On Jul 16, 2008, at 10:16 AM, Ben Webb wrote:
> Daniel Russel wrote: >>> for example, if your magic number indicates big >>> endian, but the data are actually stored as little endian, your >>> output >>> will be messed up - but the reader works perfectly as it should do. >> Currently, as far as I can tell, there is no checking of any magic >> number in MRCReaderWriter. The only thing which controls the byte >> ordering is the hack having to do with looking for excessively large >> (and now excessively small) voxels. > > The machine stamp ('magic number') is not reliably written into all > MRC > files, hence the workaround to detect byte-swapped files. MRC files > written out, on the other hand, should have correct machine stamps. Ick. Currently only the header is byteswapped if the machine stamp says the endianness is wrong. It seems like the data probably should be too. And then the heuristic applied.