no show (Windows 7, Julia 0.5, electron)

[PaulSoderlind]

Hi,
and thanks for the nice package. Seems to work well in Jupyter (Win 7, Julia 0.5), but doing include("test.jl") in the REPL gives a very random experience: most of the time Electron just opens a blank white window. Sometimes the figure shows up. I have so far not observed a pattern in this.

Paul S

[sglyon]

I'm sorry you're seeing inconsistency with Electron. I haven't noticed anything similar, but I don't use Windows. Perhaps it is an issue with Windows + Electron? @MikeInnes does this sound like it might be the problem?

@PaulSoderlind I'm worried there isn't much I can do if there isn't a reproducible way to make the problem happen. If you do figure out a consistent way to trigger the issue please let me know

[PaulSoderlind]

Hi,
and thanks for the reply. The problem basically happened almost all the time, but sometimes/somehow didn't.

Today, Blink was updated to 0.4.3 and I checked out the latest PlotlyJS.jl master (0.5.0+). This changed things. Running a script shows the 2nd plot (and all further plots) in Electron. The first plot is not shown. Still, this is much better than before.

/Paul S

[sglyon]

Thanks for the follow up. That is very strange that the first plot doesn't show up.

@MikeInnes is the blink server lazily started? If so, perhaps that's why only the first plot has issues.

That's the only think I can think of off the top of my head that would cause the first plot to behave differently from any other plot.

[tbreloff]

FWIW I have seen the same behavior... occasionally the first plot doesn't load, likely due to a race condition of some sort.

[malmaud]

I've seen this too

[MikeInnes]

Yeah we changed it to lazy starting recently. Though I think it should be waiting for it to start before opening the first window. Could try removing the lazy load and see if that helps?

[sglyon]

Ok thanks for the heads up is there a way I can have the server start from my __init__ function in plotlyjs.jl, or do I have to change things in blink?

[MikeInnes]

Should be able to call this function. I guess we need to modify that to return once the listening event is fired, or something.

[malmaud]

If there's a race condition though, we should really be trying to fix the underlying issue instead of just making the server start faster to reduce its frequency.

[sglyon]

I think what @MikeInnes suggested (return once listening event is fired) would solve the race condition @malmaud talks about. Does that seem right?

If so, it would be great to see that.

[sglyon]

Closing as this has been inactive for almost a year. Please reopen if this is still an issue.