CODEWITHSUNDEEP
pythonrsnakemake
Tadas Bareikis
Tadas Bareikis

Reputation: 25

Executing snakemake workflow without a Snakefile

I am using R and reticulate library to run some python code in a mainly R project. My goal is to somehow wrap a snakemake workflow in an R script so that the Snakefile uses the same environment as the rest of R code. If it was a regular script instead of a Snakefile, it would be possible to do so with reticulate, however I have not yet succeeded in converting a Snakefile to a python script. I am aware of the option to snakemake '--print-compilation' which compiles the Snakefile to a python script, but I have not been able to run it successfully even with a minimal example.

Has anyone had any luck with running snakemake purely from python environment?

Upvotes: 1

Views: 399

Answers (1)

dariober
dariober

Reputation: 9062

In the unlikely case I understood your problem, a solution could be to pass between one snakemake rule and another an R session saved as an image. In this way different rules would share the same environment. For example:

rule all:
    input:
        'stuff.txt',
        'plot.pdf',

rule one:
    input:
        rdata= config['rdata'], # This may come from outside snakemake
    output:
        rdata= 'step2.Rdata',
    run:
        R(r"""
        load('{input.rdata}')
        x <- 1:10
        # More stuff in R...
        save.image('{output.rdata}')
        """)

rule two:
    input:
        rdata= 'step2.Rdata',
    output:
        stuff= 'stuff.txt',
        plot= 'plot.pdf',
    shell:
        R(r"""
        load('{input.rdata}')
        # ...
        write.table('{output.stuff}')
        """)

Execute as snakemake ... --config rdata=my_input.Rdata. For the R() function see scripting-with-r. Note that, as far as I can tell, snakemake is designed to be run as a command line program rather than as a library inside another script.

(If you post a brief example of your problem you may get better answers)

Upvotes: 3

Related Questions