Regression bug: LASSOing has become significantly slower

[daattali]

This was introduced some time between 2018-01-26 to 2018-03-07

Here's a plot of 10,000 points:

df <- data.frame(x = runif(10000, min = 0, max = 20000), y = runif(10000, min = 0, max = 20000))
plot_ly(df, x = ~x, y= ~y) %>% add_markers()

Running this 2 months ago, I was able to LASSO points with no problem. Now it's essentially unusable.

Converting it to toWebGL() helps, but is still much slower than it was before.

See video below for comparisons

(related to #1211)

[cpsievert]

Strange. Again, this might be Windows-specific weirdness.

To help isolate the issue, it would be great if you could (try to) replicate these issues with pure plotly.js. I have a codepen here where you can switch scatter to scattergl to switch from SVG to WebGL. Also, feel free to play with different versions (version 1.33.0 was when the scattergl rewrite happened) https://codepen.io/cpsievert/pen/WzNvVN

PS. I'm fairly sure lasso selection in scatter has always been sluggish with this many points. Here is a minimal example with plotly.js 1.30.0 (Aug 21, 2017). When I compare that to 1.35.0 (current version), there isn't much of a noticeable difference -- https://codepen.io/cpsievert/pen/wmvaeQ

[daattali]

Thanks for providing the biolerplate js code, even though it's that simple it would have taken me a long time to come up with it as I'm not familiar at all with plotly.js

[cpsievert]

Closing as this seems fixed plotly/plotly.js#2459 (comment)

[daattali]

Yep it's better in a browser! Within RStudio it's agonizingly slow, but in a browser it's all good now