I'm creating a simple snakemake pipeline that contains global variables in the Snakefile. What's the recommended way to use these global variables in the Python scripts called in my rules?
I'm currently using argparse command line arguments as described here (Snakemake: pass command line arguments to scripts) but am wondering if there's a better way.
If the variable is specified in the Snakefile, then it could be passed via params. For example,
Snakefile
# global variable to use
FOO = 100
rule test:
input: "a.in"
output: "a.out"
params:
foo=FOO # pass the variable value as 'foo'
script: "scripts/test.py"
scripts/test.py
#!/usr/bin/env python
# access the variable through the `snakemake` object
print(snakemake.params.foo)
See Snakemake documentation on external Python scripts.
Note, that generally I find it better practice to place a variable like the above example in a config.yaml instead. That helps centralize adjustable parameters, providing a single point of configuration for reuse. Despite the availability of snakemake.config in external scripts, I still prefer to explicitly pass configuration values as params, so as to make it clear which rules depend on what configuration values.
config.yaml
foo: 100
Snakefile
configfile: "config.yaml"
rule test:
input: "a.in"
output: "a.out"
params:
foo=config["foo"]
script: "scripts/test.py"
scripts/test.py
#!/usr/bin/env python
# access the variable through the `snakemake` object
print(snakemake.params.foo)
If the value is provided in the config.yaml, one can also then (optionally) override it at the CLI:
snakemake --config foo=150