If you are already set up to use conda, then it's as simple as:
Add the conda-forge channel:
> conda config --add channels conda-forge
Set the channel priority to "strict":
> conda config --set channel_priority strict
Create an environment for PyGNOME with all requirements:
If you only need to run PyGNOME:
> conda create -n gnome python=3.10 --file conda_requirements.txt
If you need to build, etc PyGNOME:
> conda create -n gnome python=3.10 --file conda_requirements.txt --file conda_requirements_build.txt --file conda_requirements_test.txt
Activate the gnome environment:
> conda activate gnome
Build the gnome package:
> cd py_gnome > python -m pip install ./
or:
> pip install --no-build-isolation -e ./
to get an "editable" (develop) version
You now should be good to go, but to make sure:
Run the tests:
> cd tests/unit_tests > pytest --runslow
NOTE: The "runslow" tests requiring downloading data for the tests -- you can eliminate that flag to get most of the tests to run without that delay.
Anaconda is a Python distribution that has most of the difficult-to-build packages that PyGNOME needs already built in. Thus it's a nice target for running GNOME on your own system. "conda" is the packaging manager used to manage the system.
PyGNOME can "in theory" be used with any Python distribution, and if you are familiar with complex python packaging, then you might be able to make it work. But you will need to find or build a number of dependent packages that are not easy to manage. Additionally, for compiling our C/C++ components, we have moved to a build system called scikit-build-core. And it will be quite difficult to replicate the steps that are built into this package. In short, we don't recommend it.
Anaconda provides a fairly complete python system for computational programming -- it is a large install, but comes with a lot of nice stuff pre-packaged that all works together.
miniconda is a much smaller install -- it provides only Python and the conda package management system. You can install miniconda, and then install only the packages you need to run PyGNOME.
miniforge is similar to miniconda, but designed to work primarily with the conda-forge packages. As everything required for PyGNOME is available via conda-forge, and many are not available anywhere else, miniforge is the easiest way to get PyGNOME up and running. You can install miniforge, and then install only the packages you need to run PyGNOME.
Either of these should work fine with PyGNOME, as long as you create an independent environment for it.
NOTES:
PyGNOME requires Python version 3.9 or greater (version 3.10 is currently used operationally -- 3.11 and 3.12 should work, but it not well tested)
Anaconda (and miniconda and miniforge) can be installed in either single-user or multi-user mode:
https://docs.continuum.io/anaconda/install
We (and Anaconda) recommend single-user mode (Select an install for “Just Me”) -- that way, administrator privileges are not required for either initial installation or maintaining the system.
You need the Windows 64 bit Python 3 version. Installing with the defaults works fine. You probably want to let it set the PATH for you -- that's a pain to do by hand.
Anaconda provides a 64 bit version -- this should work well with PyGNOME. Either the graphical installer or the command line one is fine -- use the graphical one if you are not all that comfortable with the *nix command line.
Everything PyGNOME needs is available for either Intel or the "M" series of processors (also called arm64 or Apple Silicon). Choose a miniconda/miniforge that matches the processor in your computer.
The Linux 64bit-python3.10 is the one to use.
NOTE: We do not support 32 bit on any platform.
conda is the package manager
that Anaconda is built on. So when working with Anaconda, you use the conda
package manager for installing conda packages. pip can also be used
within conda, but it's best to use use conda for as much as you can.
As a rule, if you need a new package, you should try to conda install it, and then, if there is no conda package available, you can pip install it.
We have made sure that every package you need for PyGNOME is available for conda.
Conda-Forge is a community project that supplies a huge number of packages for the conda package manager. We have tried to assure that everything you need to run PyGNOME is available via the conda-forge channel.
Install: Anaconda or alternatively: Miniconda or Miniforge
Once you have Anaconda, Miniconda, or Miniforge installed, the rest of the instructions should be the same.
Once you have a conda system installed, you should start by getting everything up to date, as sometimes packages have been updated since the installer was built.
First, update the conda package manager itself. Enter the following on the command-line:
> conda update conda
anaconda.org is a web service for hosting
conda packages for download. The way this is done is through anaconda
"channels", which can be thought of simply as places on anaconda.org
where collections of packages are bundled together by the people hosting them.
The "conda-forge" project:
conda-forge is a community project to build and make available a wide variety of packages for conda -- it should support everything that PyGNOME needs.
