Pull-parser mode

[mkmatlock]

It would be really nice for memory and processor limited applications to have a pull-parser mode, so we can do something like:

parser.init( new FileInputStream( new File("file.bib") ) );
while(parser.next()){
   BibTeXObject object = parser.get();
   ...
}

and stay within a small memory footprint, even when processing very large (5+ Mb) bibtex files.

[vruusmann]

Are you thinking about using JBibTeX on smart devices? Keeping 5 MB worth of BibTeX content in memory doesn't seem like a big problem even in a situation like that.

The main problem with pull API is that there's no easy way to automate final cross-reference resolution and consistency checks (e.g. BibTeXParser#resolveCrossReferences()). Would it be acceptable to have "crossref" fields left as null then?

[mkmatlock]

You could allow the user to extend and contribute a cross-reference lookup interface to the parser constructor. One possible user implementation of this interface would be to do a database lookup. I actually currently use such an interface with a factory to build models of BibTeX entries for my reference manager app when they are retrieved from an on-disk database.

Something like:

public interface IReferenceResolver {
    public BibTeXEntry resolveEntry(Key key);
    public BibTeXString resolveString(Key key);
}

My particular application is on the android platform, where my users have encountered memory limit problems. The platform can be very restrictive about memory usage, and these restrictions vary on a device by device basis. Resources for the application UI and code take up a large portion of the memory resources available, and having large bibtex databases in-memory sometime pushes it over the edge, and can also affect app performance.