Installing as many dependencies as possible for pip package using Anaconda - python

When I use conda environments I often find myself using packages that are only on PyPI requiring pip to install them, but for which many of the dependencies could be installed using conda. If I just install them using pip I'll end up with an environment with several pip-managed packages.
Currently I have to start a pip install (say for torchkge), and notice that it requires for instance pandas and starts downloading it. I ctrl+C and then conda install pandas before redoing the pip install until I see that pip sees that all of the dependencies are satisfied and none need to be installed.
Is there a simple way of doing this dependency check and installing the available dependencies with conda?

You can use below command. For example if I need to check dependencies of Pandas:
pip show pandas
In result set you will find the modules required under Requires keyword.
OR
You can also use pipdeptree
$ pip install pipdeptree
then run
$ pipdeptree
and it will show you your dependencies in a tree form.
For pandas you can check:
$ pipdeptree -p pandas

Related

ModuleNotFoundError: No module named 'geopandas' [duplicate]

This error raised while installing geopandas. I've looking for its solution on the web, but none of them really explain what happened and how to solve it..
This is the full error:
Collecting geopandas
Using cached https://files.pythonhosted.org/packages/24/11/d77c157c16909bd77557d00798b05a5b6615ed60acb5900fbe6a65d35e93/geopandas-0.4.0-py2.py3-none-any.whl
Requirement already satisfied: shapely in c:\users\alvaro\anaconda3\envs\tfdeeplearning\lib\site-packages (from geopandas) (1.6.4.post2)
Requirement already satisfied: pandas in c:\users\alvaro\anaconda3\envs\tfdeeplearning\lib\site-packages (from geopandas) (0.20.3)
Collecting fiona (from geopandas)
Using cached https://files.pythonhosted.org/packages/3a/16/84960540e9fce61d767fd2f0f1d95f4c63e99ab5d8fddc308e8b51b059b8/Fiona-1.8.4.tar.gz
Complete output from command python setup.py egg_info:
A GDAL API version must be specified. Provide a path to gdal-config using a GDAL_CONFIG environment variable or use a GDAL_VERSION environment variable.
----------------------------------------
Command "python setup.py egg_info" failed with error code 1 in C:\Users\Alvaro\AppData\Local\Temp\pip-install-oxgkjg8l\fiona\
pip install wheel
pip install pipwin
pipwin install numpy
pipwin install pandas
pipwin install shapely
pipwin install gdal
pipwin install fiona
pipwin install pyproj
pipwin install six
pipwin install rtree
pipwin install geopandas
here are the source links:
http://geopandas.org/install.html#installation
https://pip.pypa.io/en/latest/user_guide/#installing-from-wheels
https://www.lfd.uci.edu/~gohlke/pythonlibs/#numpy
If you still have problems, consider uninstalling the above (pip uninstall) and reinstalling.
I solved this problem by running the following commands:
pip install pipwin
pipwin install gdal
pipwin install fiona
pip install geopandas
Works successfully on Windows.
Geospatial Data Abstraction Library (GDAL) is a library designed for vector geospatial data formats. It's a prerequisite for installing Fiona, the Python API for OGR (which doesn't really stand for anything), which is in turn a prerequisite for Geopandas. On UNIX-like systems the gdal-config script tells Fiona stuff about your particular gdal installation.
It seems that your gdal-config is not in one of the usual places on your PATH, so Fiona was unable to find it.
If you're using Anaconda, best is to remove gdal with conda remove gdal and then do a fresh conda install geopandas.
As a general rule, if you're using Conda you should never use pip to install something inside it unless you're absolutely sure conda offers no support for it. (Many package can be found on conda by specifying the right channel - -c argument.) And specifically in the case of geopandas, the maintainers recommend using conda over pip, since pip requires you to install the dependencies correctly.
I had a lot of issues myself installing geopandas, mostly showing error when downloading fiona and gdal. I did every step above and did a conda install geopandas but failed. The only thing worked for me is to install fiona and gdal wheel separately.
go to the link by Christoph: gohlke:https://www.lfd.uci.edu/~gohlke/pythonlibs/#fiona
You can search for fiona and gdal wheel files. Make sure you choose the file as per your python version, if it is 3.7 then there would be cp37.
Download the file
go to command prompt, put cd and then pip install , install GDAL wheel file, then fiona, then just do pip install geopandas.
This solution worked for me.
To install gdal, I followed the following steps:
downloaded the version that satisfies my computer (64 bit) from
https://www.lfd.uci.edu/~gohlke/pythonlibs/ . The file was GDAL-3.1.4-cp37-cp37m-win_amd64.whl
Put the file in a folder on the desktop.
From cmd, i moved to that directory and executed python -m pip install GDAL-3.1.4-cp37-cp37m-win_amd64.whl
This is followed by installing fiona the same way: python -m pip install Fiona-1.8.18-cp37-cp37m-win_amd64.whl
For shapely, i executed conda install -c conda-forge shapely
After that, i was able to install keplergl as usual: pip install keplergl
install descartes: conda install -c conda-forge descartes (or python -m pip install descartes).
In this way, i didn't have to play around with the 'Environmental Variables' as this may affect other programs
Cheers..
Installing geopandas
Geopandas has very complex multi-language dependencies, some of which need to be built with consistent compiler versions across packages. Because of this, the geopandas docs recommend installing using conda in a new environment using conda-forge only. Here are some general best practices to keep in mind:
conda is the recommended installation method. You can install geopandas from pip or source, but it's going to be a bumpy ride and it's not recommended. If you're installing conda for the first time, I recommend you start with miniconda (or better yet miniforge, a conda-forge-first miniconda variant), not anaconda, to keep your base env lean.
