bcftools merge incorrectly drops symbolic ALTs via vertical merge

[ACEnglish]

I have discovered that symbolic variants' END position isn't being considered when running bcftools merge, thus creating a vertical merge.
The documentation states that the command isn't intended for vertical merges, which I believe implies it will not perform a vertical merge, but it is performing a vertical merge, sometimes.

Example

A.vcf

##fileformat=VCFv4.1
##contig=<ID=chr1,length=248956422>
##FILTER=<ID=PASS,Description="All filters passed">
##INFO=<ID=SVLEN,Number=1,Type=Integer,Description="SV length">
##INFO=<ID=END,Number=1,Type=Integer,Description="End position of structural variation">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	Sample
chr1	147022730	SV1	N	<DEL>	.	PASS	SVLEN=-570334;END=147593064	GT	0/1

B.vcf

##fileformat=VCFv4.1
##contig=<ID=chr1,length=248956422>
##FILTER=<ID=PASS,Description="All filters passed">
##INFO=<ID=SVLEN,Number=1,Type=Integer,Description="SV length">
##INFO=<ID=END,Number=1,Type=Integer,Description="End position of structural variation">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	Other
chr1	147022730	SV2	N	<DEL>	.	PASS	SVLEN=-990414;END=148013144	GT	1/1

bcftools merge --no-index -m none A.vcf B.vcf

##fileformat=VCFv4.1
##FILTER=<ID=PASS,Description="All filters passed">
##contig=<ID=chr1,length=248956422>
##INFO=<ID=SVLEN,Number=1,Type=Integer,Description="SV length">
##INFO=<ID=END,Number=1,Type=Integer,Description="End position of structural variation">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##bcftools_mergeVersion=1.21+htslib-1.21
##bcftools_mergeCommand=merge --no-index -m none A.vcf B.vcf; Date=Tue Jan 28 15:57:35 2025
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	Sample	Other
chr1	147022730	SV1;SV2	N	<DEL>	.	PASS	SVLEN=-570334;END=147593064	GT	0/1	1/1

A temporary work around is to use -m id

##fileformat=VCFv4.1
##FILTER=<ID=PASS,Description="All filters passed">
##contig=<ID=chr1,length=248956422>
##INFO=<ID=SVLEN,Number=1,Type=Integer,Description="SV length">
##INFO=<ID=END,Number=1,Type=Integer,Description="End position of structural variation">
##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##bcftools_mergeVersion=1.21+htslib-1.21
##bcftools_mergeCommand=merge --no-index -m id A.vcf B.vcf; Date=Tue Jan 28 15:58:26 2025
#CHROM	POS	ID	REF	ALT	QUAL	FILTER	INFO	FORMAT	Sample	Other
chr1	147022730	SV1	N	<DEL>	.	PASS	SVLEN=-570334;END=147593064	GT	0/1	./.
chr1	147022730	SV2	N	<DEL>	.	PASS	SVLEN=-990414;END=148013144	GT	./.	1/1

However, assigning unique IDs to variants across files/experiments is non-trivial.

Note that vertical merging happens with/without --no-index.

Original reporter: ACEnglish/truvari#256

[pd3]

Hopefully this should be fixed, please try it out!