Force overwrite of directory?

I’m using QIIME2 with snakemake. One of the things snakemake does is create output directories when specified in the snakefile.

When running things like the Moving Pictures tutorial, this becomes problematic:

qiime diversity core-metrics-phylogenetic \
  --i-phylogeny rooted-tree.qza \
  --i-table table.qza \
  --p-sampling-depth 1109 \
  --m-metadata-file sample-metadata.tsv \
  --output-dir core-metrics-results

Nets an error:

Error: --output-dir directory already exists, won't overwrite.
Error in job coreDiversityMetrics while creating output file core-metric-results/weighted_unifrac_pcoa_results.qza.

Because the output directory is created by snakemake before the directory is created by QIIME2.

Is there a way in QIIME2 to force overwrite if a directory already exists? If not it’s an ugly hack:

Hi @metagenomes! What if instead of using the --output-dir flag, you specified all of the outputs:

qiime diversity core-metrics-phylogenetic \
  --i-phylogeny rooted-tree.qza \
  --i-table table.qza \
  --p-sampling-depth 1109 \
  --m-metadata-file sample-metadata.tsv \
  --o-rarefied-table core-metrics-results/table.qza \
  --o-observed-otus-vector core-metrics-results/observed-otus-vector.qza \ 
  --o-shannon-vector core-metrics-results/shannon-vector.qza \
  --o-evenness-vector core-metrics-results/evenness-vector.qza \
  --o-jaccard-distance-matrix core-metrics-results/jaccard-distance-matrix.qza \
  --o-bray-curtis-distance-matrix core-metrics-results/bray-curtis-distance-matrix.qza \
  --o-jaccard-pcoa-results core-metrics-results/jaccard-pcoa-results.qza \
  --o-bray-curtis-pcoa-results core-metrics-results/bray-curtis-pcoa-results.qza \
  --o-jaccard-emperor core-metrics-results/jaccard-emperor.qzv \
  --o-bray-curtis-emperor core-metrics-results/bray-curtis-emperor.qzv

I know that is a huge pain in the rear, but you can overwrite outputs without an error when you use the --o flags, versus the --output-dir flag. Let me know if that gets you moving - thanks! :t_rex:


Given that this is part of a pipeline, you will only have to list these outputs once and snakemake will validate all of these outputs for you.

In the context of snakemake, I think using explicit outputs is pretty elegant. It’s good that snakemake is controlling this, instead of qiime. :man_shrugging:


Thanks guys, explicit outputs is fine - I wasn’t aware I had such fine grained control at my fingertips there, I don’t have an issue with this - it’s definitely more elegant than the linked solution!


This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.