Potential input checking and error handling improvements

Work-in-progress

Purpose

The purpose of the following potential improvements is to make it easier to stand up and use an ActivitySim model. This list of improvements is in response to phase 6a task 7.

Potential Improvements

Idea	Level-of-effort	Notes	Priority
Check input data consistency	Days	Plan to check all the key relationships. Check primary key table joins across input tables - HH home zone vs. TAZ/MAZ land use file, MAZtoTAP file TAPs vs. TAP skims, TAZs in the land use file vs. TAZ skims, etc.	High
Check for well formed config files	Days	Check settings files, expression files, and config files at the start of a run to avoid parsing issues later on.	Low
Better describe silly errors	Days depending on LOE	For example, when a submodel yaml file cannot be found, print more useful info through more exception handling	Low
Print info about no alternative found	Hours	When no alternative is found in a choice model, log which terms in the utility turn off all alternatives since this is really helpful info in debugging.	High
Turn of restarting if desired	Hours	Add a global setting to optionally turn off restarting in order to save disk space (and some runtime) by not saving all the states of the tables in the pipeline and instead just save one version of each table This is especially useful when running lots of model scenarios in application mode.	Medium
Improve RAM/chunking settings	Days depending on LOE	Make it easier for new users to set a reasonable chunking/processors settings since this is tricky. Can we create better defaults or just improve the documentation? Maybe specify an amount of overall machine RAM to use instead of a chunksize?	High
Eliminate distance==0 problems	Days depending on LOE	Would need users to identify distance skims in the settings files since ActivitySim doesn't know what "distance" is and then ensure it is > 0 where "appropriate" to avoid funny simulation situations. Alternatively the skim reader could check for the diagonal sum of skims as being 0, or for rows or columns summing to 0, or for the MAZtoMAZ or MAZtoTAP tables summing to 0 is certain ways.	Medium
Clean up command prompt logging	Hours depending on LOE	Be more intentional about what is logged to the console vs. log file since there's a lot of info currently.	Medium
Avoid long trace file names	Hours	Reduce filenames (especially in estimation mode) because of Windows filename limits.	Low
Tidy up documentation related to settings file inheritance	Hours	Settings file inheritance is used by several examples, but it can be confusing to new users.	Low
Log dependent package versions	Hours	This is useful for debugging, especially since different versions of pytables (for hdf5) behave a little bit differently between versions and OSes	Low
Log time to solve each expression	Hours	To identify which expressions take a long time to run in order to hopefully rewrite them in faster form	Low