Ubuntu Question posted in
The official Ubuntu documentation can be found here.

How to upgrade NumPy in Ubuntu?

My OS is Ubuntu 22.04.4 LTS. The "pip list" command tells that I have the NumPy 1.21.5. But I need a newer version of the NumPy.

I am trying to install the NumPy package whith next command:

pip install --upgrade numpy==2.1.2

But here is an error:

<...>
Run-time dependency python found: NO (tried pkgconfig, pkgconfig and sysconfig)

../meson.bould:41:12: ERROR: python dependency not found

I find that I must install -dev package. So I do that:

sudo apt install python3-dev

…but the NumPy installation error is still there.

What shall I do to upgrade the NumPy?

Tags:

4

Answers


  1. 0

    The command you’re using has a syntax error. It should be

    pip install --upgrade numpy==2.1.2

    Notice the == instead of =. This might fix it.

    Login or Signup to reply.
  2. 0
    1. numpy is a Python module, it has nothing to do with the OS.
    2. pip automatically finds and installs the latest version of the module as supported by your python version when you write pip install <package>. If you see a newer version on pypi but is not being installed by pip, it’s because your python version does not support that numpy version. Confirm if it is supported for your python version or upgrade your python version.
    3. default python command on ubuntu points to python 2 and not 3. Check if it is the case and change your config to point to python3 accordingly.
    Login or Signup to reply.
  3. 0

    Use a virtual-env first

    cd <to-your-working-directory>
python3 -m venv venv
source venv/bin/activate

    Your prompt will change accordingly, to indicate you are using the virtual environment. Any next session (either a new terminal, new terminal tab or simply a new login) only requires the source venv/bin/activate in that working directory, not the python -m venv part.

    Now install NumPy

    python3 -m pip install numpy

    and install whatever else you need.

    This will install a newer NumPy in that virtual environment; it is shielded from the system Python, so it can’t break your system.

    Now do your analysis / coding / whatever you want to do.

    Note that I have explicitly used Python 3 here. The python command (and similar the pip command) may refer to Python 2, and that won’t work with newer packages such as NumPy 2. Using python3 explicitly will ensure you don’t accidentally use Python 2.

    Usuing python3 -m pip in a virtual environment, is technically not necessary anymore(*), but it’s a good enough habit to be clear what is going on, that I’ll leave it like that.

    (*) within a virtual environment, python and pip will refer to the Python you used to create that v-venv, which was the system Python 3.

    Login or Signup to reply.
  4. 0

    What’s most likely going on here is: You are using the system version of both python and numpy. With "system version" I mean a version installed as an apt package (so via Ubuntu’s package management) as opposed to being installed as a pip package (so via Python’s default package manager).

    What is usually not a good idea (and often simply does not work, as you experienced), is trying to update a Python package with pip that has been installed with apt.

    The better idea would usually be: keep the Python environment that you want to work with and the system’s Python environment separate. This separation can happen on different levels:

    • You can use the system version of python, but might want to install your own version of numpy. This can be achieved, for example, using a virtual environment, as has been suggested in AKX’s comment. The article linked there is a good deep-dive into the subject and also might provide you with further motivation as to why to choose this approach. The answer of 9769953 describes the necessary steps of setting up a virtual environment and installing numpy there.
    • You might even want to keep your python version separate from the system version of python. This can be achieved, for example, using anaconda/miniconda/miniforge. With conda and its ecosystem, you have yet another package management tool and repository available, which is also Python-focused, but includes more than Python packages.

    There are plenty more options starting from there (obligatory xkcd cartoon, which, though being outdated, still captures the essence of the situation), and it may be hard to judge which setup will be best for your needs. For the pros and cons of virtual environments vs. conda in particular, you might also want to have a look at the answers to this related question that compares virtualenv (a superset of venv) to conda.

    Login or Signup to reply.
Please signup or login to give your own answer.

Back To Top
Search