Any performance benefit of using streams?

[rexxars]

Hi,

I've been using the streams API (piping to sharp, piping result to a stream) for my current application, but have certain cases where using streams is a bit of a pain.

For instance, I want to check the format when the metadata event is emitted, and if the file is an SVG, I don't want to transform the image at all, but just directly pipe it to the response stream, but since the input stream is already consumed by sharp, I can't get hold of the content anymore.

The alternative would of course be to just buffer the entire input into memory, so I could reuse the Buffer instance. I'm wondering if I would lose any performance doing it this way, though. I've read a couple of issues regarding streams which leads me to believe that at least output streams are not actually.. streaming. Is the same true for input streams?

[lovell]

The tl;dr is no, but please do read #179 for discussion about adding stream/pipe support to libvips.

You can get away with calling metadata() on an incomplete image Buffer as long as the complete image header is present. See #298 for building this logic into sharp itself.

[rexxars]

Thanks for the reply. I'll subscribe to updates on the streaming issue and hope that it will be possible in some way in the future.