When using conda, you should not mix and match conda channels.
When installing geopandas, try creating a fresh environment rather than installing into your base environment. If you have anaconda installed, it comes with a large number of packages from the "defaults" channel installed in your base environment. I recommend deleting anaconda and installing miniconda, then installing into a new environment.
Try to create a new environment with everything you plan to use all at once rather than iteratively modifying the environment. In other words, if you want to use geopandas with scikit_learn, folium, and rasterio, install them together with a single conda create command
As a last resort, delete your conda installation and re-install miniconda. Desperate times call for desperate measures, and this usually resolves gnarly installation nightmares.
To create a fresh conda environment in which you install all necessary dependencies at the same time, using the conda-forge channel:
conda create -n my-geopandas-env -c conda-forge geopandas [all other packages you need]
For example, I might set up an environment with something along the lines of...
conda create -n my-geopandas-env -c conda-forge python=3.9 \
ipython ipykernel geopandas scipy seaborn fiona matplotlib cartopy
Bundling your installations into a single environment creation step like this reduces the chance of packages falling out of sync. To speed this process up, you could first install mamba or mambaforge, a faster drop-in replacement for conda, into your base environment and then run the above commands with mamba instead of conda.
Generally, it's best to avoid installing much of anything in your base environment (cross-environment system utilities like mamba are some of the few exceptions). If you already have a complex base environment (maybe you started with anaconda rather than miniconda) this may be the time to delete your entire conda installation and start from scratch (I know that's terrifying... sorry! but it'll save you heartache in the future). mamba is great for speeding this process up.
Connecting your editor to the conda environment
Once you have installed all of the packages you need, activate your environment with conda activate my-geopandas-env. See the conda guide to managing environments for more info.
Jupyter/ipython
Some editors/IDEs including jupyter require additional packages - jupyter requires that ipython and ipykernel be installed in order to load the environment within the notebook or editor - that's why I included ipykernel in my list above. See the ipykernel docs for more info.
Other IDES
To link this environment to an IDE such as VSCODE, spider, etc., find the location of this python version with conda run -n my-geopandas-env which python then point your editor to this python executable. Check the docs of your specific editor to get more targeted info about how to set up a conda environment for use with your editor:
Spider: FAQ on using an existing environment and Spider wiki guide to working with packages and environments
VSCode: Using python environments in vscode
PyCharm: Configure a conda virtual environment
I don't have conda installed, then using just pip I followed these steps:
Download GDAL and Fiona wheels directly on:
GDAL: https://www.lfd.uci.edu/~gohlke/pythonlibs/#gdal
FIONA: https://www.lfd.uci.edu/~gohlke/pythonlibs/#fiona
Then:
pip install <gdal.whl>
pip install <fiona.whl>
In my case I did pip install GDAL-3.4.1-cp38-cp38-win_amd64.whl and Fiona-1.8.21-cp38-cp38-win_amd64.whl. Where cp38 stands for python 3.8.
After that you are able to install geopandas with pip as well.
pip install geo pandas
For me, the only solution was to install the ready binaries from here
https://www.lfd.uci.edu/~gohlke/pythonlibs/#gdal
Then just install locally
pip install GDAL-3.1.4-cp38-cp38-win_amd64.whl
One way in which I could install geopandas was through the Anaconda Navigator. Get into the environment and install the package 'geopandas'. After that I could import the geopandas package in spyder
I will add
!pip install descartes
to #JDOaktown list.
I started with pip install geopandas and got the error, but later tried with conda install --channel conda-forge geopandas and the error disappeared.
Successfully installed in RHEL 7.8.
It automatically downloaded the required packages. This might be helpful
Installing collected packages: certifi, pyproj, shapely, attrs, click, click-plugins, munch, cligj, fiona, geopandas
Successfully installed attrs-20.3.0 certifi-2020.11.8 click-7.1.2 click-plugins-1.1.1 cligj-0.7.0 fiona-1.8.17 geopandas-0.8.1 munch-2.5.0 pyproj-3.0.0.post1 shapely-1.7.1
If you want to install GDAL, Geopandas, Shapely, Fiona etc in a windows Virtual Environment download .whl files for all of them and first install GDAL using
pip install gdal-.whl
Following this command edit the activate.bat file in you venv\Scripts folder and add
GDAL_CONFIG = \venv\Lib\site-packages\osgeo
Then you can install rest using pip install
I started off with a clean environment gdal_test in Conda environments, but made the mistake of using the old activate gdal_test instead of conda activate gdal_test. This made Conda Environment resolving take forever, which is why I resolved to other methods at first.
Takeaway: let conda handle it, with a proper new environment.
I ran into this problem not with anaconda/windows, but with python:3.6 Docker image. Google search always led me to this question, so I think I will share how I resolve my issue in case others also end up here.
Based on here, you need to install system relevant packages in the Dockerfile before running pip install geopandas or pip install requirements.txt:
RUN apt-get update && apt-get install -y --no-install-recommends \
build-essential \
libatlas-base-dev \
libgdal-dev \
gfortran
The following worked on macOS:
brew install gdal --HEAD
Verify the installation by running gdal-config --version
Following that pip installation as normal worked without a problem.

