.

[1]
https://echarts.apache.org/examples/en/editor.html?c=bar-y-category-stack

Thanks

*Ovilia*


On Tue, Dec 7, 2021 at 2:14 PM Ehsan Saberian <eh...@gmail.com>
wrote:

> Hi Ovilia,
>
> Thanks for your answer. We need to use the time axis so that the intervals
> get auto handled by the chart when zooming-in/out. With the category axis,
> this requirement can not be met.
> I decided to keep using the chart with time and wait for the fix in the
> next version.
> I also added another issue (#16177) to the github repo about the
> binning/Grouping of time series data. It's about stacking adjacent bar
> charts when the xAxis type is time in different zooms like the example i
> provided from FusionTime in
> https://codesandbox.io/s/fusion-charts-forked-8iv04?file=/src/chart.js .
> It's different from the #16099 which is about the width of the bars.
>
> Regards,
> -Ehsan
>
>
>
> On Mon, Nov 29, 2021 at 6:42 AM Ovilia <ov...@gmail.com> wrote:
>
>> Hi Ehsan,
>>
>> This behavior is because the width of bars in time series are decided by
>> the number of bars in the viewport but ECharts doesn't provide a width
>> option in the value (time) unit. That means, when zooming, the width
>> stays the same as long as the number of bars does not change.
>>
>> If you have urgent requirements about it, I would suggest using category
>> axis instead of time series so that you can control how data should be
>> "binned" into categories.
>>
>> Thanks
>>
>> *Ovilia*
>>
>>
>> On Mon, Nov 29, 2021 at 11:03 AM Ehsan Saberian <eh...@gmail.com>
>> wrote:
>>
>>> Hello,
>>> I posted an issue in your github repo about 5 days ago. It's about
>>> binning
>>> of time series data whe XAxis type is 'time'.
>>> Solving the problem is very urgent for our team project.
>>> I would appreciate it if you can check that soon. The issue link is:
>>> https://github.com/apache/echarts/issues/16099
>>>
>>> Regards,
>>> -Eh
>>>
>>

Re: Binning of timeseries data, urgent question

[Ehsan Saberian <eh...@gmail.com>]

Hi Ovilia,

Thanks for your consideration. I'll be waiting for V5.4 and will be in
touch.

Regards,
-Ehsan

On Tue, Dec 7, 2021 at 1:42 PM Ovilia <ov...@gmail.com> wrote:

