kanaries / graphic-walker Goto Github PK

Mentioned here

Tableau allows changing the aggregation of datetime fields (so that you can drill down and up); is this functionality supported with pygwalker and if not can it be please added?
Kanaries/pygwalker#58

and voted here Kanaries/pygwalker#72

1. open graphic walker
2. click config button
3. open color palette for primary color
   (sometimes you can find the color rgba is ull)
4. click somewhere else

-> page crash

Reproducible by https://graphic-walker.kanaries.net/

The tooltip don't show the correct aggregated numbers. And another small problem is: even I drop several fields into the "details" zone, it will only show the latest one.

Change the dimension/measure in the Data view, but do not change in the visualization.

I know it is silly, which is to obey the design purpose of graphic-walker as a graphics system based on the grammar of graphics.

But there are people who like power BI more than tableau.....

GW performance problem.

• computation performance
• render performance
• RAM usage
• main thread tasks optimize

Where can I find a good example to know how the dataSource, rawFields and spec properties of the GraphicWalker component need to be specified? I did not find a good example for different types of fields. Thank you.

I found the new pivot table feature very powerful, but it forces me to put my measurement to either x-axis or y-axis, which is not intuitive.
As shown in the pic, I can put my measurements to either x or y axis. Sounds flexible but not intuitive for an excel user because in excel there is a field dedicated for measurements. Also, when the number of dimensions in the field list increases, it is easy to make the field list messy.

I looked into the pivot table feature in Tableau and Excel, and seems they are using different strategies.

• Tableau: user define the measurement in the "label" field of the "mark" panel, and only one measurement is allowed in each table. However, it allows user to further take advantage of the visual channels (color, size, etc.)

• Excel: user define the measurements in the "values" field and multiple measurements are allowed, but there is no such visual channel in a table.

In one word, they both use an extra field for measures, so I believe it would be better if KG does it in the same manner

Any idea / suggestion is appreciated 😊

When I converted a dimension variable to a measure one, its calculation will be "undefined". I think there should be a default calculation for it, may be sum?

I pasted an example below, where I used Bike Sharing dataset and tried to converted humidity from dimension to measure, and the x-axis turned out to be undefined(humidity)

Hi! I came across this very nice package, and I would like to integrate it into a platform for neuroscience research. Unfortunately, the platform requires 18. Are there any plans to update graphic-walker to support React 18? Thank you.

The filter channel got locked so that no more pills could be appended into filter channel and the one in filter channel could not be removed whatever either.

Reproduction

No stable reproduction found yet.

Drag a field into the filter channel, then try to (1) drag another into it or (2) remove the existed one.

Expected Behavior

1. A new filter rule will be appended.
2. The filter rule will be removed wherever you drop it outside the filter channel.

Actual Behavior

Nothing actually happened.

In pivot tables with multiple fields mapped to one axis, an option to repeat labels would help to make the table easier to scan.

e.g. in the following full adder truth table:

Is there someway to manage my zoom in only on the y-axis? The variability of my data is in the 0.001% range and it just looks like a straight line no matter how much I zoom because the zoom appears to be locked on both the x and y axis.

Also it appears to default to have y=0 on the bottom, my values are in the 10^6 and again it makes things look like straight line.

In the "data" tab, allow user to have a preview of the distribution, like it in RATH:

I recently embedded Graphic Walker in Jupyter Notebook (https://github.com/Kanaries/pygwalker).

Then I found Top Tabbars and Left Sidebars always cause a strange offset while dragging pills from the field list, see the GIF

Hi,

Key visuals of numbers I'd love to reproduce in graphic-walker - and then integrate, rely on row facets, see 'row facet of formulae' "picture".

https://observablehq.com/@declann/some-cashflows

(+ text mark/channel, issue #66 )

What are the challenges to add row/col facets?

Thanks!
Declan

Import data from CSV file, Chinese characters are displayed as garbled characters .

Issue: current bin transformation cannot be customized for bin number(which is defaulted by 10).
TODO:

• Allow the user to customize the bin's config (bin number)
• Default bin is still set as 10 in the field menu, the config panel is only for those who need customization (progressive design).

Hi,

graphic-walker features many marks but not text.

I find text mark critical for some key visuals, starting with the 'grid of formulae' "picture" here:

https://observablehq.com/@declann/some-cashflows

Would there be a difficulty supporting a text mark?

+Creating a separate issue for facets.

Thanks for the epic tool and really epic interface !
Very excited to discover this.

Regards,
Declan

users can set custom range of values on the x or y axis, I think make sense.

link pygwalker issue: Kanaries/pygwalker#192

for vs code application does not work? what about the dark theme?

<PureRenderer
    name={spec.name}
    visualConfig={spec.config}
    visualState={spec.encodings}
    type='remote'
    computation={computationCallback!}
/>

set spec.config.limit to 30, it can't work.

When adding the same field twice, with different aggregation selected, both charts start displaying the same data:

graphic-walker-notebook.txt

Hi, an option to use independent scales in facets would be useful, e.g. in the following chart we can't see the employees marks because the scale for other formulae is too big.

This is chart 2 in https://gist.github.com/declann/3e973796671f81d332c2ce36e3471e27

In this chart it's also useful to be able to remove legend as the color channel is redundant.

Is there a wish for this level of flexibility into graphic walker?

If not, I can probably do vega lite spec patching, supposing I embed graphic walker.

Thanks
Declan

In the code, there are some url links to point to many locations, that is making the product require a connection to the internet.

