bcftools +split will fail if sample names contain the "/" character

[freeseek]

If a sample name contains the / character, such as A/B, bcftools split will fail with error message:

[E::hts_open_format] Failed to open file "./A/B.vcf" : No such file or directory
[init_data] Error: cannot write to "./A/B.vcf": No such file or directory

This is mostly the result of the VCF format not restricting sample names from using any weird characters, while UNIX filesystem don't allow / character in file names.

One quick solution would be to change this line of code in split.c:

        for (k=l; k<str.l; k++) if ( isspace(str.s[k]) ) str.s[k] = '_';

to

        for (k=l; k<str.l; k++) if ( isspace(str.s[k]) || str.s[k] == "/" ) str.s[k] = '_';

However, do notice that this fix would work okay for UNIX filesystems, while a similar issue would arise with Windows filesystems with a wider range of characters.

[pd3]

Thank you for raising the issue. This is now addressed by 200bbba.

[freeseek]

There is a minor regression bug in the last fix. The code:

        char *suffix = NULL;
        if ( args->output_type & FT_BCF ) suffix = "bcf";
        else if ( args->output_type & FT_GZ ) suffix = ".vcf.gz";
        else suffix = ".vcf";

is missing a . before bcf, causing the output file names to be incorrect.

[pd3]

This is now fixed, thank you.

(Note that the current develop is waiting for samtools/htslib#1327 to be merged)

[freeseek]

No need for me to push a new release just for this issue. I have it sorted out on my end.