Hi all,
I was testing a new workflow which I developed on the current stable release, Galaxy 15.03, and uploaded to the Tool Shed yesterday:
https://github.com/peterjc/galaxy_blast/tree/master/workflows/blast_top_hit_... http://toolshed.g2.bx.psu.edu/view/peterjc/blast_top_hit_species
I then switched to my development instance running the dev branch from github,
1. Log in as an Admin 2. Install http://toolshed.g2.bx.psu.edu/view/peterjc/blast_top_hit_species 3. Upload suggested test dataset (in workflow README file): http://nematode.net/Data/nacobbus_aberrans_transcript_assembly/N.abberans_re... 4. Click on "All Workflows" (bottom of left pane tool list) 5. Click on "Species of top BLAST hits (imported from uploaded file)" 6. On preview screen, examine step 3, blastx 7. Notice it will run against Protein BLAST database "nr" 8. Click "Run workflow"
As noted in the README file, this workflow assumes the local blastdb_p.loc file includes an entry "nr" for a current/recent mirror of the NCBI "non-redundant" protein BLAST database.
Initially my development galaxy had an empty blastdb_p.loc file, so I expected the workflow to spot this problem at the preview step (before ever clicking "run workflow").
Instead it proceeds and attempted to run BLASTX without no database (which of course failed):
blastx -query "/mnt/galaxy/repositories/galaxy/database/files/000/dataset_81.dat" -db "" -query_gencode 1 -evalue 0.001 -out "/mnt/galaxy/repositories/galaxy/database/files/000/dataset_82.dat" -outfmt "6 qseqid sseqid pident length mismatch gapopen qstart qend sstart send evalue bitscore staxids sscinames scomnames sblastnames sskingdoms" -num_threads "${GALAXY_SLOTS:-8}" -strand both -matrix BLOSUM62 -seg yes -max_target_seqs 1
That's the bad news: It fails, but it fails with a reasonably clear error:
Fatal error: Exit code 2 () BLAST Database error: Database name is required.
Good news: If I update blastdb_p.loc to add the "nr" database it works.
Very bad news: If I update blastdb_p.loc to add other databases (but not an entry for "nr"), the workflow defaults to the first database in blastdb_p.loc
Summary: Workflow requesting key entry "xxx" from example.loc (a) example.loc empty, gets empty path, fails cleanly. (b) example.loc contains xxx, gets correct path, works. (c) example.loc non empty but lacks xxx, runs with WRONG path.
Peter