Create install package of current pip packages (similar to pip freeze)

I'm needing to create an install package which contains all of the pip packages currently installed. Generally I think of this as using pip freeze > requirements.txt. However since I'm needing the actual install package I then need to do pip download -r requirements.txt.
It seems strange to have to download the packages if they are already on my computer opposed to being able to package them up. This also creates an install .whl for each package opposed to a single one. (This is not a show-stopper)
What is the best method for generating an install package of the current environment for pip packages?

pip install from a file only if needed

I have a packages file (dependencies.conf) for pip including a bunch of packages that my app needs:
argparse==1.2.1
Cython==0.20.2
...
In my build process, I download all packages using:
pip install --download=build/modules -r conf/dependencies.conf
Then in the deployment process, I want to install these files only if the installed version is different than what I need and in the correct order (dependencies)
I'm currently using the following:
for f in modules/*; do pip install -I $f; done
But this is wrong since it doesn't validate the version (-I is there in order to downgrade packages if needed) and it doesn't handle the right order of dependencies.
Is there a simple method to do that? (I'm basically trying to update the packages in machines that don't have internet connection)
Get the version using PIP, using the following command
eg. pip freeze | grep Jinja2
Jinja2==2.6
as explained in the following link Find which version of package is installed with pip
then compare this with the version, and run pip install with the appropriate version if necessary

