Marcel Martin

Turning the hash type into a template parameter would indeed be the way to go. One question: What happens at the moment when there’s a hash collision? At which time...

Thanks, right, there was something about slightly better sensitivity, I remember now. I did mean the randstrobe collisions. The mechanisms for resolving collisions in the hash table are what makes...

Link to mentioned paper: [A tandem simulation framework for predicting mapping quality](https://genomebiology.biomedcentral.com/articles/10.1186/s13059-017-1290-3) The current MAPQ computation function in StrobeAlign uses the formula from minimap2, see [Section 2.3.1 in the paper](https://academic.oup.com/bioinformatics/article/34/18/3094/4994778#121258111).

> I don't see any section numbers in the linked paper Sorry, I shouldn’t comment so late at night – I have fixed the link now.

@ksahlin sent me this comment by e-mail and asked me to post it here: > Good catch with the min_matches bug! Since our “hits” (pairs of linked seeds) are different...

Copying @PaulPyl’s comment from #79: > I didn't have a lot of time this week to check it more clearly but I was wondering about the current implementation of MAPQ...

I temporarily changed `get_MAPQ` to not cap the mapping qualities at 60. With this, we can see what the minimap2 formula outputs. The median score is 2013, so it’s no...

Oh, made a mistake when computing the median, it’s actually ~1780. And with the fix in #84 merged, this went down to 378.

(via https://github.com/python/typeshed/issues/9860) Here’s an example of a seemingly innocuous usage of the multiprocessing module that triggers this error: ```python import multiprocessing context = multiprocessing.get_context() class MyProcess(context.Process): # error: Name "context.Process"...

> I tried to create a similar environment locally for pythonbindings and the build passes. it may be a problem with the SOURCE_DIR variable, but I am not sure what...