bash script to run a python command with arguments in batch

Question

I am trying to automate running a python command with arguments using multiple files (one at a time) and writing the output in output directories with the same name of the input file but without the extension.

The files are fasta files containing the proteins from bacterial genomes. The python routine is designed to extract those proteins with certain properties and printing various outputs in the directory /myruns/test. If I run it with one genome, the name of test would be replaced.

Example:

fast file XYZ.faa will be taken as input and the output files will be placed in /myruns/XYZ.

Manually it works very well, but I want to use it in batch with multiple *.faa files and create directories with their corresponding names, otherwise the last file processed will erase the content of the previous one.

So far I have built made the following script:

#!/bin/sh for filename in *.faa ; do python predict_genome.py \ --fasta_path /Users/mvalvano/DeepSecE/myruns/${filename} \ --model_location /Users/mvalvano/DeepSecE/model/checkpoint.pt \ --data_dir data \ --out_dir myruns/test --save_attn --no_cuda done exit 0 

This script works, and the output files are saved in a test directory that is specified in the --out_dir argument. My question is how can I replace test in the --out_dir argument with a function that would name the directory with the same name as the input file. I have tried a few options but they do not seem to work.

Thanks Mike

Answer 1

Replace test with "${filename%.faa}" to get the name of the file with .faa removed. You should also quote "${filename}" to avoid problems in case of filenames with spaces.

#!/bin/sh for filename in *.faa ; do python predict_genome.py \ --fasta_path /Users/mvalvano/DeepSecE/myruns/"${filename}" \ --model_location /Users/mvalvano/DeepSecE/model/checkpoint.pt \ --data_dir data \ --out_dir myruns/"${filename%.faa}" \ --save_attn --no_cuda done exit 0 

With input files

bar.faa foo.faa 

the script will run

python predict_genome.py --fasta_path /Users/mvalvano/DeepSecE/myruns/bar.faa --model_location /Users/mvalvano/DeepSecE/model/checkpoint.pt --data_dir data --out_dir myruns/bar --save_attn --no_cuda python predict_genome.py --fasta_path /Users/mvalvano/DeepSecE/myruns/foo.faa --model_location /Users/mvalvano/DeepSecE/model/checkpoint.pt --data_dir data --out_dir myruns/foo --save_attn --no_cuda 

Possible problems with this script:

Since you specify --fasta_path /Users/mvalvano/DeepSecE/myruns/"${filename}", your script will only work without error if the current directory is /Users/mvalvano/DeepSecE/myruns/ or if this directory contains at least the same set of *.faa files as the current directory. (*.faa will expand to the file names in the current directory.)

When /Users/mvalvano/DeepSecE/myruns/ is the current directory, the argument --out_dir myruns/foo might expect or create a directory /Users/mvalvano/DeepSecE/myruns/myruns/foo with double myruns.

Answer 2

Maybe it would make more sense to write it as:

#! /bin/zsh - topdir=/Users/mvalvano/DeepSecE ret=0 for filename in $topdir/myruns/*.faa(N); do outdir=$filename:r mkdir -p -- $outdir && python -- $topdir/predict_genome.py \ --fasta_path $filename \ --model_location $topdir/model/checkpoint.pt \ --data_dir $topdir/data \ --out_dir $outdir \ --save_attn --no_cuda || ret=$? done exit $ret 

Where we use only absolute paths removing the doubt about what relative paths are relative to.

(here switching to zsh (since that /Users suggests macos) for its :rootname modifier (from csh), its Nullglob qualifier and to remove the need to quote all expansions).