If a file extension is explicitly specified in the upload name, it is
always used directly. Detection of common file extension combinations is
also performed. Currently, only ".tar.gz" and ".tar.xz" are detected. If
you would like to add support for more common combinations, please open
an issue or pull request.
If no file extension is explicitly specified, jaf falls back to MIME
type detection via the github.com/gabriel-vasile/mimetype library.
EXIF scrubbing can be enabled via the `ScrubExif` config key. When
enabled, all standard EXIF tags (as defined by the IFD mappings in the
go-exif library) are removed on uploaded JPEG and PNG images.
The `ExifAllowedIds` and `ExifAllowedPaths` config keys can be used to
selectively allow specific tags to survive the scrubbing. This can be
useful if you want to preserve image orientation information for
example. The IDs for standard tags can be found in [1].
The path specification for `ExifAllowedPaths` relies on the format
implemented in go-exif which is "documented" in machine-readable format
in [2]. Multiple paths can be specified, separated by a space. The
path format is as follows:
1. For tags in the main section: `IFD/<GroupName>/<FieldName>`.
Examples: `IFD/Orientation`, `IFD/Exif/Flash`,
`IFD/GPSInfo/GPSTimeStamp`. You will probably want to use both
[1] and [2] in combination if you plan to specify allowed tags by
path.
2. Tags in the thumbnail section follow the same format but paths
start with `IFD1/` instead of `IFD`.
[1]: https://exiv2.org/tags.html
[2]: a6301f85c8/assets/tags.yaml
The value for FileDir was updated in PR #3 but we missed also updating
the test in config_test.go. This is a hint that we should maybe setup a
GitHub Action to run tests on PRs. :-)
The benefit of keeping all managed file names in memory instead of
checking on demand does not outweigh the increased memory usage.
Additionally, this method allows users to manually move files into the
served directory without fearing they might be overwritten by jaf.
