Avoiding the consumption of InputStream when detecting mime type?

[iku000888]

Thank you for the wonderful library that works like a charm.

While using this I did notice that when detecting the mime type from an InputStream instance,
the bytes get consumed and I cannot simply reuse the same object.

There are work arounds available (mostly Java), and that is what InputStreams do,
but it would be nice if mime-type-of is free of side effects (at least from the app developer perspective).

Thanks!

P.S. Would greatly appreciate any recommended methods for re-using InputStream instances.

[michaelklishin]

@iku000888 that makes sense, please submit a PR that does that (using 100% Java interop if this is what you're most familiar with, that's fine).

[michaelklishin]

Given that nobody has messed with input stream mark, InputStream#reset should be sufficient to make it reusable for most intents and purposes. And I think it's perfectly fine for Pantomime to use it. WDYT?

[alexott]

I've used it in one of the project - it mostly worked, maybe except very big files, like big Zips, or something like - can't remember right now...

[iku000888]

@michaelklishin Thanks for the feedback!

Here is a bit more of investigation results before I form a PR.
It looks like mime-type-of consumes the stream for raw Java Input Streams,
but not for the clojure.java.io/inpu-stream (Which is a BufferedInputStream).

Maybe there could be a suggestion to the user to coerce the input stream instead?

;stream is consumed
user> (let [stream (java.io.FileInputStream. "project.clj")] [(mime-type-of stream) (mime-type-of stream)])
["text/plain" "application/octet-stream"]

;stream is not consumed
user> (let [stream (clojure.java.io/input-stream "project.clj")] [(mime-type-of stream) (mime-type-of stream)])
["text/plain" "text/plain"]

[michaelklishin]

@iku000888 as long as the result is an InputStream (which is an interface, not an implementation), it should work. If buffering helps your case, I think we can introduce it.