Pip freeze vs. pip list

Why does pip list generate a more comprehensive list than pip freeze?
$ pip list
feedparser (5.1.3)
pip (1.4.1)
setuptools (1.1.5)
wsgiref (0.1.2)
$ pip freeze
feedparser==5.1.3
wsgiref==0.1.2
Pip's documentation states:
 
 
freeze
Output installed packages in requirements format.
list
List installed packages.
What is a "requirements format"?
One may generate a requirements.txt via:
$ pip freeze > requirements.txt
A user can use this requirements.txt file to install all the dependencies. For instance:
$ pip install -r requirements.txt
The packages need to be in a specific format for pip to understand, such as:
# requirements.txt
feedparser==5.1.3
wsgiref==0.1.2
django==1.4.2
...
That is the "requirements format".
Here, django==1.4.2 implies install django version 1.4.2 (even though the latest is 1.6.x).
If you do not specify ==1.4.2, the latest version available would be installed.
You can read more in "Virtualenv and pip Basics",
and the official "Requirements File Format" documentation.
To answer the second part of this question, the two packages shown in pip list but not pip freeze are setuptools (which is easy_install) and pip itself.
It looks like pip freeze just doesn't list packages that pip itself depends on. You may use the --all flag to show also those packages.
From the documentation:
--all
Do not skip these packages in the output: pip, setuptools, distribute, wheel
The main difference is that the output of pip freeze can be dumped into a requirements.txt file and used later to re-construct the "frozen" environment.
In other words you can run:
pip freeze > frozen-requirements.txt on one machine and then later on a different machine or on a clean environment you can do:
pip install -r frozen-requirements.txt
and you'll get the an identical environment with the exact same dependencies installed as you had in the original environment where you generated the frozen-requirements.txt.
Look at the pip documentation, which describes the functionality of both as:
pip list
List installed packages, including editables.
pip freeze
Output installed packages in requirements format.
So there are two differences:
Output format, freeze gives us the standard requirement format that may be used later with pip install -r to install requirements from.
Output content, pip list include editables which pip freeze does not.
pip list shows ALL installed packages.
pip freeze shows packages YOU installed via pip (or pipenv if using that tool) command in a requirements format.
Remark below that setuptools, pip, wheel are installed when pipenv shell creates my virtual envelope. These packages were NOT installed by me using pip:
test1 % pipenv shell
Creating a virtualenv for this project…
Pipfile: /Users/terrence/Development/Python/Projects/test1/Pipfile
Using /usr/local/Cellar/pipenv/2018.11.26_3/libexec/bin/python3.8 (3.8.1) to create virtualenv…
⠹ Creating virtual environment...
<SNIP>
Installing setuptools, pip, wheel...
done.
✔ Successfully created virtual environment!
<SNIP>
Now review & compare the output of the respective commands where I've only installed cool-lib and sampleproject (of which peppercorn is a dependency):
test1 % pip freeze <== Packages I'VE installed w/ pip
-e git+https://github.com/gdamjan/hello-world-python-package.git#10<snip>71#egg=cool_lib
peppercorn==0.6
sampleproject==1.3.1
test1 % pip list <== All packages, incl. ones I've NOT installed w/ pip
Package Version Location
------------- ------- --------------------------------------------------------------------------
cool-lib 0.1 /Users/terrence/.local/share/virtualenvs/test1-y2Zgz1D2/src/cool-lib <== Installed w/ `pip` command
peppercorn 0.6 <== Dependency of "sampleproject"
pip 20.0.2
sampleproject 1.3.1 <== Installed w/ `pip` command
setuptools 45.1.0
wheel 0.34.2
My preferred method of generating a requirements file is:
pip list --format=freeze > requirements.txt
This method keeps just the package names and package versions without potentially linking to local file paths which 'pip freeze' alone will sometimes give me. Local file paths in a requirements file make your codebase harder to use for other users and some developers don't know how to fix this so I prefer this method for ease of adoptability.
pip list
List installed packages: show ALL installed packages that even pip installed implictly
pip freeze
List installed packages: - list of packages that are installed using pip command
pip freeze has --all flag to show all the packages.
Other difference is the output it renders, that you can check by running the commands.
For those looking for a solution. If you accidentally made pip requirements with pip list instead of pip freeze, and want to convert into pip freeze format. I wrote this R script to do so.
library(tidyverse)
pip_list = read_lines("requirements.txt")
pip_freeze = pip_list %>%
str_replace_all(" \\(", "==") %>%
str_replace_all("\\)$", "")
pip_freeze %>% write_lines("requirements.txt")

