edawson / svaha2 Goto Github PK

Long vcf lines (such as those with the SV sequence in the ref/alt fields) silently fail parsing. This is standard get_line behavior but is a major issue.

Temporarily, I've fixed this by making the maximum line length really large. This is not an efficient solution. I need to think about how to write a better line-by-line or htslib-style parser (or just htslib) for variants.

The GFA output of svaha2 is currently valid GFA 1. However, minigraph requires GFA overlaps in its output. This means the current output of svaha2 is invalid as minigraph input, which is a real shame.

I have repeatedly run into this issue so it's high time I write it down. Variants in a vcf with adjacent positions will cause a segfault, probably because of some sloppy pointer invalidation I coded in mid-thesis.

The right way to fix this is to make a map from genomic position -> a list of alleles/nodes, rather than a single entity. This increases the code complexity but would be a useful improvement.

Because we use a reference genome for input, we can easily generate rGFA. This also gets around the many issues of representing paths in GFA 1 and GFA 2.

We need only three tags to make minigraph-compatible rGFA output:

For reference graphs, this is pretty simple. We'll need to implement pulling some type of VCF name in as well though.

Hello，when I use command：
svaha2 -r tgraph.fa -v tgraph.vcf > test.gfa
to generate GFA file,I failed with the error info：

`FASTA index found for tgraph.fa
Parsed fasta index with 1 entries.
Loading variants from tgraph.vcf
Parsed 6 variants.
Created map of contig to breakpoints from variants.
Contig x has 12 breakpoints.

Adding max-node-length breakpoints.
Contig x has 14 breakpoints.
Sorted and uniquified breakpoints
Building reference backbone for x.
terminate called after throwing an instance of 'std::out_of_range'
what(): at: key not present
Aborted`

I need your help,please!