Lubomir Bourdev wrote:
If your input is not a stream, the code doesn't use streams; it operates straight to FILE*. But if you start with a stream, then it uses a stream.
"Not a stream" does not always imply that it's a file. The case that I, and I think Domagoj, would like to see handled better is when the data is in memory. For example, libjpeg has a function jpeg_mem_src() which could be used as follows: void read_jpeg_image_from_mem(const char* begin, const char* end, gil::image dest) { ..... jpeg_mem_src(...., begin, end-begin); ..... } So there is basically no overhead. Wrapping a block of memory up in a stream, and then unwrapping it again to pass it to the image library's C API, and pulling in another Boost library like Boost.IOStreams to do so, looks like overkill. Regards, Phil.