In order to find packages available on conda-forge, it should be added to your conda channel configuration:
> conda config --add channels conda-forge
Note
MiniForge comes pre-configured to work with conda-forge, this step is not necessary if you are using the miniforge distribution.
When you add a channel to conda, it puts it at the top of the list. So now when you install a package, conda will first look in conda-forge, and then in the default channel. This order should work well for PyGNOME.
You can see what channels you have with:
> conda config --get channels
It should return something like this:
--add channels 'defaults' # lowest priority --add channels 'conda-forge' # highest priority
In that order -- the order is important
You also need to set the channel priority to "strict":
> conda config --set channel_priority strict
This will assure that you will get packages from conda-forge, even if there are newer ones available in the defaults channel.
The conda system supports isolated "environments" that can be used to maintain different versions of various packages for different projects. For more information see:
http://conda.pydata.org/docs/using/envs.html
NOTE: We highly recommend that you use a conda environment for GNOME.
If you are only going to use Python / conda for PyGNOME, then you could use the base environment. However, there are a number of packages that PyGNOME needs to be at specific versions, so it is best to keep it separate from any other work you are doing.
NOTE: You can do these steps with the Anaconda Navigator GUI if you have that installed
Create an environment for PyGNOME:
> conda create -n gnome python=3.10 --file conda_requirements.txt --file conda_requirements_build.txt --file conda_requirements_test.txt
This will create an environment called "gnome" with Python itself and everything that it needs to be built, run, and tested -- it will be quite a bit, so it may take a while.
To use that environment, you activate it with:
> conda activate gnome
and when you are done, you can deactivate it with:
> conda deactivate
After activating the environment, you can proceed with these instructions, and all the packages PyGNOME needs will be installed into that environment and kept separate from your main Anaconda install.
You will need to activate the environment any time you want to work with PyGNOME in the future
PyGNOME is not currently available as a conda package, as it is under active development, and many users will need access to the source code.
Once you have a conda environment set up, you can compile and install PyGNOME.
You will need the files from the PyGNOME sources. If you have not downloaded it yet, it is available here:
https://github.com/NOAA-ORR-ERD/PyGNOME
You can either download a zip file of all the sources and unpack it, or you can "clone" the git repository. Either choice is valid.
Unless you want to contribute to the project, you should use the "main" branch.
zip and tar archives of the PyGNOME source code can be found here:
https://github.com/NOAA-ORR-ERD/PyGNOME/releases
This will get you the entire source archive of a given release, which is a fine way to work with PyGNOME. However if, in the future, you want to use any new changes that have been made to the code, you will need to re-download the new release in its entirety.
If you clone the repository, you will be able to update the code with the latest version with a simple command (git pull). This will download only the files that have changed and requires no archive extraction, so it will not only be a faster operation, but we think you will find it is also more convenient.
First you will need a Git client. On Linux, it should be available from your package manager using one of the following commands:
> apt_get install git # Debian & Linux Mint or > yum install git # CentOS & Red Hat Enterprise Linux
NOTE: There are a few other Linux package managers out there. Look at this `exhaustive list <https://en.wikipedia.org/wiki/List_of_software_package_management_systems#Linux>`__ to find the one your Linux distribution uses
On OS-X, Git comes with the XCode command line tools:
http://osxdaily.com/2014/02/12/install-command-line-tools-mac-os-x/
On Windows, the "official" Git for Windows installer is a good bet:
https://git-for-windows.github.io/
Once you have the client, it's as easy as:
> git clone https://github.com/NOAA-ORR-ERD/PyGNOME.git
This will create a ./pygnome directory with all the code in it.
- git branches:
- git supports a number of different "branches" or versions of the code. You will most likley want to use the "main" branch (the default) unless you specifically want to experiment with a new feature.
If you have not already created an environment in which to run PyGNOME, follow the instructions above.
To use the gnome environment you created, it needs to be activated with:
> conda activate gnome
If you don't want to create an environment (or already have one), you can install what PyGNOME needs into an existing environment:
> cd ./pygnome # or wherever you put the PyGNOME project > conda install --file conda_requirements.txt --file conda_requirements_build.txt --file conda_requirements_test.txt
NOTE: PyGNOME has a lot of specific dependencies -- it can be very hard for conda to resolve them with an large installed package base. If you have trouble, it's easiest to make a new environment just for PyGNOME.
This should install all the packages required by PyGNOME.
