Warning spam from pybnf dependencies

[emitra17]

Using fresh versions of all dependencies, with normalization enabled, the terminal gets spammed with many copies of the warning "data.py:304: RuntimeWarning: invalid value encountered in true_divide", presumably from a divide by 0 for some parameter sets.

[emitra17]

Since we have a recurring issue of new package updates producing new unwanted output, it would be really nice to permanently solve this by suppressing all output except that generated by our printing module.

[emitra17]

+1 to needing to fix the general case.
Other offenders include YAMLLoadWarning and distributed.client "gather keys" errors. We've gotten reports from confused users worried these are affecting PyBNF performance.

Best case scenario would be to send stderr to the log instead of the terminal.

[SauravMaheshkar]

@emitra17 can I take this up ? I'd love to contribute to the repository and this seems like a good place to start.

[wshlavacek]

Your help with this would be appreciated!

…

Sent from my iPhone

On Dec 17, 2021, at 9:13 AM, Saurav Maheshkar ***@***.***> wrote:  @emitra17 can I take this up ? I'd love to contribute to the repository and this seems like a good place to start. — Reply to this email directly, view it on GitHub, or unsubscribe. Triage notifications on the go with GitHub Mobile for iOS or Android. You are receiving this because you are subscribed to this thread.

[SauravMaheshkar]

I ran pip install -U pybnf on a fresh colab instance and didn't see any of the aforementioned errors. Do these errors come up during editable installs ?

[wshlavacek]

I'm not sure why using the -e option would matter. I wonder if Jacob (cc'ed) fixed the issue but didn't close it.

…

On Mon, May 16, 2022 at 12:35 PM Saurav Maheshkar ***@***.***> wrote: I ran pip install -U pybnf on a fresh colab instance and didn't see any of the aforementioned errors. Do these errors come up during editable installs ? — Reply to this email directly, view it on GitHub <#274 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AETOR53CZ2ZMKJDU7H2CETDVKKIQTANCNFSM4HDMVLYA> . You are receiving this because you commented.Message ID: ***@***.***>

[wshlavacek]

Fixed: Dependency warning spam suppressed

Added logging.captureWarnings(True) in init_logging(). This routes all Python warnings module output (numpy RuntimeWarnings, YAMLLoadWarning, etc.) through the logging system, so they go to the log file instead of the terminal.

Combined with the existing setup that routes distributed, tornado, and asyncio loggers to the log file, this addresses the broader request to keep dependency noise out of the user's terminal while preserving it in the log for debugging.