Custom Visual Performance is lacking?

[jnat5]

I am currently trying to create my own custom networking visual using either of the following R packages (using the pbiviz rhtml template):

• https://datastorm-open.github.io/visNetwork/
• http://christophergandrud.github.io/networkD3/

However I'm finding performance is extremely slow, even on a very small amount of data (8 nodes), so that when users filter the report does not have a good UX.

The visualisation works fine in RStudio, hence I think it's an issue with Power BI. I've also tried only importing the packages I need, and doing absolutely minimal data processing (just selecting columns then removing any unused data frames out of memory). It's also slow in the Power BI Service.

Has anybody else experienced this isse with R custom visuals? Any help would be appreciated!

I was thinking maybe using the Javascript package underneath would be more performant - vis.js (i.e. the issue may be something with Power BI and R)?  

[v-shex-msft]

Hi @jnat5,

Did dm-p 's suggestions help with your scenario? if that is the case, you can consider Kudo or accept the helpful suggestions to help others who faced similar requirements.

If these also don't help, please share more detailed information to help us clarify your scenario to test.

How to Get Your Question Answered Quickly 

Regards,

Xiaoxin Sheng

[dm-p]

Hi @jnat5,

R visuals are delegated to an external interpreter by the visuals layer, which is web-based - this is either your local interpreter (if using Desktop) or an interpreter in the Service hosted by MS (if published).

There is a 'round trip' / 'cold start' cost often associated with this as it is managed over HTTP and processed into a static image (or HTML if  using RHTML in the custom visuals SDK) before being rendered in the DOM of the developer visual as such.

You will indeed get much better performance if you can work directly with the TypeScript SDKs as this is 'native' to Power BI.

I got into custom visual development in a similar way - started with R and found this performance bottleneck somewhat frustrating, so now I do all dev in TypeScript. Highly optimised code can be very performant.

If you don't want to go full web dev, and can manage with frameworks like Vega-Lite, I'm working on a visual called Deneb, which may bridge this gap.

Regards,

Daniel