Skip to main content

Simple script to download .whl packages from

Project description

Gohlke Grabber

NOTICE: On 12 June 2022, this message appeared on the site Christoph Gohlke maintains: "Updated on 12 June 2022 at 20:19 UTC. Funding for the Laboratory for Fluorescence Dynamics has ceased. This service will be discontinued before July 2022." However, later that month this notice disappeared. The status of the site is unclear - this means that this tool may stop working as well; there's nothing I can do about that within reason, so I will likely remove this project if Gohlke can no longer maintain their site.

Simple script to download .whl packages from the pre-built Python packages at

Christoph Gohlke maintains 32-bit and 64-bit binaries for many popular scientific Python packages. These can save you some trouble in cases where getting the package from PyPI (using pip install package_name) causes pip to try and build underlying C or C++ code. This can of course be made to work on Windows, but requires the installation and configuration of a C/++ compiler and libraries - both of which come standard with a Linux installation, but not with Windows.

So, if you have issues installing a package, you trust Gohlke's build, and you want something easy that helps automate the download, grab a copy of and call it like shown below or in

Of course, once you have a wheel (a file with the .whl extension), you can install it using:

pip install path\to\saved\location\name.whl

Please don't bother Christoph Gohlke or their organisation if there are issues with this tool. If it breaks, that's my fault and you should bother me with it, or ideally propose how to fix it. They just provide a valuable service at no cost and merely deserve credit.


pip install gohlkegrabber


Dependencies that will be installed :


Getting Started


After installing, to get a recent copy of gdal:

from gohlkegrabber import GohlkeGrabber
gg = GohlkeGrabber()
gg.retrieve('c:/temp', 'gdal')

Or, directly from the command line:

ggrab c:\temp gdal

Note that ggrab takes the same arguments as the .retrieve() method, except that positional arguments come after named arguments, as this is the convention on OS CLIs. For example:

ggrab -v 1.18 --platform win32 .\bin numpy

The CLI command ggrab also takes an additional argument --cache if you want to specify a cached index file to use, for example:

ggrab --cache c:\temp\cache.html . numpy

If you run ggrab from the command line, you can also pass --bare or -x

pip install gohklegrabber
for /f "tokens=*" %i in ('ggrab --bare c:\temp numpy') do set ggrab_last_package=%i
pip install %ggrab_last_package%

Or in a batch file:

@echo off
pip install gohklegrabber
for /f "tokens=*" %%i in ('ggrab --bare c:\temp numpy') do set ggrab_last_package=%%i
pip install %ggrab_last_package%

In greater detail

When you create a GohlkeGrabber, it automatically downloads the index from the website (or reads a cached copy) and figures out all the packages on offer. Of course, this requires an active connection to the web.

You can list the available packages:


Note that .packages is a dict - of course you can just use the dictionary directly and the data therein yourself as well. For example, this is what the start of the numpy entry looks like:

  'numpy-1.16.5+mkl-cp27-cp27m-win32.whl': {
    'link': '',
    'version': '1.16.5+mkl',
    'build': None,
    'python': '2.7',
    'abi': 'cp27m',
    'platform': 'win32'
  'numpy-1.16.5+mkl-cp27-cp27m-win_amd64.whl': ...

To download the latest version (default) of numpy, for Windows 64-bit (default), and for the most recent version of Python (default) for which it is available, you would call:

fn, metadata = gg.retrieve(output_folder, 'numpy')

fn will be the filename of the wheel that was downloaded. metadata will be a dictionary with the metadata for the downloaded wheel. Both will be None if no package could be downloaded that matched the request.

An example of what the metadata would look like:

  'link': '',
  'version': '1.17.4+mkl',
  'build': None,
  'python': '3.8',
  'abi': 'cp38',
  'platform': 'win_amd64'

Note that this is just the appropriate entry from the .packages dict.

To get a copy for a specific Python version (e.g. 2.7), Windows version (e.g. 32-bit) and package version (e.g. '<1.17'), you can provide extra parameters to the call in no particular order:

fn, metadata = gg.retrieve(output_folder, 'numpy', python='2.7', platform='win32', version='<1.17')

Any file downloaded will be stored in the output_folder.

If the file already exists, it won't be downloaded again, unless you pass overwrite=True to the .retrieve() call.

If you create the GohlkeGrabber with a cached parameter, it will save the downloaded web page to that location, or load that file instead of downloading it again, if it already exists.

gg = GohlkeGrabber(cached='work/cache.html')


This project is licensed under the MIT license. See LICENSE.txt.

Change log


  • remove notice, as the notice on the source website has been taken down.
  • fix a partial match problem with a starts_with typo
  • adjust to change on website where identifiers are now no longer prefixed with _; detect similar change


  • add a notice with regards to the service used being discontinued - this tool may need to be discontinued as well.


  • support gzipped index html


  • adjusted to change on website where identifiers are now prefixed with _
  • added simple matching for partial identifiers


  • fixed behaviour to download current Python version of package unless 'last' is specified for '--python'

0.3.6 / 0.3.7

  • Changed version import for, as it was causing dependency problems


  • Downgraded Python requirement to 3.6


  • Improved message for missing packages


  • 'Bare' mode added to capture written wheel
  • Project structure cleanup (script folder, version location)


  • Versioning issues resolved
  • Documentation fix
  • Short command line switches


  • Added command line tool. Added 'User-Agent' to file retrieve as well as index.


  • Flipped default for python parameter, favouring the current Python over the most recent


  • added a user agent header field, as the site no longer serves a basic Python client


  • open release, version conflict

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

gohlkegrabber-0.3.12.tar.gz (11.2 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page