> Hi Ehsan,
>
> Thanks for your detailed description and yes, what you described is
> exactly the same
> as what I'm suggesting. We might reach you for further discussion if
> necessary
> when we implement the feature.
> It's scheduled in v5.4 so you may not expect to see it in a month. If you
> are interested
> in making a pull request, it would be certainly appreciated and we could
> offer more
> help about that.
>
> BTW, it would be nice if we could CC our mailing list
> dev@echarts.apache.org
> so that anyone else who is interested in the topic can also join the
> discussion. :)
>
> Thanks
>
> *Ovilia*
>
>
> On Tue, Dec 7, 2021 at 4:02 PM Ehsan Saberian <eh...@gmail.com>
> wrote:
>
>> Yes it's true. The core functionality of them is the data aggregation on
>> different zooms.
>>
>> About what I mentioned about stacking, I think it can be similar in some
>> ways, because we also need aggregating and then stacking different time
>> series (different colors) when they're adjacent.
>> Suppose we have 3 different main series (which get different colors) and
>> each series has some date values.
>>
>> firstTimeSeries data: (blue color)
>> (1988, 7, 1), 100
>> (1988, 7, 4), 135
>> (1988, 8, 2), 200
>> (1988, 9, 3), 300
>> (1988, 10, 4), 400
>>
>> secondTimeSeries data: (green color)
>> (1988, 7, 2), 100
>> (1988, 8, 3), 200
>> (1988, 9, 4), 300
>> (1988, 10, 5), 400
>>
>> thirdTimeSeries data: (yellow color)
>> (1988, 7, 3), 100
>> (1988, 8, 4), 200
>> (1988, 9, 5), 300
>> (1988, 10, 6), 400
>>
>> In this case when we're seeing the chart in monthly binning, We expect
>> that for example for the 7th month we have a bar stacked of three colors
>> which the blue part of the stack has 100 + 135 count, the green part has
>> 100 count and the yellow part has 100 count.
>> I also created a codesandbox example for what i'm describing.
>> https://codesandbox.io/s/echarts-zoomin-timeseries-sktvb?file=/src/App.js
>>
>> to be more precise, instead of seeing this:
>> [image: image.png]
>>
>> we expect to see something like this (for the 7th month)
>>
>> [image: image.jpeg]
>>
>> Thanks,
>> -Ehsan
>>
>>
>> On Tue, Dec 7, 2021 at 9:59 AM Ovilia <ov...@gmail.com> wrote:
>>
>>> Thanks Ehsan,
>>>
>>> The most essential part for these issues is the ability to aggregate
>>> data (or named grouping/binning), which is scheduled in v5.4.
>>>
>>> What you mentioned as "stacking" doesn't seem to behave like current
>>> "stacking" [1] with Apache ECharts where each stacking element belongs to
>>> different series and in most cases have different colors. As for the time
>>> series, if data should be grouped into monthly, each month should have only
>>> one bar with the sum of the data within the month, rather than stacking
>>> bars each of which represents a single data.
>>>
>>> If you have different ideas about this, please let me know and let's see
>>> what we can provide for most reasonable situations.
>>>
>>> [1]
>>> https://echarts.apache.org/examples/en/editor.html?c=bar-y-category-stack
>>>
>>> Thanks
>>>
>>> *Ovilia*
>>>
>>>
>>> On Tue, Dec 7, 2021 at 2:14 PM Ehsan Saberian <eh...@gmail.com>
>>> wrote:
>>>
>>>> Hi Ovilia,
>>>>
>>>> Thanks for your answer. We need to use the time axis so that the
>>>> intervals get auto handled by the chart when zooming-in/out. With the
>>>> category axis, this requirement can not be met.
>>>> I decided to keep using the chart with time and wait for the fix in the
>>>> next version.
>>>> I also added another issue (#16177) to the github repo about the
>>>> binning/Grouping of time series data. It's about stacking adjacent bar
>>>> charts when the xAxis type is time in different zooms like the example i
>>>> provided from FusionTime in
>>>> https://codesandbox.io/s/fusion-charts-forked-8iv04?file=/src/chart.js
>>>> .
>>>> It's different from the #16099 which is about the width of the bars.
>>>>
>>>> Regards,
>>>> -Ehsan
>>>>
>>>>
>>>>
>>>> On Mon, Nov 29, 2021 at 6:42 AM Ovilia <ov...@gmail.com> wrote:
>>>>
>>>>> Hi Ehsan,
>>>>>
>>>>> This behavior is because the width of bars in time series are decided
>>>>> by
>>>>> the number of bars in the viewport but ECharts doesn't provide a width
>>>>> option in the value (time) unit. That means, when zooming, the width
>>>>> stays the same as long as the number of bars does not change.
>>>>>
>>>>> If you have urgent requirements about it, I would suggest using
>>>>> category
>>>>> axis instead of time series so that you can control how data should be
>>>>> "binned" into categories.
>>>>>
>>>>> Thanks
>>>>>
>>>>> *Ovilia*
>>>>>
>>>>>
>>>>> On Mon, Nov 29, 2021 at 11:03 AM Ehsan Saberian <
>>>>> ehsan.saberyan@gmail.com> wrote:
>>>>>
>>>>>> Hello,
>>>>>> I posted an issue in your github repo about 5 days ago. It's about
>>>>>> binning
>>>>>> of time series data whe XAxis type is 'time'.
>>>>>> Solving the problem is very urgent for our team project.
>>>>>> I would appreciate it if you can check that soon. The issue link is:
>>>>>> https://github.com/apache/echarts/issues/16099
>>>>>>
>>>>>> Regards,
>>>>>> -Eh
>>>>>>
>>>>>

Re: Binning of timeseries data, urgent question

[Ovilia <ov...@gmail.com>]

Hi Ehsan,

Thanks for your detailed description and yes, what you described is exactly
the same
as what I'm suggesting. We might reach you for further discussion if
necessary
when we implement the feature.
It's scheduled in v5.4 so you may not expect to see it in a month. If you
are interested
in making a pull request, it would be certainly appreciated and we could
offer more
help about that.