When we deploy it in a closed network environment, or as a desktop app in a bad network, the user will have bad experience or no experience at all(if he dosen't have a connection).

Thus we would like to compile the package with resolved links as assets, or some other method.

Thank you!

First chart

Create a new chart:

the data from the previous chart still exists.

It's cool how fast it works to generate the UI. I have one doubt if anyone can explain. I tried loading a dataframe with categorical variables but this library didn't worked. It didn't generate any dashboard. I tried after selecting only those columns which are int or float then it was working fine. Let me know am i wrong somewhere? New to tableau though.

From reddit user's comment

It seems sometimes gw failed when there are only nominal fields.

Dear Maintainers,

I have posted this issue here, double posting for visibility: Kanaries/pygwalker#256

I have also found that when using derived data points (like create a Year field from timestamp), then it cannot be placed into the filter and the entire app hangs. So it's impossible to select any date range at the moment.

Could you please fix the date filtering functionality? Thank you very much in advance.

I'm considering using the @kanaries/graphic-walker library for generic data visualisation, and I'm finding it extremely useful. However, I have a requirement where I need to handle large datasets efficiently. Currently, the library fetches the entire dataset upfront, which can be problematic when dealing with large amounts of data. I'm assuming that based on demos; please let me know if I'm wrong. We have a requirement to display spatial datasets which can go up to 50GB in size.

I wanted to inquire if it's possible to modify the workflow of the library to support the lazy loading of the dataset as a React Component. Here's the proposed workflow:

• Initially, only fetch the measures and dimensions from the data source. JSON API is preferable, but direct SQL connection is also doable.
• Allow the user to make chart selections such as the x-axis, y-axis, size, and color.
• A button (e.g., "Generate Visualization") should be used to trigger the lazy loading of the dataset. It can either filter via the JSON API,or fetch the actual group/subset via SQL query, based on the selected measures and dimensions.
• Once the dataset is fetched, generate the visualization using the selected chart options.

Please let me know if this is something that can be done using this library as an import without needing to customize the internals.

When I add a filter item, use vizStore.exportViewSpec() export json.

part of json:

{"type":"one of","value":"Set(\"GET\",\"POST\")"}

'GET' and 'POST' is my filter items

Nice project! Can you send a pull request to add this to https://vega.github.io/vega-lite/ecosystem.html?

1. Should remove

2. Should replace with console.error() or console.warn()

Found in https://youtu.be/3WjWeH3HIMo

It appears when pyg is used in jupyter notebook (JupyterLab 3.5.2) there are some inconsistencies in the theme. It works fine when I use JupyterLab Light, but with JupyterLab Dark it looks like below.
Can i somehow change pyg theme manually?

But the interface is a little wonky inside of my Jupyter notebook. All the buttons on the side look kinda faint, and the GUI looks small with everything at a small font size. Also there is a mouse delay. I'm sure a lot of these problems could be solved on my end, but the package seems to work a lot better in the Google or Kaggle notebooks.

From Reddit user's comment

Hi Kanaries team,
Hopefully this question as titled is not totally nonsense😂
Many thanks!

Hello,

thank you for the very nice tool, do you plan to extend the data source formats with parquet?

If I create multiple Graphic Walker instances showing different charts, their states will be identical.

Currently the charts are not deletable since created, so I wonder whether it is possible to allow users to discard the unwanted charts.

I have some questions about the opacity

• I set column "weight" as opacity, there are 2 values in this col, 1 and 255.
• A sample line in my df is {'name':'some name', 'month':'2022-02', 'weight':255}, name as Y and month as X, and weight means the opacity of the cell such as a heatmap.
• If this col is "dimension", then drag on opacity, it looks fine. But if set as "measure", it looks been calculated for the whole column.
• And even if I set the weight from 10 to 100, or 1 to 255, or 0.1 to 1, the opacities look the same, is it some normalization happened? Can I set a specific value for the opacity?

Hi

I want to use this library in an angular application. But I am unable to find any helpful implementation. I am also trying to use react component in angular but am still unable to use it properly, it gives errors for the library

Does anyone have any idea regarding this how can i implement this library in an angular application and start using this for different feature development purposes?

The image I exported

Ths is how it looked in my explorer

Noticed the background is transparent but image explorers usually make a white background, so that the text contents are hard to recognize. I wonder if there will be an option to control the background, or just to capture the appearance in my browser. thx.

Steps:

Load a topoJSON file.
Create a choropleth plot based on the loaded file.
Load another topoJSON file.

Expected Result:

The previous topoJSON data should be cleared, and the newly loaded file should be used for the plot.

I'm experimenting with gwalker inside Jupyter notebook.

df = pd.read_csv('./bike_path.csv', parse_dates=['datetime'])
gwalker = pyg.walk(df, hideDataSourceConfig=False)

There doesn't seem to be a way in the UI to shift a field from it's current designation (as a measure) to the other designation(dimension) or vice versa.

How do I control how an arbitrary column is interpreted by Gwalker.

And in a related question, if my original dataframe has a 'datetime' column, but in a visualiization I might want to use datetime.month or datetime.year, is there anyway to specify that use in the UI, or do I need to specify new columns in the df (month, year, etc) ?

Thanks

The image I exported

Ths is how it looked in my explorer

Noticed the background is transparent but image explorers usually make a white background, so that the text contents are hard to recognize. I wonder if there will be an option to control the background, or just to capture the appearance in my browser. thx.

On mobile/tablet no DND works for me in GW so there is no way to interact with GW.

Developing mobile DND or else an additional interaction approach like dropdowns would make this usable on mobile/tablets.