Exception in prediction when hitting horizon #25
Description
Describe the bug
Using prog_server, I got an exception when making a prediction and the prediction was stopped by hitting the horizon. Flask returned the html output of the exception back to the client, rather than an HTTP error code.
Relevant Requirements
Nothing directly relevant, but I found it when I was investigating the propagate to event functionality.
To Reproduce
Steps to reproduce the behavior:
- Download this file and remove the .txt extension to make it a shell script again
prog_server_sws_model.sh.txt - Install the dependencies needed by the bash script, listed at the top
- Start prog_server using the SWS script that adds the SWS models to the server, wait for it to initialize
- In another window run the prog_server_sws_model.sh shell script. Note it calls for a prediction and gives a short horizon=5.0
- prog_server is expected to throw an exception
- If debugging, there is an IndexError that you can press "resume" and it is handled (out of debugger it doesn't stop application)
- The problem one is "ValueError('All samples were none')", which is being raised from uncertain_data_metrics.py line 82.
- This ValueError exception is sent back to the web client, which may or may not be what you want the interface to be. Maybe that is useful, maybe you would want to send an HTTP error of some sort.
Expected behavior
When the horizon is hit, I would expect the return to be the set of events being monitored and their values.
Screenshots
I don't think you'd need screenshots, but let me know if they would help.
Desktop (please complete the following information):
- OS: Ubuntu 20.04
- Python Version 3.8.10
Additional context
none