I have some code that I am incorporating into a Snakemake pipeline. I also want to be able to run the code independently of Snakemake, so I want to write flexible code. I have a config file that can be read by Snakemake and an alternative config file that will be parsed if the code is run outside Snakemake.
I had the following structure in a rule in my Snakemake file:
rule some_rule:
input:
input_file
output:
output_file
script:
runfile.py
and then in runfile.py I tested to see if the script was being run within Snakemake as follows:
if 'snakemake' in globals():
# Get config params from Snakemake
else:
# Load config params from alternative config file
This was fine, except that I am working within a virtual environment and the Python version I want to use is different from the one Snakemake uses by default. So I had to restructure the rule as follows:
rule some_rule:
input:
input_file
output:
output_file
shell:
/path/to/python runfile.py
but now because I am using shell instead of script I don't have access to the global parameter 'snakemake' in my python file.
So my question is either
(i) Can I specify a Python version and still use script?
or
(ii) Can I use shell but still access the 'snakemake' object within the script that will allow me to test whether or not the script is being run in a Snakemake pipeline?
You could use Conda - it would both provide fine-grained control over execution contexts and improve the pipeline reproducibility. See Documentation.
First, you'll need an environment definition YAML file. For example,
envs/py37.yml
channels:
- defaults
dependencies:
- python=3.7
Add in any other requirements you need in that file. Then your rule would be
Snakefile
rule some_rule:
input:
input_file
output:
output_file
conda:
envs/py37.yml
script:
runfile.py
Lastly, you now need to use the additional flag --use-conda when launching this job, e.g.,
shell
snakemake --use-conda output_file