Read Operations

[ashwani-rathee]

This will include read operations on tags like:

• readmetadata() : read all tags
• readtags() : reads single or multiple tags

Focus:

• Filepath, IO, data support is to be provided
• benchmark still to be done properly, with docs and tests proper
• improvement in code in general is also required
• no memory related issues after processing.
• Improvement in generalizability of these functions

tags.jl need to export all tags which it doesn't yet and need to decide on the names(IMAGE-WIDTH still looks long and maybe there is a better way to do it), there are some issue with clang.jl wrapper generation that needs a fix so I use a old version from here: https://github.com/ashwani-rathee/io-project/blob/main/exifviewer/gen/LibExif.jl Also need to check if we have some image availabe in testimages.jl for testing purpose or do I need to include it?

Able to do:

julia> read_metadata("test/test_images/test.jpg")
Dict{Any, Any} with 12 entries:
  "EXIF_TAG_DOCUMENT_NAME"     => "Test_Image"
  "EXIF_TAG_ORIENTATION"       => "Bottom-right"
  ⋮                            => ⋮

julia> read_tags("test/test_images/test.jpg", tags=[ARTIST, Y_RESOLUTION, X_RESOLUTION])
Dict{Any, Any} with 3 entries:
  "EXIF_TAG_Y_RESOLUTION" => "72"
  "EXIF_TAG_ARTIST"       => "Ashwani"
  "EXIF_TAG_X_RESOLUTION" => "72"

[johnnychen94]

Feel free to add new images to TestImages.jl.

[ashwani-rathee]

Couple of things to cover in next commit:

• [x] combine read_metadata and read_tags
• [x] fix the annoying behavior of allifds and ifds
• [x] study the memory allocations and how to best deallocate/deref, also use try/finally block
• [x] workout the maker note basics
• take time to understand the libexif code
• find sources of errors and focus on tests, write better tests

Also if possible:

• study support on other file type given libexif supports directly reads only jpeg and raw exif data(this opens gateway to other types like  .TIF, .WAV, .PNG .WEBP cause they embed in form of data blob so we need to find the location of blob which can be directly understood by libexif)

[ashwani-rathee]

Getting permission for some more test images from libexif team: https://github.com/libexif/libexif/issues/88 I tested mnote functionality on https://github.com/libexif/libexif/blob/master/test/testdata/canon_makernote_variant_1.jpg and it works

julia> a = read_tags(filepath1; read_all=true,read_mnote=true)
Dict{String, Any} with 105 entries:
  "MNOTE_FLASH_EXPOSURE_COMPENSATION" => "0.00 EV"
  "MNOTE_SELF-TIMER"                  => "Off"
  "MNOTE_FOCUS_RANGE"                 => "Auto"
  "MNOTE_FOCAL_PLANE_X_SIZE"          => "7.39 mm"
  "MNOTE_EXPOSURE_COMPENSATION"       => "0.00 EV"
  "EXIF_TAG_SENSING_METHOD"           => "One-chip color area sensor"
  "MNOTE_EASY_SHOOTING_MODE"          => "Landscape"
  ⋮                                   => ⋮

And yes we have their permission now!

[ashwani-rathee]

Not entirely sure why the segmentation faults are happening now tbh, sometimes I am able to read all the maker note tags from those new test images(https://github.com/ashwani-rathee/exif-sampleimages) but sometimes it all just starts failing. But I think it comes from here when reading the maker note tags, also I have seen the mnote tags fail in tests but they generally seem to work in repl.

mnote = LibExif.exif_mnote_data_get_name(md_ptr, i) # so it give a pointer to location of name of tag but i am unable to 
# decide how far to read so its like [83,100,91,63,0,120...] and we have mnote which is address of 83, 
# so ideallly we should have size of name, so we just read till 0 and then convert it to UInt8 and then to string
if (mnote == C_NULL) continue end
data = unsafe_wrap(Array, mnote, 30) # but given we don't have size of name(haven't been able to find it), i read 30 address which might 
# not the smart thing to do, haven't been able to find the size of name in documentation

Or I am missing some understanding of resources management, freeing pointers or something like that

Also, I think its best if we remove fixformat for now, there are some conspicous errors showing up there too.

[johnnychen94]

@ashwani-rathee I think this PR is almost finished, can you try to finish the unresolved comments so that we can merge it (for GSoC evaluation purposes)?

[johnnychen94]

@ashwani-rathee nice work!