Subject: Re: [boost] RFC: edit_distance / edit_alignment library
From: Bjorn Reese (breese_at_[hidden])
Date: 2013-12-17 05:08:57
On 12/16/2013 10:45 PM, Erik Erlandson wrote:
> Come to think of it, now that the interface sits on top of boost::parameter, it ought to be possible to simply present the single 'edit_distance' function, and enable the path-output implementations iff an '_output = outputter_class' is provided.
Very good idea.
> Tangentially, I considered 'edit_cost' instead of 'edit_distance' mostly because in the general case, the function may violate the definition of a distance metric, for example if the cost of insertion is different than the cost of deletion. In that case, edit_cost(seq1, seq2) != edit_cost(seq2, seq1).
Which reminds me... If you are running without substitution (_allow_sub
set to false), then we may need to have a lower cost for deletions than
for insertions, because for substitutions we want a script with
delete(A) followed by insert(A), rather than insert(A) followed by
delete(A). The latter may result in 'A' being removed rather than
replaced.