Bypass confirmation prompt for pip uninstall

I'm trying to uninstall all django packages in my superuser environment to ensure that all my webapp dependencies are installed to my virtualenv.
sudo su
sudo pip freeze | grep -E '^django-' | xargs pip -q uninstall
But pip wants to confirm every package uninstall, and there doesn't seem to be a -y option for pip. Is there a better way to uninstall a batch of python modules? Is rm -rf .../site-packages/ a proper way to go? Is there an easy_install alternative?
Alternatively, would it be better to force pip to install all dependencies to the virtualenv rather than relying on the system python modules to meet those dependencies, e.g. pip --upgrade install, but forcing even equally old versions to be installed to override any system modules. I tried activating my virtualenv and then pip install --upgrade -r requirements.txt and that does seem to install the dependencies, even those existing in my system path, but I can't be sure if that's because my system modules were old. And man pip doesn't seem to guarantee this behavior (i.e. installing the same version of a package that already exists in the system site-packages).
starting with pip version 7.1.2 you can run pip uninstall -y <python package(s)>
pip uninstall -y package1 package2 package3
or from file
pip uninstall -y -r requirements.txt
Pip does NOT include a --yes option (as of pip version 1.3.1).
WORKAROUND: pipe yes to it!
$ sudo ls # enter pw so not prompted again
$ /usr/bin/yes | sudo pip uninstall pymongo
If you want to uninstall every package from requirements.txt,
pip uninstall -y -r requirements.txt
on www.saturncloud.io, Jupiter notebooks one can use like this:
!yes | pip uninstall tensorflow
!yes | pip uninstall gast
!yes | pip uninstall tensorflow-probability
Alternatively, would it be better to force pip to install all dependencies to the virtualenv rather than relying on the system python modules to meet those dependencies,
Yes. Don't mess too much with the inbuilt system installed packages. Many of the system packages, particularly in OS X (even the debian and the derived varieties) depend too much on them.
pip --upgrade install, but forcing even equally old versions to be installed to override any system modules.
It should not be a big deal if there are a few more packages installed within the venv that are already there in the system package, particularly if they are of different version. Thats the whole point of virtualenv.
I tried activating my virtualenv and then pip install --upgrade -r requirements.txt and that does seem to install the dependencies, even those existing in my system path, but I can't be sure if that's because my system modules were old. And man pip doesn't seem to guarantee this behavior (i.e. installing the same version of a package that already exists in the system site-packages).
No, it doesn't install the packages already there in the main installation unless you have used the --no-site-packages flag to create it, or the required and present versions are different..
Lakshman Prasad was right, pip --upgrade and/or virtualenv --no-site-packages is the way to go. Uninstalling the system-wide python modules is bad.
The --upgrade option to pip does install required modules in the virtual env, even if they already exist in the system environment, and even if the required version or latest available version is the same as the system version.
pip --upgrade install
And, using the --no-site-packages option when creating the virtual environment ensures that missing dependencies can't possibly be masked by the presence of missing modules in the system path. This helps expose problems during migration of a module from one package to another, e.g. pinax.apps.groups -> django-groups, especially when the problem is with load templatetags statements in django which search all available modules for templatetags directories and the tag definitions within.
pip install -U xxxx
can bypass confirm

Categories

Resources