integer overflow?

[lynxoid]

Hi guys,

Ran mash on a collection of reference 16s rRNA sequences (3.6Gb) and got this warning while sketching:

WARNING: For the k-mer size used (21), the random match probability (0.000726951) is above 
the specified warning threshold (0.01) for the sequence "..rdp/current_Bacteria_unaligned.fa" of 
size 18446744072614073515. Distances to this sequence may be underestimated as a result. 
To meet the threshold of 0.01, a k-mer size of at least 20 is required.

Seems like the message should not be displayed and/or an interger overflow happened when calculating the size of the reference.

My suspicion is that maybe there an assumption that the reference fasta will always contain a single sequence?

[aphillippy]

Hey Darya, yes, that genome size looks completely wonky. We'll take a look. Thanks!

[tseemann]

I just got the same thing on a massive collection of S.aureus genomes:

WARNING: For the k-mer size used (21), the random match probability (0.000641033) is above the specified warning threshold (0.01) for the sequence "saureus" of size 18446744072235684419. Distances to this sequence may be underestimated as a result. To meet the threshold of 0.01, a k-mer size of at least 20 is required.

The sequence size is ~16 Gbp of contigs.

[tseemann]

Ok, I built head with your 64 bit int patch and it now sketches them ok. 6aad387

Can you make a 1.0.2 release?

[ondovb]

Fixed in v1.0.2