Migrate print() to the Logger?

[Spinnernicholas]

I’m not sure if this was a deliberate design choice, but I noticed that pixeltable has a configured logger while still relying on many direct print() calls. These print calls primarily confirm database operations and are helpful if you’re getting started or running a few individual commands by hand. However, they can flood the console when you run thousands of operations. In my case, I’ve even had a Jupyter notebook become essentially unresponsive after thousands of print() statements.

Ideally, I’d still like to use print() in my scripts for quick debugging, but have the pixeltable outputs redirected to a log file for later review—without resorting to complicated Python stdout redirection.

Question:
Is there an easy way to achieve this with the current setup? If so, please let me know. If not, I propose the following change:

1. Keep the current behavior as the default (i.e., using print()).
2. Provide a configure_logging() option (or similar) to route all output through the logger.

This way, users can choose between seeing output directly in the console or sending it to a log file. Let me know your thoughts!

[mkornacker]

Hi Nick, we actually also do logging to a log file (under ~/.pixeltable/logs). But you make a good point, we should probably allow the user to configure whether the status messages go to the console and/or the log.

Let me bring this up with the team early next week, and we'll get back to you shortly. Thanks for making the suggestion!

[mkornacker]

Hi Nick, quick update: we're going to add a 'verbosity' config option that you'll be able to set to 0 to disable console output.

[Spinnernicholas]

Awesome!

I have some other little annoying things I have found. Should I just keep posting threads on here? I'm going to end up being like 90% of the threads pretty soon and I don't want to be a nuisance.

---

One of these is related to JSON types in expressions with UDFs. For example:

{"stringdata":"this is a string"}

Does not work:

• mytable.where(mytable.jsoncolumn.stringdata.contains('string')).collect()
• mytable.where(mytable.jsoncolumn['stringdata'].contains('string')).collect()

Works:

• mytable.where(pxt.functions.string.contains(mytable.jsoncolumn.stringdata, 'string')).collect()

My assumption based on the error is that it is trying to use 'contains' as part of the internal json path. I don't even know how you would reasonably fix this but it is annoying.

[aaron-siegel]

Hi Nick,

Thanks for all this feedback! I think it's actually something different going on, which is that mytable.jsoncolumn.stringdata has type Json, but contains expects a string. Try the following and see if it works:

• mytable.where(mytable.jsoncolumn.stringdata.astype(pxt.String).contains('string')).collect()

That might be a bit more natural.

We'd welcome any other feedback you have. We're continually working to improve the user experience, so whatever else you find annoying or high-friction in Pixeltable, we'd love to know about it.

-a

[mkornacker]

Hi Nick, you can now configure the volume of console output.

See 'verbosity' in the config:
https://pixeltable.github.io/pixeltable/config/

Let us know if that takes care of the problem for you.

Marcel