http://codepen.io/cpsievert/pen/QGGMej
I originally discovered problems re-creating updating this page to use v1.20.0 -- http://cpsievert.github.io/pedestrians/stl-tour/
I get different errors depending on whether viewing in Chrome or Firefox...
I assume you're saying this example falls under this category:
cartesian + gl2d subplot on the same graph are not supported
But this is not a subplot...they are 2 different graphs on the same page
What's the problem? I don't see any console errors?
Do you see errors for the second link? For the codepen, you have to uncomment this line:
//"mode": "markers",
When I do that, it freezes my browser tab in Chrome
@cpsievert I'm able to get it to crash with this simplified case involving just one plot: http://codepen.io/rsreusser/pen/woorOo?editors=0010
Plotly.plot('graph1', [{
"x": [1],
"y": [1],
"mode": "markers",
//"type": "scattergl"
}]);
Perhaps it's a problem with length-one traces?
Interesting. Maybe @dfcreative would be interested in looking into this.
Yea, @rreusser, that is probably the real issue here...and now that second link is now rendering ok for me...I must be seeing things!!
@cpsievert thanks for the report!
Sorry about that hasty close. I must be going blind.
This bug was introduced in 1.19.0.
I noticed the bug a fair few days ago, but I did not have the time to dig into why the scatterplot kept crashing when I started to receive data.
After finally getting free times after finishing my exams for the semester, I finally found out it was due to adding 1 data point. Opened up the issue tab to post a codepen with the issue to find out someone beat me to it.
@monfera looks like one of your patches caused this regression. Would you mind sharing some insights on why this could be happening?
It seems that the simultaneous presence of these conditions cause the browser tab to hang:
markers (problem doesn't come up if it's markers+linesscatterglInitial thoughts: apparently, it requires a switch into the non-fancy scattergl version. Perhaps a domain projection gets into a degenerate case with one data point (domain high and low values are the same) - though this condition wouldn't be new. The most recent change only impacted the return value - pick index offset - of the render function, so maybe it's related to some downstream effect of that.
Update: it seems to be an earlier regression, working on it now.