User-Reported Defect: InitializePopulation fails with unclear message

[MeWu-IDM]

Describe the defect
This error is reported by external user @Nyerere who tried to use spreadsheet, in their "lookup" tab there is empty values for Female / Male column, I tried to help debugging

To reproduce
Steps to reproduce the behavior:

1. Go to "Lookup Sheet", create a row with label but empty values for Male/Female
2. Run pacehrh::InitializePopulation()

Expected behavior

• If these rows are not able to be used, they should be removed with a warning to the user

Actual Behavior
User is getting an error that they do not understand how to proceed

Error in if (sex == "f" && (lt$Female[i] == FALSE)) { : 
  missing value where TRUE/FALSE needed
Called from: FUN(X[[i]], ...)

[MeWu-IDM]

Related: @BHagedorn-IDM spotted if column I ad J are empty, https://github.com/InstituteforDiseaseModeling/PACE-HRH/blob/main/pacehrh/R/pace_population_config.R#L177-L184 use the default min and max which may not be what user intended to do, we should at least give a warning or make these 2 values required.

[celiot-IDM]

to do, we should at least give a warning or make these 2 values required.

I'm not sure this is the right answer either.

Sensible defaulting is a powerful tool for improving usability. In this case I would argue a few things:

• In the absence of any other information, the most sensible default for the start of a range is the lowest possible value, and the most sensible default for the end of a range is the highest value.
• Another basic principle is reducing the dependence on fixed hard-coded configuration values. The built-in age range is 0-100. Say the user has to explicitly enter 100 as the end point for any interval intended to represent "age X and above". Then say somebody decides to extend the built-in range to 110. Every hard-coded reference in configures ranges also has to change.
• The argument that the user might not have intended this is countered by the argument that this might be exactly what they intended. "Age X and above" is a perfectly legitimate construct, so having a compact representation for it will (I expect) be seen by some users as a nice touch.
• The idea that the system should generate a warning message any time it makes an implicit decision for the user is attractive, but the sheer volume of these informative messages could become a different problem. (Our standard model_inputs config file would generate 10 such warnings. Would we need a way to turn these warnings off if the system is system is in fact doing exactly the right thing for the user?)
• Finally, we can discuss whether there's a better solution than a blank cell, perhaps a sentinel value like -1?