Suggestion: Declare script as input

[mfansler]

Nice example - I came here from the SO answer.

Unlike the other rule that uses a script: key, the use of shell: key shown below means that Snakemake will not track the dependency between this rule and the scr/test-script.R file:

snakemake-with-R/Snakefile

Lines 18 to 22 in a18f782

	shell:
	"Rscript scr/test-script.R \
	--input {input.data} \
	--output {output.out} \
	--myFactor {params.myFactor}"

I would recommend it as better practice to

• declare the script in the input:

  input:
    script='scr/test-script.R',
    data='data/testData.rds'

• use that Jinja variable in the shell: command, e.g., RScript {input.script} ...

This way, if the script has changes made to it, Snakemake can use them to invalidate any downstream objects that may have used an older version.

At least, that's what I picked up from using that style of integration (custom arguments to the scripts) for a couple years. Personally, I moved on to favor the script: style (using the snakemake object), and find it clearer to maintain.