we would like a function with the following features:

input - takes a csv file (or data.frame)
parameters - names for ID field, lat, long fields (default to "lat" and "long"), fields for datestart and dateend, optional parameter for furthest distance (to constrain predictions to points that fall within the model area) - default to 0.05 decimal degrees)
checks - check that there is an ID field (otherwise how can they use this)
check that there are valid lat and long
run geomerge
return a long data.frame with one row per day-location making sure to preserve the ID field
optionally, take a filepath-filename to write.csv the results
report back a warning if everyone is outside of the range of the model (a hint that lat and long are messed up)

if the variable that is the key in joinfrom is not a character we try to make it one and that doesn't work right now.

Because the key values are being stored in rownames (in makepointmatrix, for use in the knn), we are trying to coerce (unsuccessfully) to character
https://github.com/allanjust/aodlur/blob/master/R/nearestbyday.R#L42

Variable should be able to be named anything (currently unstated that it must be "day") - it is just another dimension on which we are requiring a join. Add flexibility:

1. make the variable name a parameter (default to day)
2. check that it has the same type between jointo and joinfrom

simplify and then perhaps you won't need to export this as a second function

runs quickly and so it will simplify everything, even if the jointo.pts is sometimes reused

for a neighborhood variable that doesn't use the value closest to the site

1. class/consistency of inputs
2. day is a date (or at least consistent in jointo and joinfrom)
3. jointovarname is in jointo, joinfromvarname is in joinfrom, valuefield is in joinfrom
4. check that joinfrom covers all of the needed days (those in jointo)

keep the same parameter conventions if possible