[Bioperl-l] results problem with StandAloneBlast

[Genevieve DeClerck]

Problem solved, albeit, in a slightly hacky way.

I tried to make seek() work for a good long while with the SearchIO 
blast results object, but I just couldn't get it to work. (Probably b/c 
seek wants to see a genuine file handle-- not a SearchIO filehandle.) I 
used SearchIO's fh() to get the handle and could while(<$fh>) through 
the data but when I used seek($fh,0,0) to reset the cursor position in 
the handle in prep for another loop, i got an error complaining about my 
use of seek() by indicating that "SEEK" could not be found in Seekable.pm.

I concluded that it was not going to be possible and instead made an 
array if SeqFeature objects which contain all the relevant blast output 
data (i.e. the m8/hit table stuff).

It still seems unfortunate that one can't reuse the SearchIO object for 
cases when the SearchIO blast report needs to be accessed mltiple times.

Thanks for your help,
Genevieve



Sendu Bala wrote:

> Genevieve DeClerck wrote:
> 
>>Thanks for your comment Sendu, it was very helpful. I think this must be 
>>what's going on.. I am using $blast_report->next_result in both 
>>subroutines. It appears that analyzing the blast results first w/ my 
>>sort subroutine empties (?) the $blast_result object so that when I try 
>>to print, there is nothing left to print. (and visa-versa when I print 
>>first then try to sort).
>>So, from the looks of things, using next_result has the effect of 
>>popping the Bio::Search::Result::ResultI objects off of the SearchIO 
>>blast report object??
> 
> 
> Not quite. It's more or less exactly like opening a file and then trying 
> to read it all twice like this:
> open(FILE, "file");
> while (<FILE>) {
>      print # prints each line in the file
> }
> while (<FILE>) {
>      print # never happens, we never enter this while loop
> }
> 
> To get the second while loop to print anything we need to say seek(FILE, 
> 0, 0) before it. Or in the first while loop store each line in an array, 
> and then make the second loop a foreach through that array.
> 
> 
> 
>>It seems I could get around this by making a copy of the blast report by 
>>setting it to another new variable...(not the most elegant solution) but 
>>I'm having trouble with this...
>>
>>If I do:
>>
>>    my $blast_report_copy = $blast_report;
>>
>>I'm just copying the reference to the SearchIO blast result, so it 
>>doesn't help me. How can I make another physical copy of this blast 
>>result object? Seems like a simple thing but how to do it is escaping me.
> 
> 
> Not really a good idea, and it may not work anyway if the object 
> contains a filehandle. But for a simple object you might recursively 
> loop through the data structure and copy each element out into a similar 
> data structure.
> 
> 
> 
>>But better yet, the way to go is to 'reset the counter,' or to find a 
>>way to look at/print/sort the results without removing data from the 
>>blast result object. How is this done though??
> 
> 
> It would be rather nice if this worked:
> my $blast_report = $factory->blastall($ref_seq_objs);
> my $blast_fh = $blast_report->fh();
> while (<$blast_fh>) {
>      # $_ is a ResultI object, use as normal
> }
> seek($blast_fh, 0, 0); # this would be great, but does it work?
> while <$blast_fh>) {
>      # go through the results again in your second subroutine
> }
> 
> An alternative hacky way of doing it, which may also not work, would be 
> to go through your $blast_report as normal, but then before going 
> through it a second time, say
> my $fh = $blast_report->_fh;
> seek($fh, 0, 0);
> 
> Finally, the most sensible way (assuming bioperl provides no methods of 
> its own for this) of solving the problem is, the first time you go 
> through each next_result, next_hit and next_hsp, just store the returned 
> objects in an array of arrays of arrays. Then the second time get the 
> objects from your array structure instead of with the method calls.
> _______________________________________________
> Bioperl-l mailing list
> Bioperl-l at lists.open-bio.org
> http://lists.open-bio.org/mailman/listinfo/bioperl-l
>