samtools sort stdout

[pederworning]

Hello

I am mapping reads from bacterial sequences to reference genomes and I am using bwa and samtools. I am trying to set up a pipeline and want to use stdin and stdout and pipe the data to avoid temporary files. I am using samtools-0.1.13.

Do you know if you can get the output of samtools sort to stdout?

The manual pages says that the -o option means stdout for samtools sort , but it doesn't seem to work on my system.

Here is the command line I have used to make quality filtering and sorting in one line and the error message I get:
>samtools view -b -q 30 5139_6_7.paired_ends.bam | samtools sort -o - > temp
Usage: samtools sort [-on] [-m <maxMem>] <in.bam> <out.prefix>

>samtools view -b -q 30 5139_6_7.paired_ends.bam | samtools sort - - > temp
And this makes as expected an empty temp file and the sorted alignment in "-.bam"

Do you know how I can avoid the *.bam output file?

Thanks
Peder Worning

[daler]

Try providing an output prefix anyway. Even if you specify -o, I think samtools will use that prefix to create temp files if the BAM to be sorted won't fit in memory.

So to modify your example,

samtools view -b -q 30 in.bam | samtools sort -o - deleteme > out.bam

should result in a new out.bam file; deleteme.* may be created as intermediate files but will be cleaned up after the sort

[dawe]

Do you know how I can avoid the *.bam output file?

Thanks
Peder Worning

AFAIK there's no way to do that. samtools sort doesn't handle stdout.

d

[dawe]

Do you know how I can avoid the *.bam output file?

Thanks
Peder Worning

AFAIK there's no way to do that. samtools sort doesn't handle stdout.

d

[gaffa]

The samtools "Working on a stream" page (http://samtools.sourceforge.net/pipe.shtml) says:

"In fact, most of samtools commands, except indexing and external sorting, recognize an input file `-' as stdin and an output file `-' as stdout."

I suppose the rationale is if you're sorting an input list, you can't send items away before you know which position in the output list they will have, making working on the stream trickier.