Why I'm doing this
I wanted to write a simple web app (purely client-side) which would stitch two images together. I tried to do it with <canvas>, but I soon found out that on iOS the <canvas> size can't be bigger than 4096x4096px.
So I thought that this is a perfect use case for some Rust through Wasm.
The problem
After some research, I settled on the image library and I wrote a prototype. It works just fine, maybe it's a bit slow but I can work on that later.
The problem is that the crate is so low-level that I have to take care of rotating the JPGs by myself. I found a good article about JPG orientation, found a simpler implementation in JS based on <cavas>, but I'm struggling to figure out what would be the best way to do it with the image crate.
Some possible solutions
The image::imageops module
There's the image::imageops module with functions like flip_horizontal_in_place or rotate180_in_place. With these two I could accomplish two out of the three operations outlined in that article without allocating creating a whole new ImageBuffer object. Flipping diagonally OTOH would need to be implemented as rotate 90 degrees clockwise then flip horizontally, the former of which creates a new ImageBuffer. Not the optimal solution but so far this is the only way forward that I can see.
The imageproc::geometric_transformations module
The imageproc::geometric_transformations has stuff like Projection and warp. Projection is stored as a matrix which might be useful to do the orientation like in the article, but I really know almost nothing about matrices. On top of that, the warp function says that "The returned image has the same dimensions as image", so I'm not really sure how I could use it for orientation. Maybe that's the warp_into's job?
Use some other crate for orienting the JPGs
I tried looking for other crates, but couldn't find any that would work with Wasm and work on buffers, not files. For me it just seems to be a common enough operation that someone would've taken care of that already, but perhaps I'm wrong. 
Is there any other solution that I missed? Do I need to implement that by hand or did I miss a crate that can already do that? Is using image::imageops the best thing to do here or can I use this library or imageproc in a way that would make it easier to orient an image than I described?