BTW, it would be nice if we could CC our mailing list dev@echarts.apache.org
so that anyone else who is interested in the topic can also join the
discussion. :)

Thanks

*Ovilia*


On Tue, Dec 7, 2021 at 4:02 PM Ehsan Saberian <eh...@gmail.com>
wrote:

> Yes it's true. The core functionality of them is the data aggregation on
> different zooms.
>
> About what I mentioned about stacking, I think it can be similar in some
> ways, because we also need aggregating and then stacking different time
> series (different colors) when they're adjacent.
> Suppose we have 3 different main series (which get different colors) and
> each series has some date values.
>
> firstTimeSeries data: (blue color)
> (1988, 7, 1), 100
> (1988, 7, 4), 135
> (1988, 8, 2), 200
> (1988, 9, 3), 300
> (1988, 10, 4), 400
>
> secondTimeSeries data: (green color)
> (1988, 7, 2), 100
> (1988, 8, 3), 200
> (1988, 9, 4), 300
> (1988, 10, 5), 400
>
> thirdTimeSeries data: (yellow color)
> (1988, 7, 3), 100
> (1988, 8, 4), 200
> (1988, 9, 5), 300
> (1988, 10, 6), 400
>
> In this case when we're seeing the chart in monthly binning, We expect
> that for example for the 7th month we have a bar stacked of three colors
> which the blue part of the stack has 100 + 135 count, the green part has
> 100 count and the yellow part has 100 count.
> I also created a codesandbox example for what i'm describing.
> https://codesandbox.io/s/echarts-zoomin-timeseries-sktvb?file=/src/App.js
>
> to be more precise, instead of seeing this:
> [image: image.png]
>
> we expect to see something like this (for the 7th month)
>
> [image: image.jpeg]
>
> Thanks,
> -Ehsan
>
>
> On Tue, Dec 7, 2021 at 9:59 AM Ovilia <ov...@gmail.com> wrote:
>
>> Thanks Ehsan,
>>
>> The most essential part for these issues is the ability to aggregate data
>> (or named grouping/binning), which is scheduled in v5.4.
>>
>> What you mentioned as "stacking" doesn't seem to behave like current
>> "stacking" [1] with Apache ECharts where each stacking element belongs to
>> different series and in most cases have different colors. As for the time
>> series, if data should be grouped into monthly, each month should have only
>> one bar with the sum of the data within the month, rather than stacking
>> bars each of which represents a single data.
>>
>> If you have different ideas about this, please let me know and let's see
>> what we can provide for most reasonable situations.
>>
>> [1]
>> https://echarts.apache.org/examples/en/editor.html?c=bar-y-category-stack
>>
>> Thanks
>>
>> *Ovilia*
>>
>>
>> On Tue, Dec 7, 2021 at 2:14 PM Ehsan Saberian <eh...@gmail.com>
>> wrote:
>>
>>> Hi Ovilia,
>>>
>>> Thanks for your answer. We need to use the time axis so that the
>>> intervals get auto handled by the chart when zooming-in/out. With the
>>> category axis, this requirement can not be met.
>>> I decided to keep using the chart with time and wait for the fix in the
>>> next version.
>>> I also added another issue (#16177) to the github repo about the
>>> binning/Grouping of time series data. It's about stacking adjacent bar
>>> charts when the xAxis type is time in different zooms like the example i
>>> provided from FusionTime in
>>> https://codesandbox.io/s/fusion-charts-forked-8iv04?file=/src/chart.js .
>>> It's different from the #16099 which is about the width of the bars.
>>>
>>> Regards,
>>> -Ehsan
>>>
>>>
>>>
>>> On Mon, Nov 29, 2021 at 6:42 AM Ovilia <ov...@gmail.com> wrote:
>>>
>>>> Hi Ehsan,
>>>>
>>>> This behavior is because the width of bars in time series are decided by
>>>> the number of bars in the viewport but ECharts doesn't provide a width
>>>> option in the value (time) unit. That means, when zooming, the width
>>>> stays the same as long as the number of bars does not change.
>>>>
>>>> If you have urgent requirements about it, I would suggest using category
>>>> axis instead of time series so that you can control how data should be
>>>> "binned" into categories.
>>>>
>>>> Thanks
>>>>
>>>> *Ovilia*
>>>>
>>>>
>>>> On Mon, Nov 29, 2021 at 11:03 AM Ehsan Saberian <
>>>> ehsan.saberyan@gmail.com> wrote:
>>>>
>>>>> Hello,
>>>>> I posted an issue in your github repo about 5 days ago. It's about
>>>>> binning
>>>>> of time series data whe XAxis type is 'time'.
>>>>> Solving the problem is very urgent for our team project.
>>>>> I would appreciate it if you can check that soon. The issue link is:
>>>>> https://github.com/apache/echarts/issues/16099
>>>>>
>>>>> Regards,
>>>>> -Eh
>>>>>
>>>>