Bug in locate_errors

[markvanderloo]

The following gives an invalid solution (version 0.5.1)

library(errorlocate)
d <- data.frame(
  x = 10, y = NA, z=5
)
v <- validator(x + y == z, x>=0, y>=0)

out <- locate_errors(d,v)
out$errors

replace_errors(d,v)

[edwindj]

Bug is fixed, actually were multiple bugs rolled into one master bug:

• when a variable had a missing value the mip problem was not well formulated (most of the times it worked in some cases it didn't). Should be fixed now

• as an optimization records without fails (as in validate::confront and aggregate(cf, by="record")) were skipped in errorlocate. This resulted in skipping the record above in the example: because y == NA validate::confront returns NA and not fail.

• This has been changed into that errorlocate now only skips records when there are no nNA or fails on a record.