Frontend Performance

[prastut]

@digi0ps we need to figure out how to reduce load on the frontend. Mainly if we have 3 hours or long matches we can't poll the server for the data since the user would only be interested in an overall sense of view. Some things that I am thinking on top of my head:

• Line Chart performance wouldn't matter because it produces a point in 30 seconds. Still it would be a lot when looking over a one-day cricket match.

• Scatter Chart performance is a big issue for now. The backend people are sorting filtering since there are a lot of tweets. Still no amount of filtering would solve the problem of data since that's our USP. Right now we are thinking in terms of performance but from the POV of our product, having data helps -> exactly similar to having number of likes and comments on Facebook. Our backend can process that data, our frontend should also do.

The hard solution is figuring out layers of loading. For eg on initial zoom, load data for the last 30 minutes:

• Line Chart won't have any problems.
• Depending on user zoom and pan interactions we can have tweets loading in place as well as when the user pan.
• On different zoom levels we can have more data coming in to refine the view. For eg:
  • user level at zoom 1 would see the entire line chart.
  • No point loading a very fine line, instead can plot a best fit of overall points. When the user zooms, we load the fine points.
  • No point loading all the tweets. First we show friends and influencers tweets and depending on the user zoom level, we load more tweets in place.

This requires heavy understanding of web performance as well as D3 Data selections:

• https://bl.ocks.org/mbostock/3808218
• https://medium.com/@c_behrens/enter-update-exit-6cafc6014c36

[digi0ps]

Yeah, I will look into those.

[harwee]

load data for the last 30 minutes

30 minutes is too High, reduce it to around 5 - 10 minutes

user level at zoom 1 would see the entire line chart.

User level 1 zoom should be preferably, last 5 to 10 mins data