Informative, user-friendly and unified error messages about easystats HOT 12 CLOSED

True true, but I liked the idea of providing directly the install code to run 😬

maybe:

stop("Package \"X\" needed for this function to work. Please run 'install.packages(X)'.")

so the user can simply copy paste the thing?

That's a super minor thing though, I am also okay with Hadley's holy Word :)

from easystats.

For readability (and usability) I suggest:

stop("This function requires the package 'X' to be installed.\nPlease install by running 'install.packages(\"X\")' and run the function again ;-)")
#> Error: This function requires the package 'X' to be installed.
#> Please install by running 'install.packages("X")' and run the function again ;-)

from easystats.

For my packages, I just copied Hadley Wickham (usually a safe strategy):

stop("Package \"X\" needed for this function to work. Please install it.")

from easystats.

Ha! Hadley's holy word aside, while its a useful place to start, I like the idea of making this our own as well in line with @DominiqueMakowski suggestion. While the assumption could be made that if they're using our packages then they should already know how to install a package, I like the idea of customizing the warning message for our purposes and goals (i.e., making stats in R easy; this makes good sense to me)

from easystats.

Now that I think of it, the easiest way would be to offer the possibility to automatically install the thing, similarly to what caret or other packages do.

"Package \"X\" needed for this function. Press ENTER to install or ESCAPE to abort."

from easystats.

I vote for @pdwaggoner suggestion, as the word "install" only appears once (and not three times in two sentences).

from easystats.

I second that!

from easystats.

I think such user interaction is more dangerous than useful. Imagine this happens when running scripts, tests, vignettes, markdowns, ... Scripts are frozen until user does something and no error message or whatever is shown.

I do not recommend such actions. And if users don't know how to install a package, they can ask for help somewhere or maybe they're not our (or R's) target group anyway ;-)

from easystats.

Mmmh that's true, although that's why there is the interactive state variable. Usually, the user is prompted and asked to respond only if interactive() == TRUE. Else, the traditional stop would occur. Do you think it's not a sufficient safeguard against potential issues related to user input?

from easystats.

We could indeed offer this, however, in my experience, in most cases you need to restart the R session because some namespaces are loaded, or installation fails etc. That's why I personally would say a message is enough.

from easystats.

in most cases you need to restart the R session

so true 😩 So yeah now I agree. Still, returning the right install code so that the user can just copy-paste seems useful (I've used that myself quite a lot)

from easystats.

I think this can be closed?

from easystats.