Redundancy wrt applyFilters() & applySamp()

[drnickisaac]

applyFilters() does two jobs: sample the posterior and apply the filters, including clipping.
applySamp() does just the first.
Separating them is useful if you don't want to do any filters. It also makes the code much easier to debug. As a result there is some redundancy. This is not ideal because over time the two bits code might diverge, with unexpected consequences. A logical change would be to take the sampling part out of applyFilters(), so it reads in the output from applySamp().

[drnickisaac]

In fact this is already starting to happen. In @EllieDyer 's fork, applyFilters() has been modified to take RDS files rather than rdata (in fact the code needs to be flexible enough to read both). This change has not been extended to applySamp()

[drnickisaac]

I think that applyFilters() can be removed. You can get the same result by using applySamp() then applyClipping(). The advantage of this is that you can save time if you want multiple clipping options from the same dataset (or even no clipping).

[drnickisaac]

This code:

sampDat <-  lapply(roster, 
                     wrappeR::applySamp,
                     parallel=TRUE, sample=TRUE)
filterDat_Gr <- rbind.fill(lapply(sampDat,
                                                 wrappeR::applyClipping,
                                                clipBy = "group")) 

Should give the same answer as:

filterDat_Gr <-  lapply(roster, 
                     wrappeR::applyFilters,
                     clipBy = "group",
                     parallel=FALSE, sample=TRUE)

Test this with a simple example. Then either remove applyFilters() or reduce to a shell that calls the other two steps.

[03rcooke]

I have updated applySamp with updated applyFilters code but will test applySamp and applyClipping together before turning applyFilters into a shell

[03rcooke]

I think applyFilters can just be turned into a wrapper for applySamp now