Better RMarkdown functionality

[ropensci]

Better RMarkdown functionality

[elinw]

I think part 1 of your issue is already known and referenced in the "skim_print() doesn't work for classes other than numeric, factor or character" issues #71. If you test the skim_print() refactor code #101 this is addressed.

Yes the output is a list of tibbles one for each variable type which is the design given that the columns are (potentially) different for each type. If you have a thought about a different way of handling that it would be great to explain, but my suggestion is to use the skim object as the basis for building a custom print method. This aspect of the package makes it really flexible. One could also merge the tibbles and just have NA for the statistics that don't apply for the type.

My issue with the markdown is what I think you are saying in the last paragraph. The way the tables are wrapping in the console is somewhat a matter of preference (you can modify that just like for any tibble by changing the width to Inf), but when I knit to html I really don't think that wrapping is helpful (what's the point of knitting to results you will hide) and I'd rather either have full width with a horizontal scroll or degrade the table gracefully so that the data are shown underneath.

[puterleat]

Another potential issue here is that in RStudio RMarkdown document this prints a sequence of data frames:

skim(mtcars)

The columns of which are var, type, missing, etc. But this fails because the actual return value is the tidy dataframe:

skim(mtcars) %>% select(var, mean, sd)

I had to write names(skim(mtcars)) to work out what was going on. I'm not sure if this is a problem with skimr or Rstudio though.

[elinw]

I think you are mixing together the skim object (which has a long, vertical structure) and the skim print object (which has a wide, horizontal structure separated out by type (in a list)). The type separation in the print object is because each type has a different set of functions (e.g. factors return different things than numeric). If you review the readme file it explains a lot of this.