When I run pip freeze
I see (among other expected packages) pkg-resources==0.0.0
. I have seen a few posts mentioning this package (including this one), but none explaining what it is, or why it is included in the output of pip freeze
. The main reason I am wondering is out of curiosity, but also, it seems to break things in some cases when trying to install packages with a requirements.txt
file generated with pip freeze
that includes the pkg-resources==0.0.0
line (for example when Travis CI tries to install dependencies through pip
and finds this line).
What is pkg-resources
, and is it OK to remove this line from requirements.txt
?
Update:
I have found that this line only seems to exist in the output of pip freeze
when I am in a virtualenv
. I am still not sure what it is or what it does, but I will investigate further knowing that it is likely related to virtualenv
.
According to https://github.com/pypa/pip/issues/4022, this is a bug resulting from Ubuntu providing incorrect metadata to pip. So, no there does not seem to be a good reason for this behaviour. I filed a follow-up bug with Ubuntu. https://bugs.launchpad.net/ubuntu/+source/python-pip/+bug/1635463
To backup the previous answer, it should be safe to remove that line from your requirements.txt. Here is an example Make file stanza that safely freezes your package list (drop in your Makefile and run with
make freeze
):As for the part of your question "is it OK to remove this line?":
I have the same issue here developing on an ubuntu 16.04 with that very line in the requirements. When deploying on a debian 8.5 running
"pip install -r requirements.txt"
pip complains that pkg-resources is "not found" but there is a global package installed "python-pkg-resources" so the dependency should be satisfied. Same on ubuntu: The package exists there as well.As stated here it seems to be some "implicitly installed package".
So: If you are on a Debian/Ubuntu having python-pkg-resources installed it should be safe to remove that line. I did so and everything is running fine. However since I am no expert on this you should keep in mind that this might lead to complications when deploying on another machine.