[Biojava-l] Editing SymbolLists

[David Waring]

Before I sit down an write the EditableSymbolList, I would like to get some
consensus on it. You suggest, why not just replace the existing
SimpleSymbolList with an editable implementation. This makes sense to me. I
note that there are comments in both Abstract and SimpleSymbolList stating a
desire to make them immutable e.g. in SimpleSybmolList a comment states that
"addSymbol is deprecated Can we make it immutable?". Is this still a desire
of some? Should we have one SimpleSymbolList implementation, and let
ChangeListeners determine whether it was editable or not? Is there a problem
out there were people can accidently edit a SymbolList? If SimpleSymbolList
were to suddenly support edit, would anything break?

I like the idea of a copy constructor that Thomas suggested. I might even go
so far as suggest a String constructor too.
	seq = new EditableSymbolList(alphabet,"gaattc");


It is my guess that even when a SymbolList will be edited, most of the
activity will still involve reading, and therefore an array based
implementation is still going to be fastest. Implementations that make
editing faster will certainly make reading slower. Do people aggree?

David

>
> On Fri, Aug 24, 2001 at 06:09:57PM -0700, David Waring wrote:
> > Perhaps I am missing something big here, but I can not figure
> out how in the
> > world I can edit a Sequence or a SymbolList.
> >
> > There is a class Edit, and there seems to be strong support for handling
> > changes and passing them to anything that is listening, but
> SimpleSymbolList
> > is immutable (even the addSymbol method is deprecated).
> SimpleSequence wraps
> > a SimpleSymbolist so that is immutable (at least the SymbolList is). The
> > other subclasses of SymbolList don't seem to be the
> appropriate. As best I
> > can tell if I wanted to be able to read a fasta file, and make
> a change to
> > the sequence, I would have to.
> >
> > Create an EditableSymbolList.
> > Create a FastaSequenceBuilder that used this.
> > Write a new method or two in SeqIOTools to read my file into my
> new Sequence
> > that has an EditableSymbolList.
> >
> > Is this correct? Does no one else have any desire to edit a SymbolList?
>
> The API for editing has existed on the SymbolList interface for
> a long time, but up until now, there seems to have been very
> little interest, hence no mutable implementation.
>
> It would be really good to have an EditableSymbolList (better
> still, replace SimpleSymbolList with an editable implementation).
> If you feel like coding this up, we'd be really interested to
> here about it!
>
> As you say, you'll probably then also want a special SequenceBuilder
> implementation which creates these editable implementations.  An
> alternative, rather simpler, approach would be to just make sure
> that the editable SymbolList has a copy-constructor, just as
> the current SimpleSymbolList does.  Then you can do:
>
>   SequenceIterator si = SeqIOTools.readFastaDNA(stream);
>   Sequence seq = si.nextSequence();
>
>   SymbolList editableSeq = new SimpleSymbolList(seq); // mutable copy
>   editableSeq.edit(new Edit(20, 10, SymbolList.EMPTY_LIST));
>
> Or something similar...
>
> Does this make sense?
>
>
> > PS: ChunkedSymbolList does not appear in the javadoc anywhere,
> what gives?
>
> ChunkedSymbolList is a private implementation -- the only way
> you can get hold of them is via the ChunkedSymbolListBuilder.
>
> ChunkedSymbolLists are special, optimized implementations,
> designed for the case of loading large sequences from files
> or other streams -- the hope was that the majority of users
> wouldn't need to know about them.
>
>
>    Thomas.