At a high-level apt
is something maintained by your system. Specifically anything in the debian family will use apt
to manage things like drivers, compilers, things that require lower-level integration.
This means for things like numpy
and scipy
that require system-level integration with FORTRAN libraries, including the pip
dependency won't actually work.
Some python packages that are tightly-linked with the system-level dependencies maintain apt
packages that simply give you the full package all at once without having to coordinate between the two. The minus is that because Canonical's review process is pretty meticulous (as it should be) you will be getting, 9/10 a less-recent version of the library you're trying to use.
So, in short: you will often require apt
packages to enable more recent pip
installs, and while the same python dependencies may be available via apt
, these libraries are typically much older and may not have required functionality.
A common workaround is to simply use the system dependencies from one of these packages rather than the full package. You can do this by use the build-deps
flag. A common example given below:
apt-get build-dep python-scipy
pip install scipy
Which will actually give you the most up-to-date version of scipy while working within your virtualenv.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…