newer nf-schema validation fails with glob when returning multiple files

[maxulysse]

From references

	-> Entry 1: Error for field 'vcf_dbsnp_vcf' ([/home/max/workspace/references](s3://ngi-igenomes/igenomes//Homo_sapiens/GATK/GRCh37/Annotation/GATKBundle)/{1000G_phase1.indels.b37,Mills_and_1000G_gold_standard.indels.b37}.vcf.gz): could not check existence of '[/home/max/workspace/references](s3://ngi-igenomes/igenomes//Homo_sapiens/GATK/GRCh37/Annotation/GATKBundle)/{1000G_phase1.indels.b37,Mills_and_1000G_gold_standard.indels.b37}.vcf.gz': No signature of method: java.util.LinkedList.getFileSystem() is applicable for argument types: () values: [] (DBSNP VCF must be an existing file ending in '.vcf' that does not contain spaces. This file can be bgzipped)

from rnavar/sarek

* --known_indels (s3://ngi-igenomes/igenomes//Homo_sapiens/GATK/GRCh38/Annotation/GATKBundle/...gold_standard.indels.hg38,beta/Homo_sapiens_assembly38.known_indels}.vcf.gz): could not check existence of 's3://ngi-igenomes/igenomes//Homo_sapiens/GATK/GRCh38/Annotation/GATKBundle/{Mills_and_1000G_gold_standard.indels.hg38,beta/Homo_sapiens_assembly38.known_indels}.vcf.gz': No signature of method: java.util.LinkedList.getFileSystem() is applicable for argument types: () values: []

Most likely due to nextflow-io/nextflow#5507

[bentsherman]

I don't think that PR is related, rather the problem is the file() function itself. It can return a list or a single file depending on whether the glob pattern matches exactly one file. So if you use that function with a glob pattern you need to be careful with the return value

Whereas the files() function always returns a list of files

That Nextflow PR is trying to take steps to solve this problem by warning you if the file() function returns a list. But right now it's only a warning and it only happens with the strict syntax

[bentsherman]

Based on the error message it looks like nf-schema used the file() function expecting a single file (because it's trying to call getFileSystem()) but the glob pattern returned a list.

[nvnieuwk]

Thanks for your two cents Ben! I'll investigate this soon

[maxulysse]

Thanks for the explanation, that makes more sense to be honest

[bentsherman]

Although, my explanation doesn't explain why it stopped working in 25.10, since the file() function has always worked that way...

Happy to help triage if you need it

[nvnieuwk]

I'll investigate now, I'll come back to you if I encounter any issues

[nvnieuwk]

@bentsherman the problem seems to be in the .exists() method when using the file() function on a glob pattern. I might just skip the exists check for now on file path patterns as the format checker checks the amount of files found and thus also does an exists check.

[nvnieuwk]

#182 fixes this issue

[bentsherman]

Okay I see the issue now. Your PR makes sense

[nvnieuwk]

Strange that this worked earlier, should have been an obvious bug 🤔

[maxulysse]

Especially since it's quite central to sarek