FAILURE: extra segments in file about segalign

[rsharris]

In lastz, if the --segments option is used, and the input sequences are NOT accessible in random order, the items in the segments file must be consistent with the order in which they 'will be needed'. For most users, "accessible in random order" means YES for twoBit, NO for fasta.

For fasta this should just mean segment names are in the same order as in the fasta file, and all + segments for a given sequence appear before all - segments.

There's some detail here (starting with the paragraph that begins with "Query sequence names must appear in the same order as ..."):
http://www.bx.psu.edu/~rsharris/lastz/README.lastz-1.04.03.html#fmt_segments

That implementation choice seemed more important 15 yrs ago that it does now. Now it would make more sense to read all the segments into memory, sort them by sequence name (to facilitate binary search), and use the appropriate segments as each new fasta sequence arrives.

[gsneha26]

@glennhickey the error is definitely in one of the lastz jobs that Bob has pointed out. Ideally, SegAlign divides and sorts the segments in a way to take care of this requirement. I can look into it. Could you share the input files?

[glennhickey]

I can confirm that this error doesn't happen with this older commit ComparativeGenomicsToolkit@1d2d38c

To reproduce:

• install cactus binaries
• make a config.xml as described here (I've been meaning to add a command line but haven't gotten around to it yet)
• copy and unzip the input files from /public/home/hickey/dev/work/gpu-lastz on courtyard
• make sure segalign's in your path
• run cactus-blast ./jobstore 10mammalsplus.txt Anc05.cigar --root Anc05 --pathOverrides panTro6.fa.pp equCab3.fa.pp hg38.fa.pp canFam3.fa.pp felCat8.fa.pp --pathOverrideNames Chimp Horse Human Dog Cat --realTimeLogging --logInfo --retryCount 0 --maxCores 64 --cleanWorkDir never --configFile ./config.xml

[gsneha26]

@glennhickey I have fixed the error in the latest commit. Thanks!