[Bioperl-l] Bio::DB::Taxonomy HTTPget query problem
Hilmar Lapp
hlapp at gnf.org
Thu Mar 6 18:09:04 EST 2003
Bio::DB::Taxonomy provides a programmatic interface to NCBI's taxonomy
server. Apparently it uses Bio::Root::HTTPget to make the http request.
Here's the problem: if I run make test I get the following exception
from Taxonomy.t:
-------------------- WARNING ---------------------
MSG: Can't query website:
------------- EXCEPTION -------------
MSG: request failed: HTTP/1.1 502 Proxy Error
STACK Bio::Root::HTTPget::get blib/lib/Bio/Root/HTTPget.pm:155
STACK (eval) blib/lib/Bio/DB/Taxonomy/entrez.pm:207
STACK Bio::DB::Taxonomy::entrez::get_taxaid
blib/lib/Bio/DB/Taxonomy/entrez.pm:206
STACK toplevel t/Taxonomy.t:34
--------------------------------------
The first thing is I don't use a proxy and never have. Also, the 2
preceding queries work fine, their tests pass. Knowing our friends from
NCBI, I put in a sleep(10) before the last query - and voila, it works
just fine. What's apparently happening is that queries get blocked if
there are too many (more than 2) per time unit. This needs to be
accounted for. I also don't know whether NCBI actually provides
instructions about how to set up your own Taxonomy server. With Biosql
the underpinning database is soon going to be there, but you'd still
have to write the server.
Apart from these considerations, there are actually 3 problems here:
1) The error message from HTTPget is confusing. I assume it is indeed
the status line it gets from the HTTP server, so the problem would then
lie there and not in HTTPget, and there's probably little we can do
about it.
2) The line number in the stack trace is grossly wrong for HTTPget.
Apparently the style of __PACKAGE__->throw instead of $self->throw
yields wrong line numbers in the stack trace (maybe because comments
are not counted?). Is there a specific reason for __PACKAGE__, or can
it be replaced with $self? (ChrisD, Lincoln?)
3) Bio::DB::Taxonomy needs to adopt the delaying mechanism Lincoln
implemented for the entrez query code.
If no-one volunteers to take this on, I'll submit a bug report.
-hilmar
--
-------------------------------------------------------------
Hilmar Lapp email: lapp at gnf.org
GNF, San Diego, Ca. 92121 phone: +1-858-812-1757
-------------------------------------------------------------
More information about the Bioperl-l
mailing list