Information query about juicy.pixels HOT 9 CLOSED

definitely. not only that, but files load into a DynamicImage, and then i have to patternmatch on all those constructors to get an Image that i can imageWidth, etc on. why isn't DynamicImage a typeclass instead?

from juicy.pixels.

also, instead of a bunch of save* functions, why isn't file type an argument?
saveImage :: FilePath -> ImageType -> DynamicImage -> IO ()

data ImageType = PNG | JPG Int | BMP ...

from juicy.pixels.

DynamicImage can't be a type class, as the image decoders don't know at compilation time which color space the will have to decode (especially for the PNG case). For your use case, it's easy to add a function of the following type

{-# LANGUAGE Rank2Types #-}
withDynamicImage :: (foral pixel. Image pixel -> a) -> DynamicImage -> a

allowing something like withDynamicImage imageWidth

For the saveImage, I don't see any real advantage of providing it as the intended use case is to serialize to a fixed image format, but patches are welcome.

from juicy.pixels.

i'm not clear why your reason precludes a typeclass, but i might be missing something. is there any way to write withDynamicImage without a million patternmatches (which is just doing what a typeclass would do, but in a closed/brittle way)? in any case, why not include it in the library?

regarding saveImage, consider the use case (mine) of writing a thumbnailer -- doesn't care what the original format was, just wants to write out the same format. don't you want to let me write that in a generic way?

from juicy.pixels.

I don't know how to write decodeImage :: B.ByteString -> Either String DynamicImage with a type class, with the constraint that the caller doesn't know the pixel type (because it can only be known by the decoding function).

For the withDynamicImage, it's the lack of type mainly, but I accept pull requests.

For your specific use case, why not use the imageTo* functions? It seems to me that it answer your use case.

from juicy.pixels.

Since this issue is a bit old, is there any way to determine image width from a dynamic image with a function call? (And your link to imageTo functions seems to have gone dead.)

from juicy.pixels.

The dynamicMap could help extract width and height of a fully extracted image, I've updated the link to the imageTo functions

from juicy.pixels.

Thanks, that was perfect.

from juicy.pixels.

Done with addition of Metadata.

from juicy.pixels.