Codelists provides functionality to generate a list of codes from different
specifications.

There are three broad approaches:

1. Keep a manual list of codes.

2. One can generate a canonical set of codes given an input specification, and
use those as required to test data. This approach is good when you have
millions of rows of data and lots of checks to perform. Set-up time is longer
but checks should be quicker. It is possible to generate a crossmap table
to demonstrate how selection has occurred, and helpful for reproducibility at
a time point.

3. One can test a set of identifiers against a specification. This approach
is good when a codelist is very large, and fewer checks are needed. Set-up
time is small.

In many situations all approaches might be necessary, depending on trade-offs.
You might generate a codelist for documentation purposes but use a different
approach to check each row of source data. All approaches should give the
same answers.