This package provides native Go functionality to parse an existing EXIF block, update an existing EXIF block, or add a new EXIF block.
To get the project and dependencies:
$ git clone https://github.com/dsoprea/go-exif.git
$ cd v3
- Go >= 1.17: Due to a breakage with "go test", we only officially support 1.17 for testing/CI reasons. It may still work in earlier versions if such a need is critically required, however.
This project is concerned only with parsing and encoding raw EXIF data. It does not understand specific file-formats. This package assumes you know how to extract the raw EXIF data from a file, such as a JPEG, and, if you want to update it, know how to write it back. File-specific formats are not the concern of go-exif, though we provide exif.SearchAndExtractExif and exif.SearchFileAndExtractExif as brute-force search mechanisms that will help you explore the EXIF information for newer formats that you might not yet have any way to parse.
That said, the author also provides the following projects to support the efficient processing of the corresponding image formats:
See the SetExif example in go-jpeg-image-structure for practical information on getting started with JPEG files.
The package provides a set of working examples and is covered by unit-tests. Please look to these for getting familiar with how to read and write EXIF.
Create an instance of the Exif type and call Scan() with a byte-slice, where
the first byte is the beginning of the raw EXIF data. You may pass a callback
that will be invoked for every tag or nil if you do not want one. If no
callback is given, you are effectively just validating the structure or parsing
of the image.
Obviously, it is most efficient to properly parse the media file and then
provide the specific EXIF data to be parsed, but there is also a heuristic for
finding the EXIF data within the media blob, directly. This means that, at least
for testing or curiosity, you do not have to parse or even understand the
format of image or audio file in order to find and decode the EXIF information
inside of it. See the usage of the SearchAndExtractExif method in the
example.
The library often refers to an IFD with an "IFD path" (e.g. IFD/Exif, IFD/GPSInfo). A "fully-qualified" IFD-path is one that includes an index describing which specific sibling IFD is being referred to if not the first one (e.g. IFD1, the IFD where the thumbnail is expressed per the TIFF standard).
There is an "IFD mapping" and a "tag index" that must be created and passed to the library from the top. These contain all of the knowledge of the IFD hierarchies and their tag-IDs (the IFD mapping) and the tags that they are allowed to host (the tag index). There are convenience functions to load them with the standard TIFF information, but you, alternatively, may choose something totally different (to support parsing any kind of EXIF data that does not follow or is not relevant to TIFF at all).
This project is configuration driven. By default, it has no knowledge of tags and IDs until you load them prior to using (which is incorporated in the examples). You are just as easily able to add additional custom IFDs and custom tags for them. If desired, you could completely ignore the standard information and load totally non-standard IFDs and tags.
This would be useful for divergent implementations that add non-standard information to images. It would also be useful if there is some need to just store a flat list of tags in an image for simplified, proprietary usage.
There is a runnable reading/dumping tool included:
$ go install github.com/dsoprea/go-exif/v3/command/exif-read-tool@latest
$ exif-read-tool --filepath "<media file-path>"
Example output:
IFD-PATH=[IFD] ID=(0x010f) NAME=[Make] COUNT=(6) TYPE=[ASCII] VALUE=[Canon]
IFD-PATH=[IFD] ID=(0x0110) NAME=[Model] COUNT=(22) TYPE=[ASCII] VALUE=[Canon EOS 5D Mark III]
IFD-PATH=[IFD] ID=(0x0112) NAME=[Orientation] COUNT=(1) TYPE=[SHORT] VALUE=[1]
IFD-PATH=[IFD] ID=(0x011a) NAME=[XResolution] COUNT=(1) TYPE=[RATIONAL] VALUE=[72/1]
IFD-PATH=[IFD] ID=(0x011b) NAME=[YResolution] COUNT=(1) TYPE=[RATIONAL] VALUE=[72/1]
IFD-PATH=[IFD] ID=(0x0128) NAME=[ResolutionUnit] COUNT=(1) TYPE=[SHORT] VALUE=[2]
IFD-PATH=[IFD] ID=(0x0132) NAME=[DateTime] COUNT=(20) TYPE=[ASCII] VALUE=[2017:12:02 08:18:50]
...
You can also print the raw, parsed data as JSON:
$ exif-read-tool --filepath "<media file-path>" -json
Example output:
[
{
"ifd_path": "IFD",
"fq_ifd_path": "IFD",
"ifd_index": 0,
"tag_id": 271,
"tag_name": "Make",
"tag_type_id": 2,
"tag_type_name": "ASCII",
"unit_count": 6,
"value": "Canon",
"value_string": "Canon"
},
{
"ifd_path": "IFD",
...
The traditional method:
$ go test github.com/dsoprea/go-exif/v3/...
This release primarily introduces an interchangeable data-layer, where any
io.ReadSeeker can be used to read EXIF data rather than necessarily loading
the EXIF blob into memory first.
Several backwards-incompatible clean-ups were also included in this release. See releases for more information.
Features a heavily reflowed interface that makes usage much simpler. The undefined-type tag-processing (which affects most photographic images) has also been overhauled and streamlined. It is now complete and stable. Adoption is strongly encouraged.
EXIF has an excellently-documented structure but there are a lot of devices and manufacturers out there. There are only so many files that we can personally find to test against, and most of these are images that have been generated only in the past few years. JPEG, being the largest implementor of EXIF, has been around for even longer (but not much). Therefore, there is a lot of compatibility to test for.
If you are able to help by running the included reader-tool against all of the EXIF-compatible files you have, it would be deeply appreciated. This is mostly going to be JPEG files (but not all variations). If you are able to test a large number of files (thousands or millions) then please post an issue mentioning how many files you have processed. If you had failures, then please share them and try to support efforts to understand them.
If you are able to test 100K+ files, I will give you credit on the project. The further back in time your images reach, the higher in the list your name/company will go.
Thank you to the following users for solving non-trivial issues, supporting the project with solving edge-case problems in specific images, or otherwise providing their non-trivial time or image corpus to test go-exif:
- philip-firstorder (200K images)
- matchstick (102K images)
In addition to these, it has been tested on my own collection, north of 560K images.