Simplify `params.current_basedir`

[victorlin]

pathogen-repo-guide/ingest/build-configs/nextstrain-automation/upload.smk

Lines 30 to 39 in 4784a83

	current_basedir = str(workflow.current_basedir),
	benchmark:
	"benchmarks/upload_to_s3/{remote_file}.txt"
	log:
	"logs/upload_to_s3/{remote_file}.txt"
	shell:
	r"""
	exec &> >(tee {log:q})
	{params.current_basedir}/../../../shared/vendored/scripts/upload-to-s3 \

Usage of workflow.current_basedir in upload.smk (first and last lines above) could be simplified. Two options:

1. Use {workflow.basedir}

   • This is used elsewhere.

     pathogen-repo-guide/ingest/rules/fetch_from_ncbi.smk

     Line 154 in 4784a83

     {workflow.basedir}/../shared/vendored/scripts/fetch-from-ncbi-entrez \

   • Con: this makes the snakefile it appears in less portable in some ways, as the path is now based on the top-level Snakefile instead of the current file.

2. Use {workflow.current_basedir!s}

   • This is blocked on what seems to be a Snakemake bug: AssertionError when using workflow.current_basedir in shell block snakemake/snakemake#3796

[tsibley]

Using workflow.basedir makes the snakefile it appears in less portable in some ways, as the path is now based on the top-level Snakefile instead of the current file.

I guess you could view this the other way too: it makes the top-level Snakefile more portable.

[joverlee521]

Ah, I propagated this from nextstrain/measles#62 (comment).

I went with workflow.current_basedir because the relative path was easier to understand from the nested build-configs/nextstrain-automation/upload.smk file that is optionally included in the ingest/Snakefile.

[victorlin]

I see, thanks for linking to that thread! I think it's fine to use workflow.current_basedir, I just don't like that it needs two loosely coupled lines to work properly. Hopefully the Snakemake bug can be fixed.