Twinside
commented
9 years ago Right now, the answer is clearly no, every format working at bit level, violently extract a big bytestring from binary and put it in a bit reader structure (for performance reasons). It could be interesting, but it requires a major rewriting of nearly all file parsers, which is not likely to happen soon (for motivation reasons :])
Can the various decoding functions be augmented to support incremental input, such as is supported in the Data.Binary library used under the hood? This would allow decoding images from non-file inputs without having to load the entire image file into memory, and then parse the entire image file in memory.