Functions failing on tibbles

[R-Walmsley]

Some functions fail on tibbles. For example, the comp_mean function.

In the case of comp_mean, this is because tibbles and data frames treat dat[, "colname"] differently, with a data frame returning a vector and a tibble returning a data frame. This leads later functions (in the case of comp_mean, the gm function, because of trying to run mean on a tibble) to return NA values, screwing up computations and leading to a cryptic error message that gives little hint as to the actual problem.

A (much needed!) major rewrite of the package might improve lots of this code to be smart about what it is and is not receiving and returning (e.g. lots more checking of class of objects).

[Feakster]

In the short term, you could use the drop argument:

dat[, 'colname', drop = FALSE]

Or make use of R's list-indexing syntax:

dat['colname']

Would probably be best to write tibble, data.frame & data.table methods though.

[R-Walmsley]

The current functions actually rely on the data.frame behaviour.

But really I agree with:

Would probably be best to write tibble, data.frame & data.table methods though.

I'm currently not sure I have the capacity to do much maintenance and refactoring, so mostly trying to leave a paper trail of what could be improved.