Hi Matt,
Here's a pull request to update the code to address the issue #249.
I found that the source of the coercion was actually upstream in the initial read of the CSV in which the R function read.csv when column classes are not specific defaults to guessing.
Having adjusted the import of the CSV as character, the labelling function now attempts to coerce only the non-code columns as discussed in the issue.
Christian
Hi Matt, Here's a pull request to update the code to address the issue #249. I found that the source of the coercion was actually upstream in the initial read of the CSV in which the R function read.csv when column classes are not specific defaults to guessing. Having adjusted the import of the CSV as character, the labelling function now attempts to coerce only the non-code columns as discussed in the issue. Christian