And since you are using the pull-parser, you'll be going through quite a bit of work just to arrive at a count, then doing it all over again. Might there be some kind of recognizable keyword within the file that you could find using regular-expressions? Read the file a chunk at a time, count the number of occurrences in each chunk, add 'em up and pretend that this is your count even if it isn't exactly. Or, perhaps just as well and a great deal easier, give the user some other sign of "visible progress," such as printing a "." to the console after every some-n nodes have been processed.