Positron console stops responding

[NirOfir]

I have been using Positron daily for Python for the last couple of weeks and it's mostly been a good experience. The data explorer is great, especially the ability to sort by more than one column. One thing that does disrupt my work is that the console becomes hung after a couple of hours of work. Suddenly ctrl+enter or running code cells stops working and writing a line of code directly in the console and hitting enter doesn't work as well. The text editor is responsive, I can edit the code and save without issues, but I can't run anything without restarting Positron. I'm looking at somewhat large data (~35 million rows, 8 columns), but I don't think it's a memory issue, because task manager shows I still have plenty of memory and CPU free.
Any ideas?
Thanks!

[juliasilge]

We are tracking this problem in #9852 and are currently making plans for how to address this. As a temporary workaround, you should be able to get out of the laggy, hung state by running Console: Clear Console from the command palette; this will clear the history but not restart Python.

[NirOfir]

Thanks! I'll try that. It's good to know you're working on this issue.

[NirOfir]

Reopening this discussion as I see "Console: Clear Console" does clear the console, but it remains unresponsive.

[juliasilge]

Ah OK, so perhaps this is a separate, different problem, if you don't observe Clear Console clearing the laggy state. It sounds like you are working in a Quarto document?

• If you are in the laggy state, run Developer: Restart Extension Host from the command palette. Does this immediately clear the lag?
• If you are in the laggy state, go to your Extensions view and disable the Quarto extension. This will make a lot of features with Quarto files not work, but does it immediately clear the lag in the UI? Can you use Cmd/Ctrl+Enter in a regular .py again?
• When you get into the laggy state, how many plots are in the gallery/thumbnail area? Does Plots: Clear Plots immediately clear the lag and make the UI snappy again?
• Can you try working with the Variables pane closed (you can right click and choose "Hide View") to see if that is the source of the growing lag? When the Variables pane is not visible, it does not refresh after each code execution.
• As you start working run Developer: Open Process Explorer and keep that window open over time as you work to observe. Does one of the processes start to take a lot of memory? Maybe the processs for the extension host?

[NirOfir]

OK, I think I found a lead. It looks like using qt as the backend for matplotlib might be causing the problem. I tried switching to tkinter and everything worked fine for about a day. When I went back to qt, I got to a laggy state in less than an hour of work. Does that make sense?
I work with floating interactive plots quite a bit, because the scrolling/zooming really helps with timeseries data which is most of the data I look at.

[juliasilge]

Thank you so much! That is very clear and I have opened #12704 for us to investigate what might going on and find a solution.

[LR1933]

I'm not sure if it's the same issue, but this month I've been experiencing frequent lag with the Postron R language editor. Even when handling small samples and simple analyses, it often becomes unresponsive, requiring several seconds to register a change. Sometimes the program even stops responding altogether.
I'm using a Windows computer, so I don't know if it's a problem with my computer or with Postron. Will this be fixed in next month's update?

[juliasilge]

Can you say more about when you are observing this problem? Is it occurring after you work in Positron for a longer session, like an hour or more? If so, can you try running Console: Clear Console when you get into the laggy state? If so, we will be addressing this in #9852.

If not, it would be great to hear more about when you are observing this.

[LR1933]

I didn't pay particular attention to the time; maybe an hour? Or longer. I'm not sure. However, Console: Clear Console worked for me, but only temporarily; the lag returned after a few hours.

By the way, I hadn't encountered this situation before. I didn't analyze data much last month, so I might not have noticed; it ran very well before (the first time the interpreter started was a bit slow, but acceptable).

[juliasilge]

Thank you for sharing that! That sounds like the same problem we are tracking in #9852, which we have slotted for work quite soon.