[Bioperl-l] StandAloneBlast.t testing

[Peter Schattner]

Todd Richmond wrote:

> StandAloneBlast.pm says:
>
> # You will need to enable Blast to find the Blast program. This can be done
> # in (at least) three ways:
> #  1. Modify your $PATH variable to include your Blast directory as in (for
> Linux):
> #    export PATH=$PATH:/home/peter/blast   or
>
> #1 doesn't work for me, even though $PATH clearly includes blastall. Is that
> just a MacOS X thing? I couldn't get the tests to run correctly until I set
> $BLASTDIR and $BLASTDATADIR, even though a standard command-line blast
> worked fine. (Same problem with clustalw - can't find the application even
> if it's on the path).

This is a bug.  It has nothing to do with OSX.  Way back when,
StandAloneBlast.pm  and Clustalw.pm could find their programs by any of these
three methods (at least under Linux).  However,  currently setting the $PATH
does not work , even under Linux.   At this point,  I think the simplest
solution is to change  the documentation to require setting the environmental
variable $BLASTDIR and $CLUSTALDIR (rather than modifying $PATH) - which I will
do unless someone suggests  a better approach.

>
>
> For the BLAST databases, the standard /data directory is supposed to be the
> one that you're required to specify in the .ncbirc file, right? Shouldn't
> you be able to leave that blank, and have blastall look in it's own default
> directory?

Yes, if the database  is in the /data blast subdirectory, StandAloneBlast.pm
should find it even if $BLASTDATADIR is left blank.  This works for me
currently on Linux.  Perhaps this is a "MacOS X thing".  I don't know.

>
>
> Is there
> some reason that the whole database needs to be searched, or can we put
> together a small preformatted database and test sequence in the data folder
> for this kind of test?

Well.  I guess I was  too lazy  to put together a small preformatted database.
Sounds like a good idea - I just don't have the time now.   I made the tests
run against  ecoli.nt  and swissprot.pr because there are different tests
whether a nuecleotide  or protein blast  is performed and those were the
databases I happened to be using at the time.  At the least Ithe documentation
should be  corrected so that it's clear that you need to obtain the test
databases.  I'll make that change  unless someone has the incliniation to offer
a more robust solution

- Peter