VcfWriter does not write VCF index files anymore

[mjhipp]

...unless overwriting an existing path (after #753)

When deciding whether to write an index, VcfWriter uses isRegularFile() on the output path. This will almost always return false, because the output path is usually a new, non-existing path (there is not already a regular file there).

Example (using FilterSomaticVcf), if I have this dir:

.
├── existing-file.vcf.gz
├── input.bai
├── input.bam
├── input.vcf.gz
└── input.vcf.gz.tbi

And run the tool using a new output path:

❯ java -jar fgbio-1.5.1.jar FilterSomaticVcf -i input.vcf.gz -o new-output.vcf.gz -b input.bam
...
[2022/03/10 14:39:31 | FgBioMain | Info] FilterSomaticVcf completed. Elapsed time: 0.04 minutes.

I get the following output:

.
├── existing-file.vcf.gz
├── input.bai
├── input.bam
├── input.vcf.gz
├── input.vcf.gz.tbi
└── new-output.vcf.gz

If I instead write to the existing-file.vcf.gz, I get an index:

❯ java -jar fgbio-1.5.1.jar FilterSomaticVcf -i input.vcf.gz -o existing-file.vcf.gz -b input.bam
...
[2022/03/10 14:40:19 | FgBioMain | Info] FilterSomaticVcf completed. Elapsed time: 0.04 minutes.

I now get an index written:

.
├── existing-file.vcf.gz
├── existing-file.vcf.gz.tbi
├── input.bai
├── input.bam
├── input.vcf.gz
└── input.vcf.gz.tbi

Indexes are written using 1.4.0:

❯ java -jar fgbio-1.4.0.jar FilterSomaticVcf -i input.vcf.gz -o new-output.vcf.gz -b input.bam
...
[2022/03/10 14:48:15 | FgBioMain | Info] FilterSomaticVcf completed. Elapsed time: 0.04 minutes.

❯ tree
.
├── input.bai
├── input.bam
├── input.vcf.gz
├── input.vcf.gz.tbi
├── new-output.vcf.gz
└── new-output.vcf.gz.tbi

[clintval]

That's my fault :/ I can take a stab at a fix soon. Would be nice to have this patched before 2.0

[tfenne]

I think we're still a week or more away from a 2.0 release, but I do want to cut a beta/milestone either tomorrow or Monday for folks to start testing