This vignette is mostly aimed at package developers, though will be
useful for anyone who wants to use any of the individual
precheck_*() or check_*() functions in
isolation.
If you are the latter, using any checks individually outside of a
screen_*() wrapper function, proceed with caution and read
this article first.
What assumptions?
- Only the
screen_*()functions have argument validation
At the time the package was first developed there were already 90+ checks in the screener. It would be inefficient to revalidate the arguments in every check, for example checking that the datafile is a data.frame 90+ times.
Therefore we only provide argument validation for the top level
screen_*() functions.
- Within all of the the
screen_*()functions, individual checks are placed in a specific order, and broken up by stages.
This means the tests can fail early, giving quicker feedback if there’s fundamental issues, and also means the later checks can be more lightweight and ‘assume’ certain things about the data.frame that is being tested.
For example, precheck_meta_col_type() requires the
col_type column to be present, as this is already checked
for in precheck_col_req_meta() we minimise duplication that
check by not rechecking that column is present.
Starting assumptions
Filename checks can be carried out in isolation from other checks. The
screen_filenames()function can be ran in a single line on two simple strings where needed.Mandatory columns are checked before anything else about the files. Many of the other checks rely on assumptions that key columns exist.
Metadata files are checked before data files, they are faster to test and can give immediate feedback, plus they are crucial for understanding and checking the data file, so we need to have a passing metadata file before we can test the data file.