Extensions of FileIO interface

[timholy]

I excised ec2dec5, reverted the commit (#370) that caused the test failure, created some deprecations, and moved Imagine into its own package.

Only AndorSIF to go, right? Do you want me to tackle that?

[SimonDanisch]

Awesome!!!
AndorSIF is mostly ready and it might be better if you do the rest. I don't know the format and couldn't find any files to test this. I think it's fine to leave it in Images, as the other IO packages rely on Images anyways. Up to you ;)

[timholy]

OK. @rsrock, I'm going to take the liberty of moving the SIF reader over to JuliaIO, and make sure you have push privileges.

[timholy]

https://github.com/JuliaIO/AndorSIF.jl

Once this passes its tests, I'm merging 😄.

[SimonDanisch]

Ha! That would be quite something :)
Just remembered though, that the reading Vector{Uint8} hasn't made it into ImageMagick.jl yet.
Are we forgetting other stuff?

[SimonDanisch]

Win64 0.3 is off by one:

Expression: raw(res) --> raw(map(ScaleAutoMinMax(Ufixed16),raw(imgr))) 
Expected: 
Uint16[65535 43131
       49330 23688
       35780 41243]
Uint16[40516 36945
       43380 18967
       28964 0]
Occurred: 
Uint16[65535 43131
       49330 23688
       35780 41243]
Uint16[40516 36945
       43379 18967
       28964 0] 

This must be a rounding error due floating point precision problems?
Should we relax the test? Would be annoying if this lets the tests fail at random

[timholy]

I want to try to debug it first, so for now let's have it generate enough output to reproduce the problem.

[timholy]

The writemime test failure is expected until JuliaIO/FileIO.jl#37 gets merged.

[rsrock]

No problem, I'll catch up with all these changes when I have the chance. I assume osxnative is moving somewhere else too? Let me know where that will go, I'd like to add image writing too.

I doubt that Andor SIF is used much. At one point I tried to grab a small test image, but it was with the wrong version of the software and the format was different. Of course.

Ronald S. Rock, Jr., Associate Professor
Dept. of Biochemistry and Molecular Biology, The University of Chicago
GCIS W240, 929 E. 57th St.
Chicago, IL 60637 USA
+1 773.702.0716 (w), +1 773.702.0439 (f)
rrock@uchicago.edu

On Sep 27, 2015, at 2:47 PM, Tim Holy notifications@github.com wrote:

OK. @rsrock, I'm going to take the liberty of moving the SIF reader over to JuliaIO, and make sure you have push privileges.

—
Reply to this email directly or view it on GitHub.

[SimonDanisch]

Little checklist:

• SIF
• Imagine
• pass PBM tests
• writemime
• rounding (?) error
• OSXNative
• OSXNative rename
• NRRD
• ImageMagick
• merge Choose a default for ambiguous in FileIO

Probably incomplete ;)

[timholy]

OK, I think we're basically good to go here. This couples to the following PRs:

• Choose a default for ambiguous formats JuliaIO/FileIO.jl#38
• The new Netpbm package: https://github.com/JuliaIO/Netpbm.jl
• Pass old io tests JuliaIO/ImageMagick.jl#9

I think lots of travis tests will fail (surprisingly, this one succeeded) until packages are registered and tagged. OK to go ahead?

[timholy]

(I'm not sure what the "rounding error" means above.)

[SimonDanisch]

This one:

Expression: raw(res) --> raw(map(ScaleAutoMinMax(Ufixed16),raw(imgr))) 
Expected: 
Uint16[65535 43131
       49330 23688
       35780 41243]
Uint16[40516 36945
       43380 18967
       28964 0]
Occurred: 
Uint16[65535 43131
       49330 23688
       35780 41243]
Uint16[40516 36945
       43379 18967
       28964 0] 

[timholy]

Ah, thanks for reminding me. Yes, let's wait for it to happen again and see what the debugging output tells us.

[timholy]

Aside from a weird problem with QuartzImageIO, things are looking good. I'm interested in tagging a new (minor) release very soon. A key question: should Images (temporarily?) add

ImageMagick
@osx QuartzImageIO

to its REQUIRE file?

Also, if folks can test I'd be grateful.

[SimonDanisch]

Wow, is it Christmas already!? :)
On 29 Sep 2015 12:02, "Tim Holy" notifications@github.com wrote:

Merged #373 #373.

—
Reply to this email directly or view it on GitHub
#373 (comment).

[vchuravy]

💯 Whoo

[lucasb-eyer]

A Pkg.update(), followed by Pkg.checkout("Images") also installs FileIO. Subsequent Pkg.test("Images") succeeds. imread("some.png") nicely warns about deprecated imread and recommends load.

But with the new system, I'm back to having ImageMagick problems:

julia> load("/home/beyer/zealgo.png")
ERROR: ArgumentError: ImageMagick not found in path
 in require at ./loading.jl:233
 in check_loader at /home/beyer/.julia/v0.5/FileIO/src/loadsave.jl:11

A Pkg.free("Images") "fixes" this problem, e.g. imread works on my system in the latest release. I'll investigate this towards the end of the week.

Edit to clarify: I'm not sure it's this specific patchset that introduces the regression or any other since the last release, so the problem might well be unrelated!

[timholy]

Thanks for checking. Right now you have to say Pkg.add("ImageMagick"). See #373 (comment).

[lucasb-eyer]

Ah, right, I read that comment but comitted it to memory with a bitflip :) Adding the ImageMagick package, it all works fine on my Arch Linux for a few images I tried.

[timholy]

Awesome! Thanks for the report.

I think we can take this as evidence that we should add it to the REQUIRE 😄.

[lucasb-eyer]

That, or to the README.md if you want to force people to choose their loader.