Inconsistent exit codes

[alephreish]

Hello, I noticed that SNPGenie's successful runs exit with what seems to me to be inconsistent exit codes: 0 or 25. From my experience, runs with warnings exit with code 0 and runs w/o warnings exit with code 25 which is particularly counterintuitive. Any code other than 0 means a failure. In practical terms, this breaks e.g. snakemake rules or stuff like snpgenie.pl && mv results to_some_other_dir. I'm using snpgenie 1.0 via bioconda.

[alephreish]

At first I thought that snpgenie.pl || [ $? -eq 25 ] would be a valid workaround but the exit codes of failed calls are even less consistent and some of them return status 25 or even 0 as well:

• Results directory already exists => 255
• Problem creating results directory => 0
• There must be only one reference genome. => 25
• XXXX.gtf does not contain any sense (+) strand products. => 25
• The SNP Report XXXX.vcf has no header. => 9
• User must specify the specific SNP report format => 255

This means that it is impossible to tell if the run was successful or not w/o checking the output files.

[singing-scientist]

Dear @alephreish, thank you for flagging this problem. Indeed the program was written purely (and poorly) as a standalone tool, with the expectation that the output files would be examined by a human, well before I understood or anticipated usage and conventions in pipelines like this. I really apologize and wish I had time and support for addressing this and myriad other inefficiencies. I do hope to undertake a complete rewrite in the future if I can find support to do so. In the meantime, please do let me know if there are any π or analysis questions I can help with. Thanks for your understanding!

[alephreish]

@singing-scientist As long as the behaviour of a piece of software is predictable, it is in principle possible to integrate it in a pipeline. SNPGenie is not a very hard case in this respect: e.g. snpgenie.pl --outdir={output} &> {log} ||:&& grep 'completed successfully' {log} would ignore the exit status entirely and instead check whether SNPGenie reported to have completed successfully in the log, but a clean solution would be to fix the exit codes.

If I find the time for this, do you accept pull requests?

[singing-scientist]

@alephreish thanks so much for your understanding and your generous proposal! I think if the changes are incredibly simple, say only employing die and exit in straightforward ways, I could find the time to evaluate the changes you might propose and run the necessary tests. (I'm sure you've already noticed the script is incredibly brittle and inelegant — the maiden coding voyage of yours truly.) But I really do apologize for my lack of bandwidth, as I would have no time to do this until at least February. Thanks again and let me know your thoughts.