[Biojava-l] Wrapping SimpleGappedSequence

[Richard Holland]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

The easiest way is simply for me to alter the constructor to
SimpleGappedSequence (and equivalently to SimpleGappedSymbolList) to
copy all gaps if passed another instance of GappedSymbolList as the
parameter. I've just done this in CVS so you should be able to update
your copy and observe the new behaviour.

cheers,
Richard

Ditlev Egeskov Brodersen wrote:
> Hi again,
> 
>   thanks for the info - will do the check just to be proper. I have another
> question: In my application, I would like to wrap the retrieved
> SimpleGappedSequence objects inside another object that extends the
> functionality with application-specific stuff. Ideally, I would do this by
> extending the SimpleGappedSequence object and create it by passing the
> SimpleGappedSequence from the alignment import to the constructor of the
> parent, like so:
> 
>   class AlignedSequence extends SimpleGappedSequence {
>     public AlignedSequence(SimpleGappedSequence aGapped) {
>       super(aGapped);
>     }
> 
>     ..custom stuff..
>   }
> 
> However, the problem is that there is only one constructor for the
> SimpleGappedSequence, one which takes a simple Sequence object. I can pass
> the derived class alright, but all gap information is lost again, presumably
> because the SimpleGappedSequence constructor just takes out the seqString()
> and puts it into its own sequence object.
> 
> Shouldn't the constructor of the SimpleGappedSequence class recognise when a
> derived (and gapped) sequence object is passed, and process it accordingly?
> 
> As it stands, I am forced to include the SimpleGappedSequence as a private
> member of the AlignedSequence class, which is not near as nice since all
> statement using the class will have to do something like
> 
>   class AlignedSequence extends SimpleGappedSequence {
>     private SimpleGappedSequence gapped_sequence;
> 
>     public AlignedSequence(SimpleGappedSequence aGapped) {
>       gapped_sequence = aGapped;
>     }
> 
>     public SimpleGappedSequence getGappedSequence() {
>       return(gapped_sequence);
>   }
> 
>     ..custom stuff..
>   }
> 
>   ...
> 
>   AlignedSequence aAligned = new AlignedSequence(aGapped);
>   aAligned.getGappedSequence().seqString();
> 
> rather than simply:
> 
>   AlignedSequence aAligned = new AlignedSequence(aGapped);
>   aAligned.seqString();
> 
> In other words, is there any solution with the current setup that would
> allow me to extend SimpleGappedSequence and not loose the gap information?
> 
> --  Ditlev
> 
> --
>  
> Ditlev E. Brodersen, Ph.D.
> Lektor, Associate Professor
>  
> Department of Molecular Biology   Office:  +45 89425259
> University of Aarhus              Lab:     +45 89425022
> Gustav Wieds Vej 10c              Fax:     +45 86123178
> DK-8000 Aarhus C                  Email:   deb at mb.au.dk
> Denmark                           Lab WWW: www.bioxray.dk/~deb
> 
> 
>> -----Original Message-----
>> From: Richard Holland [mailto:holland at ebi.ac.uk]
>> Sent: 16 November 2007 10:50
>> To: Ditlev Egeskov Brodersen
>> Cc: biojava-l at biojava.org
>> Subject: Re: [Biojava-l] Parsing exising gaps
>>
>>>>   The returned gapped sequences are all properly set up with gaps,
> name etc.
>>>> But as for other users, I think there may be some problems, since the
>>>> SimpleAlignment object only has a general symbol list iterator, the
> user
>>>> will have to cast each statement extracting a sequence object, and
>>>>
>>>>       SimpleSequence aSimple = (SimpleSequence)aSequences.next();
>>>>
>>>> returns an ClassCastException at run time. So old code might not run
> with
>>>> the update as far as I can see.
> This is true. However, such code would be unsupported by us as the API
> clearly states that SimpleAlignment returns SymbolList instances, and
> does not make any guarantees about the exact implementation details of
> the objects it returns. To attempt to cast it to anything other than
> SymbolList would be a mistake! (Although actually it is now returning a
> guarantee of GappedSymbolList, which is what your code can now take
> advantage of). To assume it will return SimpleSequence is outside the
> behaviour defined by the API and therefore should not be relied upon.
> 
> A more correct behaviour would be to test each item returned:
> 
> 	SymbolList symlist = aSequences.next();
> 	if (symlist instanceof SimpleSequence) {
> 		SimpleSequence seq = (SimpleSequence)symlist;
> 		// Do simple-sequence stuff
> 	} else {
> 		// Do something else!
> 	}
> 
> In future, I will modify the API to change the SymbolList guarantee to
> a
> GappedSymbolList guarantee, but I can't do this right now as this
> really
> would break everyone's code!
> 
> We are currently planning a redesign as you may be aware, so issues
> like
> this will hopefully be resolved as part of that process. For a start,
> if
> we use Java 5 generics in future as we plan, we can strictly specify
> what kinds of objects will be returned by things such as the alignment
> API, making it easier for us to enforce API-compliant behaviour in
> user's code.
> 
> cheers,
> Richard
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.2.2 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFHPXUh4C5LeMEKA/QRAsbqAKCnpCRnIiztjZ69fE2/UaJuI9QjiACfYa0m
8EJTzWZYOyjp9VhmvsgvmNA=
=1uaB
-----END PGP SIGNATURE-----