Optimizing FITS

January’s mostly over, and I’ve only posted three times to this blog. Files that Last has been keeping me busy. My posting should pick up again before long, once I get a draft out to first readers.

One thing I’ve been looking at, with an eye to the upcoming SPRUCE Hackathon, is things that can be done with FITS. I’ve written up the results of some profiling experiments and quick attempts at optimization. FITS puts together a lot of tools for extracting file metadata, but there have been some complaints that it’s not as fast as it might be. The first results were surprising; the easiest way to get a small improvement was to factor out the initialization of namespace URIs for parsing XML. You wouldn’t think that would make any detectable difference, but the initialization of URIs in Xerces is surprisingly slow.

Another possibility to explore is improving the connection between FITS and JHOVE. Even though JHOVE is intended for use as a callable library, among other things, it’s designed to write to an output file. Some simple changes would let it provide an in-memory response without writing a file, which would be more useful to an application like FITS.

Comments are closed.