[ENHANCEMENT] Enable the free combination of dimensions and metrics in a two-dimensional data table

[123777tw]

As we known, A Power BI visualization can receive a two-dimensional data table consisting of several dimensions or metrics, such as this attachment:

grid8F318104F2EA47E4BC602709F007C1C9.1.0.0.pbiviz.zip

Due to website restrictions, I temporarily changed the extension to ".zip".

Advocating this approach has two benefits:

1. It naturally has a table layout, making it easier to organize interface styles, and even without any HTML code, it still has a natural table meaning;
2. Power BI has a limit of approximately 64kB for each cell of incoming data, and a two-dimensional free-form table will allow for more code to flood in, thus creating a grand visual effect.

[dm-p]

Hi Thomas, Thanks for your request. If you're looking for something like the attached visual, where you add multiple columns and measures to the data roles and get an HTML table, this is best done as a dedicated custom visual (like the one attached).

Adding a separate workflow and functionality for a basic table display goes against the grain of how the visual is intended to work and would take a lot of effort to implement without potentially breaking the visual for all users. There would also be a lot of consideration as to how to customize this output - do we want dedicated properties, for example? If so, this has a high administrative overhead and will continue to be challenging to support. We would also need to manage a separate data role for this mode, which then increases friction for users (data roles are very specific for how a visual should be designed to work, and there is not a lot of ability to customize how they behave outside of some simple conditional checks).

Power BI's limits are only a problem for columns, which are 32KB. Measures have a ~2MB limit, which is typically more than enough for most dynamic HTML cases.

If I have misunderstood the request, please provide a more detailed use case with intended usage scenarios, and I can think about this a bit more.

[123777tw]

Hi Thomas, Thanks for your request. If you're looking for something like the attached visual, where you add multiple columns and measures to the data roles and get an HTML table, this is best done as a dedicated custom visual (like the one attached).

Adding a separate workflow and functionality for a basic table display goes against the grain of how the visual is intended to work and would take a lot of effort to implement without potentially breaking the visual for all users. There would also be a lot of consideration as to how to customize this output - do we want dedicated properties, for example? If so, this has a high administrative overhead and will continue to be challenging to support. We would also need to manage a separate data role for this mode, which then increases friction for users (data roles are very specific for how a visual should be designed to work, and there is not a lot of ability to customize how they behave outside of some simple conditional checks).

Power BI's limits are only a problem for columns, which are 32KB. Measures have a ~2MB limit, which is typically more than enough for most dynamic HTML cases.

If I have misunderstood the request, please provide a more detailed use case with intended usage scenarios, and I can think about this a bit more.

Thank you for your positive response to the issue!

A dedicated custom visual may be a good idea, but I think it shouldn't stop at this attachment. I mean, it could be your brand new visualization, but not mine.

Talking of this attachment, it is a very old workaround I created, and I even forgot where its source code is located. But I still remember it was very simple, it hardly used any defined APIs from Power BI. You can even see that it is less than 40kB in size. It simply accepts the data table provided by Power BI via a variable and does its own thing without any feedback, such as cross filtering. I once wanted it to have similar feedback abilities, but all efforts ended in failure. I am not a professional in web front-end development. I have figured out the basic working principles of JS based solely on my interests and hobbies, and I am very wary of TS, a new and more rigorous language. The only fortunate thing is that TS is a superset of JS. You can easily imagine that I just wrote a few lines of JS code in 3 TS files and compiled this little thing. And the reason I did all of this was simply because I hadn't seen something like the one you created here at that time. When I saw this HTML Content recently, I had a feeling of regret for meeting it too late.

32kB is actually very small, it is even less than one tenth the size of the "ECharts.min.js" file. If users want to perform a complex design through the ETL process, they must split their code and allocate it to multiple cells; But when they want to stack some sections vertically again, it will be difficult to use HTML Content which only has one column of data input.
There won't be very typical cases here, but if you rely heavily on one to think about problems, I think you can use a calendar as an example. Generally, a calendar page needs to display every day of a month, with a fixed number of columns of 7 and dynamic number of rows of 4-6. Considering using your HTML Content, you may need to split a cell into a row with 7 columns, or loop reference every 7 cells and convert them into a row. If you choose the former, you will face the dilemma of dividing 32kB among 7 cells; Otherwise, you will need to design additional citation rules to ensure that every day is placed in the correct position.
It's difficult for you to make a third reasonable choice because the number of rows in the calendar is dynamic, and you need to clear out completely blank rows. As for the case you mentioned where a single measure can accommodate approximately 2MB - firstly, a single measure needs to define a complete vertical region at a time, which can be difficult to handle in practical use if you happen to be designing a weakly iterative interface; Secondly, your HTML Content only allows you to introduce one measure, and 2MB is still too cramped for the entire visualization area.
You may have always thought that such capacity is sufficient in most cases, but that may just be because your users are not yet awakened. Based on the ECharts and calendar I mentioned earlier, I can easily add several ECharts, pie charts, bar charts, line charts, or even all of them to my calendar every day cell. As long as I don't worry about performance issues, I can endlessly enrich my visualization.
As for “the grain of how the visual is intended to work” of Power BI, I don't think it is very important. Separate users who may experience "friction" by create a brand new visulization. Or further improve your website “htmlcontent.com” to minimize the “friction” as much as possible. A visualization can act like a plugin, breaking away from its original audience by endowing the host program with new functionalities. I always believe that a good tool should be able to maximize the value of its users without limiting their original abilities. No new definitions are needed, just unlock more restrictions and unleash more possibilities.
Of course, I may seem like I'm just making sarcastic remarks. Forcing my ideas into your work will definitely encounter various difficulties. And it seems like I haven't contributed any code to the repository, and I haven't even actually bought you a cup of coffee. I am just paranoid that you may accept new ideas based on your open source behavior. It is all up to you, but I still hope to wish you good luck in my broken English!

[123777tw]

Ah, I suddenly thought of something like data URI schema which might take up a lot of space, though it may not be a good practice :)