Type a call to a function and have a definition for that function appear with a keystroke. Lay the planks with your fnmate but leave the hammering til after smoko.

remotes::install_github("milesmcbain/fnmate")

There are two addins that do pretty much what their names suggest:

• Create function definition file
• Create function definition below
  • as in appended to the end of the current file. Roxygen is not added in this case.

You can bind these to key combinations of your choosing, see here.

The function to be generated is determined by the cursor position. The function call at the most nested scope that encloses the cursor is the one that fnmate will generate a definition for.

There is some Emacs lisp you can use to create bindings to fnmate contained in the vignette Using fnmate with ESS. If you have better ideas about how to distribute ESS 'addins' please let me know!

There are some options that affect how fnmate works:

• fnmate_window determines how many lines above and below the current cursor position fnmate will look for a function call that encloses the cursor. Defaults to 20 which probably covers three standard deviations of coding styles.
• fnmate_folder is the name of the folder in the current working directory to place created definition files. Defaults to "R".
• fnmate_placeholder is the placeholder value that gets put in the function body. Defaults to NULL, can be set to any text.

A lot of the time when I attack a problem I find it helps to cruise over the gnarly bits requiring fiddly code by just claiming a function exits that will magically resolve that fiddly bit for me. After I have a high level solution described, I go back and fill in the blanks. This tool helps me clearly mark out the blanks without breaking my flow on the higher level algorithm.

Recently when developing R workflow plans with drake I've found myself wanting a tool like this so that my sketch of the workflow plan can be built, even though some of the targets are just placeholders.

Also realising this idea in a robust way turned out to be way more challenging than I anticipated and necessitated coopting the R parser and its parseData output. So it became a learning exercise.