psychopy.org | Reference | Downloads | Github

Experiment vanished after builder crashed

Win10:
PsychoPy version 3.0.2:
Standard Standalone? Yes

I was creating an experiment in builder (and just finishing up), when builder suddenly became very slow and eventually crashed. When i tried to start it back up again the experiment was completely empty and it turns out the file is as well.

I’m guessing there’s no way to get it back, at least not in builder view (still have the code version). I have an older version so will have to remember what changes i’ve made. Mostly wanted to report it.

Also tried googling it a bit and found this:

Apart from this sort of dramatic error i love PsychoPy so keep up the good work!

/Örn

Thanks @Oern, do you have any more info about the crash, for example, what you were doing in PsychoPy when the crash occurred? If someone can recreate this error, it can be raised as an issue on GitHub and fixed.

Hey! I’ve tried to recreate it myself without success. I’ll try to describe what i remember:

Hardware: HP Elitebook 1040G3
intel i5-6200U @2.3GHz, x2
8GB RAM at 2133MHz
Intel HD graphics 520
(not sure if you want to know this, or something else, but just let me know)

Basicly I was building an experiment in builder view. It’s a pretty big experiment, around 8 different tasks and it takes about 40 mins to complete, so plenty of routines and components. I’ve been piloting it for the last week and was making some changes. Had been working on it for about 9 hours, and noticed that the application was becoming slower, to a point where it was hard to work. More specificly i remember i was copying and pasting text components in a routine which already had around 15 text components. I remember noticing that python was eating up 1.5GB of RAM. I began worrying that PsychoPy might crash, and so i remember pressing ctrl+s, and as far i can remember this ironically enough caused it to crash. When i tried to open it back up the flow was empty and when i looked into it further i realised that the .psyexp file was just a completely empty file.

Not sure if any of this is helpful but just keep asking if you are curious! :slight_smile:

Do you mean that you opened it in a text editor and found that it had no content?

I did that, yes. That fact that it is a 0kB file was also a hint.

This is really unusual (and of course a real pain to you). The only thing I can respond with is the same as for any file loss or corruption: do you have a daily/hourly backup system in operation (or even use git)?

Well unfortunately in this case i lost a couple of days of work, but doing things a second time is always quicker so i managed :slight_smile:
Hitting myself over the head for not using git yet though, so am going to be getting in to that now.

Git is probably not as ideal for tracking changes in an XML document like a .psyexp, as opposed to pure Python code, but might still be useful. If being used locally though, it doesn’t protect you from catastrophic data loss affecting the whole folder. i.e. if not pushing changes to GitHub frequently, it would still need to be used in conjunction with a regular local backup system.

Thanks, that’s useful advice. This was a potent lesson that I can’t be so careless with backups.

Having said that, I don’t want to minimise the issue with the file getting corrupted in the first place. Although this is very unusual, we certainly don’t want psychoPy to be doing this. If you ever have any insights on how this situation could be reliably reproduced, we’d really like to know, so this issue could be resolved and prevented from happening again.

I’ll definitely let you know!