well, it went fast :-) already ported. regarding strictness, i hope just doing that is enough: data IfEntry = IfEntry { entryTag :: !Word16, entryFormat :: !Word16, entryNoComponents :: !Word32, entryContents :: !Word32 } deriving Show Though from past experience, I'm far from sure. I'll add comments soon and then push it to hackage. emmanuel On Sun, Apr 13, 2014 at 5:08 PM, Emmanuel Touzery wrote:

On Sun, Apr 13, 2014 at 4:30 PM, Yitzchak Gale wrote:

...

Emmanuel Touzery wrote:

...

I could not find a pure haskell library to parse EXIF so I wrote one (pretty basic so far): https://github.com/emmanueltouzery/hsexif

Very nice! Why not upload it to hackage to make it easier for others to share and collaborate?

Yes, that's the plan as soon as I get the basics right and that I write some short documentation. Hopefully very quickly.

...

...

I wrote it with binary-strict. I also considered binary, but I would like the library not to throw an exception if the file that it's given is not a JPEG or is a JPEG without EXIF, but rather return an Either. I didn't manage to do that with binary

Perhaps the function Data.Binary.Get.runGetOrFail is what you are looking for? Or perhaps the incremental strict interface?

Ah yes... Actually I was a bit mislead because I found an haskell wiki stating that with binary it was impossible to catch the exceptions except in the IO monad and I took that at face value. I think runGetOrFail is what I want, I'll test it, thank you!

...

...

And then I realized that binary-strict was last updated in 2010, it seems cereal is recommended for strict binary file parsing nowadays. So, what should I do? Leave it as it is? Port to cereal? Port to binary?

The binary-strict library is no longer maintained, since binary now also provides a strict interface. Both binary and cereal are good choices.

Probably porting to binary would be easiest if you've already written it for binary-strict, assuming runGetOrFail or the incremental interface does what you want.

Ok, then I'll try the non-strict binary first then. Most JPG files will be only a couple of megabytes and it wouldn't be THAT bad loading them entirely in memory but then, it seems a bit of a shame. With lazy though I'll have to work a bit on my strictness annotations I think.

Thanks for the hints! Hopefully I can port it to lazy binary and publish it on hackage soon enough...

Emmanuel