rawspeed
rawspeed copied to clipboard
klauspost
Rename 'BE' to "Swap",
Since it is really behaving as a swapper, and does only reads "Big Endian" on little endian platforms (albeit that is the wast majority).
Pedro, is this ok for you, if I merge this?
That's actually not the case. The files really are big endian or little endian. On little-endian architectures the bits are swapped on big endian they are not. It's the nature of the C shift operator that makes the operation architecture agnostic.
Yes - however, it is only "TiffIFDBE" and "TiffEntryBE". These classes actually swap endianness.
I am not referring to the function names in RawDecoder, which are correct, and will remain as they are.
Whether or not a TiffEntry is "TiffEntry" or "TiffEntryBE" doesn't actually say if the data is LE or BE. On a BE system it will be the other way around.
So I would like to rename them to avoid that confusion.
Ah, if that's the case then DcrDecoder is wrong. I used TiffIFDBE to mean "parse a big endian Tiff". I see that's wrong now as can be seen in TiffParser. It will break on BE architectures.
I'd argue it would be cleaner to change it the other way around though, to make TiffIFDBE mean "parse big endian" and TiffIFD "parse little endian" and have both of them swap bytes when they're in the opposite architecture.
But that's a more intrusive change. If you want to keep the current code changing them to TiffIFD and TiffIFDSwap makes sense. TiffIFDBE and TiffIFDLE could be implemented on top of those two by checking the host endianness.
I was looking at this while implementing a CIFF parser for CRW files. It would make a lot more sense to push the big/little stuff into stuff like FileMap and ByteStream and have TiffIFD and TiffEntry handle both types of files.
Yes, after some back&forth, I agree that would be both simpler and more reliable, and not something that should have any big performance impact.