feature request: unbounded calculation of coverage histogram & reporting non-empty coverages

[KamilSJaron]

Hello,

Thank you for this wonderful tool. Would it be possible to ask for two small features?

We are using coverage histograms to estimate genomic properties. We have found cases where we had to take into account kmers with extremely high coverage to get meaningful estimates (related to #115; something like -cx50000000). This is impractical because we need to figure out first what is the coverage of the most covered kmer or overshoot the number and generate an enormous histogram.

I would suggest two subtle changes:

• allow something like -cx0 for 'all the coverages'
• output only coverages of the histogram with at least one kmer

When super repetitive kmers are present, there is usually just a handful of them. Therefore the histogram file is full of coverages with 0 kmers. I suggest reporting only coverages that carry at least one kmer. For instance instead of

...
15651666   0
15651667    1
15651668    0
15651669    0
15651670    0
15651671    1
15651672    0
...

we would have

...
15651667    1
15651671    1
...

[hannesbecher]

I would like to second this suggestion! Yes, yes.

[hannesbecher]

As an alternative, it would be useful to have the option to write to stdout, so one could filter the output stream with awk etc.

[marekkokot]

Hello, thanks for using kmc and kmc_tools. I know there is a couple of feature requests, I hope I will find a time to implement at least some of them. For now, it is in fact possible to write to stdout and filter with awk. If the kmc output is o you may use:

bin/kmc_tools -hp transform o histogram /dev/stdout | awk '{if ($2 != 0) {print}}'