On 16.07.2015, at 19:46, Marshall Schor <[email protected]> wrote:
> The usual way these kinds of things have been decided is to "guess" if the
> more
> likely scenario is one of a user error, or a user intent. Here, the user
> seemed
> to intend this. If this was supported, there might be other unintended
> consequences - such as setting a subject-of-analysis **after** some
> Annotations
> were made.
Well, actually *that* is a case that should be considered supported.
E.g. uimaFIT comes with a JCasBuilder class that allows to incrementally
initialize a CAS, e.g:
JCasBuilder jb = new JCasBuilder(jCas);
jb.add("This sentence is not annotated. ");
jb.add("But this sentences is annotated. ", Sentence.class);
int begin = jb.getPosition();
jb.add("And", Token.class);
jb.add(" ");
jb.add("here", Token.class);
jb.add(" ");
jb.add("every", Token.class);
jb.add(" ");
jb.add("token", Token.class);
jb.add(" ");
jb.add("is", Token.class);
jb.add(".", Token.class);
jb.add(begin, Sentence.class);
jb.close();
The annotations are added directly to the CAS while the strings are
added to an internal StringBuilder which is only set as the CAS
document text when close() is called.
Very useful I might mention! I use the same technique in various readers.
Cheers,
-- Richard
