Return plots as ggplot objects

[AndreaPi]

Are the DataExplorer plots ggplot objects? I would like to use patchwork to patch together two plot_missing plots for two different datasets (actually, the same data set, but split by a binary variable). If the plot generated by plot_missing is a ggplot object, then returning it would allow me to patch two (or more) of them together using patchwork.

PS excellent package, thanks for putting it on CRAN!

[boxuancui]

Yes it is a ggplot object. However, the function is already returning something.

Two options IMO:

1. Wrap the ggplot object in a list and return along with other things.
2. The missing value profile is already returned and it goes directly into the plot function from line 57 - 66. You can always create your own ggplot object by reusing/modifying the code.

I am more inclined to option 2, since you already get what you need from the returned object, and it is cleaner. Option 1 seems a little redundant, and might conflict with future development of this package. Again, I do not have clear vision about what that is, but would like to keep it clean and simple for now.

Open for suggestions.

And, thanks for using DataExplorer!

[AndreaPi]

Thanks for the explanation! So, if I understand correctly, plot_missing already creates a ggplot object (called output in your code). However, it doesn't return it, because it already returns (invisibly) missing_value, the dataframe containing the missing value profile.
Option 2 would work, but I don't like it because it's basically duplicating work: you suggest that I assign missing_value to a variable, e.g.

missing_value_profile <- plot_missing(my_dataframe)

and then create a plot copying the ggplot code in plot_missing. But you already create and plot output in your code, so I'd generate the plot twice.

Instead, I suggest adding an optional parameter defaulting to FALSE:

plot_missing <- function(data, title = NULL, ggtheme = theme_gray(), theme_config = list("legend.position" = c("bottom")), return_plot_object = FALSE)

and a simple if at the end of plot_missing:

  ## Set return object
 if (return_plot_object)
  return(invisible(list(output, missing_value)))
 else 
  return(invisible(missing_value))
}

[boxuancui]

It will be released in v0.7. Please use either workarounds in the mean time.

[AndreaPi]

Absolutely! Thanks a lot :-D BTW, another possible solution:

• expose a function which creates the missing value profile, e.g., missing_values_profile()
• have plot_missing() call it
• at this point, plot_missing() can simply always return the ggplot object (output). If the user wants to get the dataframe containing the missing value profile, instead, s/he can directly call missing_values_profile().

This way, plot_missing() would always return the same object, no matter what its inputs are: I don't think this is something indispensable, but maybe it could simplify your life as a mantainer. At the same time, you would still maintain only one function to generate the missing value profile (missing_values_profile()), because plot_missing() would internally call missing_values_profile().

Personally I think that the first solution (adding the optional argument return_plot_object to plot_missing()) should be fine, and it wouldn't break old code, but of course the decision is yours. Thanks again!!

[boxuancui]

Added profile_missing(), and plot_missing() calls that directly.
All plot_* functions return the ggplot objects now.