Leandro Lima

Might be worth trying Google's SparseHash (https://github.com/sparsehash/sparsehash). Lower memory usage but slower, could have some use cases (e.g. roundhound)

Andreas has more experience than us on these data structures, and he recommends using this map: https://gitlab.ub.uni-bielefeld.de/gi/sans/-/blob/kc/src/tsl/sparse_map.h (best compromise for both speed and RAM). We might have hash maps that...

The error rate update bug is a convoluted integration bug I think... 1. The first sample is mapped using the default or argument error rate: https://github.com/rmcolq/pandora/blob/955e1c0c7bcd127361d66a74342a8ac19d651a2a/src/compare_main.cpp#L400-L402 2. The error rate...

Note that when using negative binomial model, we never update or re-estimate the error rate, we always use the default or the one given as argument. This impacts the mapping,...

Hello, no, this fix isn't present on `dev`...

This is needed for scalability, added label

Hey @AmayAgrawal , sorry for my delay. `pandora compare` basically runs in two steps, each using multiple threads: 1. The first step is to map reads from the samples to...

Hey @AmayAgrawal , sorry for the huge delay on answering you. > So I was thinking to combine the output from all these individual PRGs vcf files into one final...

Not sure I understand this issue. The issue is that you have 150 entries in the `pandora compare` output directory, each one corresponding to a sample? And you are just...

Am I right to assume this is for paper?