binding for the libvips image processing library, API mode
This module wraps the libvips image processing library.
If you have the development headers for libvips installed and have a working C compiler, this module will use cffi API mode to try to build a libvips binary extension for your Python.
If it is unable to build a binary extension, it will use cffi ABI mode instead and only needs the libvips shared library. This takes longer to start up and is typically ~20% slower in execution. You can find out how pyvips installed with pip show pyvips.
This binding passes the vips test suite cleanly and with no leaks under python2.7 - python3.6, pypy and pypy3 on Windows, macOS and Linux.
We have formatted docs online here:
How it works
Programs that use pyvips don’t manipulate images directly, instead they create pipelines of image processing operations building on a source image. When the end of the pipe is connected to a destination, the whole pipeline executes at once, streaming the image in parallel from source to destination a section at a time.
Because pyvips is parallel, it’s quick, and because it doesn’t need to keep entire images in memory, it’s light. For example, the libvips speed and memory use benchmark:
Loads a large tiff image, shrinks by 10%, sharpens, and saves again. On this test pyvips is typically 3x faster than ImageMagick and needs 5x less memory.
There’s a handy blog post explaining how libvips opens files, which gives some more background.
You need the libvips shared library on your library search path, version 8.2 or later. On Linux and macOS, you can install via your package manager; on Windows you can download a pre-compiled binary from the libvips website:
Then just install this package, perhaps:
$ pip install --user pyvips
Testing your install
Try this test program:
import logging logging.basicConfig(level = logging.DEBUG) import pyvips print('test Image') image = pyvips.Image.new_from_file('/home/john/pics/k2.jpg') print('image =', image) print('image.width =', image.width)
/home/john/pics/k2.jpg with the name of a file on your machine.
If pyvips was able to build a binary module on your computer (API mode) you should see:
$ python try1.py DEBUG:pyvips:Loaded binary module _libvips ....
Otherwise, if the build failed (fallback to ABI mode), you should see:
$ python try1.py DEBUG:pyvips:Binary module load failed: No module named '_libvips' DEBUG:pyvips:Falling back to ABI mode ....
Important: if you end up installing libvips development headers after installing pyvips, you should reinstall pyvips. You should make sure pip is not reusing a cached wheel, e.g. by using pip install --no-cache-dir pyvips.
Check the output of pip show pyvips at the end to confirm you got the version you expected:
$ pip show pyvips Name: pyvips Version: 2.1.2 Summary: binding for the libvips image processing library, API mode ...
This sample program loads a JPG image, doubles the value of every green pixel, sharpens, and then writes the image back to the filesystem again:
import pyvips image = pyvips.Image.new_from_file('some-image.jpg', access='sequential') image *= [1, 2, 1] mask = pyvips.Image.new_from_array([[-1, -1, -1], [-1, 16, -1], [-1, -1, -1] ], scale=8) image = image.conv(mask, precision='integer') image.write_to_file('x.jpg')
Converting old code
To convert old code, replace the lines:
import gi gi.require_version('Vips', '8.0') from gi.repository import Vips
import pyvips Vips = pyvips
Instead of the pyvips = Vips, you can of course also swap all Vips for pyvips with eg.:
The Python binding included in libvips works, but porting and installation are more difficult than they should be.
This new binding is:
- compatible with the current Python binding (it runs the same test suite, unmodified)
- easier to install, since the stack is much smaller, and there are no issues with the overrides directory
- faster, since we implement Buffer and save some copies
- faster, since it is “thinner”. The ffi Ruby binding is about twice as fast as the gobject-introspection one, when running the test suite
- portable across CPython, PyPy and others
- more simply portable to Windows
- easy to package for pip
Local user install:
$ pip install --user -e . $ pip3 install --user -e . $ pypy -m pip --user -e .
Run all tests:
Run test suite:
$ tox test
Run a specific test:
$ pytest tests/test_conversion.py
$ tox qa
Generate HTML docs in doc/build/html:
$ cd doc; sphinx-build -bhtml . build/html
$ cd doc; \ python -c "import pyvips; pyvips.Operation.generate_sphinx_all()" > x
And copy-paste x into the obvious place in doc/vimage.rst.
Update version number:
$ vi pyvips/version.py $ vi doc/conf.py
Update pypi package:
$ python setup.py sdist $ twine upload dist/*