Image Metadata

Thumbor uses py3exiv2 to read and write image metadata.

If the Py3exif2 Python library is available, the PIL engine also stores image metadata in engine.metadata.

Reading and writing Metadata

This part is copied from the py3exiv2 Tutorial

Let’s retrieve a list of all the available EXIF tags available in the image:

>>> engine.metadata.exif_keys
['Exif.Image.ImageDescription',
 'Exif.Image.XResolution',
 'Exif.Image.YResolution',
 'Exif.Image.ResolutionUnit',
 'Exif.Image.Software',
 'Exif.Image.DateTime',
 'Exif.Image.Artist',
 'Exif.Image.Copyright',
 'Exif.Image.ExifTag',
 'Exif.Photo.Flash',
 'Exif.Photo.PixelXDimension',
 'Exif.Photo.PixelYDimension']

Each of those tags can be accessed with the [] operator on the metadata, much like a python dictionary:

>>> tag = metadata[b'Exif.Image.DateTime']

The value of an ExifTag object can be accessed in two different ways: with the raw_value and with the value attributes:

>>> tag.raw_value
'2004-07-13T21:23:44Z'

>>> tag.value
datetime.datetime(2004, 7, 13, 21, 23, 44)

The raw value is always a byte string, this is how the value is stored in the file. The value is lazily computed from the raw value depending on the EXIF type of the tag, and is represented as a convenient python object to allow easy manipulation.

Note that querying the value of a tag may raise an ExifValueError if the format of the raw value is invalid according to the EXIF specification (may happen if it was written by other software that implements the specification in a broken manner), or if pyexiv2 doesn’t know how to convert it to a convenient python object.

Accessing the value of a tag as a python object allows easy manipulation and formatting:

>>> tag.value.strftime('%A %d %B %Y, %H:%M:%S')
'Tuesday 13 July 2004, 21:23:44'

Now let’s modify the value of the tag and write it back to the file:

>>> import datetime
>>> tag.value = datetime.datetime.today()

>>> engine.metadata.write()

Similarly to reading the value of a tag, one can set either the raw_value or the value (which will be automatically converted to a correctly formatted byte string by pyexiv2).

You can also add new tags to the metadata by providing a valid key and value pair (see exiv2’s documentation for a list of valid EXIF tags):

>>> key = 'Exif.Photo.UserComment'
>>> value = 'This is a useful comment.'
>>> engine.metadata[key] = pyexiv2.ExifTag(key, value)

As a handy shortcut, you can always assign a value for a given key regardless of whether it’s already present in the metadata. If a tag was present, its value is overwritten. If the tag was not present, one is created and its value is set:

>>> engine.metadata[key] = value

The EXIF data may optionally embed a thumbnail in the JPEG or TIFF format. The thumbnail can be accessed, set from a JPEG file or buffer, saved to disk and erased:

>>> thumb = engine.metadata.exif_thumbnail
>>> thumb.set_from_file('/tmp/thumbnail.jpg')
>>> thumb.write_to_file('/tmp/copy')
>>> thumb.erase()
>>> engine.metadata.write()

Installation

Pyexiv2 depends on the following libraries:

On OSX you can use homebrew to install the dependencies:

brew install boost --with-python
brew install boost-python
brew install exiv2

pip install py3exiv2

If you are updating thumbor and already have an existing virtualenv, then you have to recreate it. If you have both a System Python and a Homebrew Python with the same version, then make sure the Virtualenv uses the Homebrew Python binary.

On Linux Pyexiv2 can be installed with apt-get:

apt-get install python-all-dev libboost-python-dev libexiv2-dev pip install py3exiv2

pyexiv2.metadata API reference

class pyexiv2.metadata.ImageMetadata(filename)[source]

A container for all the metadata embedded in an image.

It provides convenient methods for the manipulation of EXIF, IPTC and XMP metadata embedded in image files such as JPEG and TIFF files, using Python types. It also provides access to the previews embedded in an image.

__delitem__(key)[source]

Delete a metadata tag for a given key.

Raise KeyError if the tag with the given key doesn’t exist

Args: key – the metadata key in the dotted form

familyName.groupName.tagName where familyName may be one of exif, iptc or xmp.

__getitem__(key)[source]

Return a metadata tag for a given key.

Raise KeyError if the tag doesn’t exist

Args: key – metadata key in the dotted form

familyName.groupName.tagName where familyName may be one of exif, iptc or xmp.

__setitem__(key, tag_or_value)[source]

Set a metadata tag for a given key.

If the tag was previously set, it is overwritten. As a handy shortcut, a value may be passed instead of a fully formed tag. The corresponding tag object will be instantiated.

Raise KeyError if the key is invalid

Args: key – metadata key in the dotted form

familyName.groupName.tagName where familyName may be one of exif, iptc or xmp.

tag_or_value – an instance of the corresponding family of metadata

tag or a value Type: pyexiv2.exif.ExifTag instance or

pyexiv2.iptc.IptcTag instance or pyexiv2.xmp.XmpTag instance or any valid value type

property buffer

The image buffer as a string. If metadata has been modified, the data won’t be up-to-date until write() has been called.

property comment

The image comment.

copy(other, exif=True, iptc=True, xmp=True, comment=True)[source]

Copy the metadata to another image.

The metadata in the destination is overridden. In particular, if the destination contains e.g. EXIF data and the source doesn’t, it will be erased in the destination, unless explicitly omitted.

Args: other – the destination metadata to copy to (it must have been

read beforehand) Type: pyexiv2.metadata.ImageMetadata instance

exif – whether to copy the EXIF metadata, default True iptc – whether to copy the IPTC metadata, default True xmp – whether to copy the XMP metadata, default True comment – whether to copy the image comment, default True

property dimensions

A tuple containing the width and height of the image, expressed in pixels.

property exif_keys

Return the list of the keys of the available EXIF tags.

classmethod from_buffer(buffer_)[source]

Instantiate an image container from an image memoryview.

Args: buffer_ – a memoryview containing image data as bytes

property iptc_charset

An optional character set the IPTC data is encoded in.

property iptc_keys

Return the list of the keys of the available IPTC tags.

property mime_type

The mime type of the image, as a string.

property previews

List of the previews available in the image, sorted by increasing size.

read()[source]

Read the metadata embedded in the associated image.

It is necessary to call this method once before attempting to access the metadata (an exception will be raised if trying to access metadata before calling this method).

write(preserve_timestamps=False)[source]

Write the metadata back to the image.

Args: preserve_timestamps – whether to preserve the file’s original

timestamps (access time and modification time) Type: boolean

property xmp_keys

Return the list of the keys of the available XMP tags.