(make sure you are in the correct conda environment, and you have the conda-forge channel enabled)
If you get an error about a particular package not being able to be installed, then conda will not install ANY of the packages in the file. We try hard to make sure everything is available on conda-forge. If however, a package of that particular version is missing, here are some things you can try.
Edit the conda_requirements.txt file and comment out the offending package by putting a "#" at the start of the line:
... scipy>=0.17 py_gd>=0.1.5 # libgd>=2.2.2 gsw>=3.0.3 ...
That will disable that particular package, and hopefully everything else will install.
You can then try installing the offending package without a version specification:
> conda install libgd
And it may work for you.
If you want to use PyGNOME with "real oil", rather than inert particles,
you will need NOAA's adios_db package from the ADIOS Oil Database Project:
https://github.com/NOAA-ORR-ERD/adios_oil_database
This will allow you to use the JSON oil data format downloadable from NOAA's ADIOS Oil Database web app:
The adios_db package is available on conda-forge, and should have been
installed by the process above. If not, it can be installed with
> conda install adios_db
However, the adios_db package is also under active development along with PyGNOME, so if you are working with the develop branch of PyGNOME, you may need the latest version of adios_db as well. In which case, you are best off downloading the sources from GitHub and installing it from source -- similar to PyGNOME.
The latest releases (of the same branch) of each should be compatible.
To clone the repository:
> git clone https://github.com/NOAA-ORR-ERD/adios_oil_database.git
To install its dependencies:
> cd ./oil_database/adios_db > conda install --file conda_requirements.txt
Installing the package:
> pip install ./
(or pip install --no-build-isolation -e ./ to get an "editable" version)
If you run the PyGNOME tests after having installed adios_db, it will run
a few additional tests that require the adios_db. It should not need
independent testing.
But if you want to test it directly, you will need additional requirements:
> conda install --file conda_requirements_test.txt
And then you can run the tests:
> pytest --pyargs adios_db
To build PyGNOME, you will need a C/C++ compiler. The procedure for getting the compiler tools varies with the platform you are on.
The system compiler for OS-X is XCode. It can be installed from the App Store.
Apple has changed the XCode install process a number of times over the years. Rather than providing out-of-date information, we will simply state that you need the "Xcode Command Line Tools" -- look for Apple's documentation for how to install those.
Once the command line tools are installed, you should be able to build PyGNOME as described below.
For compiling python extensions on Windows with python3 it is best to use the Microsoft the Visual Studio 2019 (or later) Build Tools. They should be available here:
https://visualstudio.microsoft.com/downloads/
The free "Community" version should be fine.
Once Visual Studio is installed, a number of
"Visual Studio Developer Command Prompt" applications will be made available
on the Windows toolbar. Scikit-build-core claims that it can intelligently
configure its environment to correctly build your package, but to be on the
safe side, you will want to open up the one with a name that looks something
similar to "x64 Native Tools Command Prompt (for VS 20XX)" in order to
build PyGNOME -- this is to make sure the compiler is setup for building
x64 targets.
- Warning:
- On some locked down systems, such as those at NOAA, the
standard way to use the MS compiler will not work for a user that does not
have administration privileges. If you get errors about not being able to
run the
vcvarsall.batscript, then the compiler must be run as an administrator. If you have access to the NOAA/ORR GitLab server, a work around is supplied here: Building Python extensions on Windows. If you have this issue and are not from NOAA, ask for help on the Python forum or as an issue in the PyGNOME gitHub project.
Linux uses the GNU gcc compiler. If it is not already installed on your system, use your system package manager to get it.
- apt for Debian based distros (Ubuntu, Mint, Kali, ...)
- yum for CentOS
- ...
At this point you should have all the necessary third-party tools in place, and you can build the PyGNOME package itself. But How you build the package depends on how you plan to use it.
Most people will likely want to simply use the package for building and running simulations. For this, run the following:
> cd <your_pygnome_git_repo>/py_gnome > python -m pip install ./
Just keep in mind that any updates to the project will need to be rebuilt and re-installed in order for changes to take effect.
NOTE: You may have noticed that we run the pip module inside python instead of running the `pip` executable directly. We have noticed on some platforms (Windows) that conda virtual environments, when activated, sometimes don't properly update the $Path environmental variable, causing pip to be run from the base conda environment instead of the current one. The result is that PyGNOME gets installed there instead of our current conda environment. Running pip as a module ensures we are referencing the correct environment for installation
If you are planning to develop or debug the PyGNOME source code itself, then you may want to perform a "editable" install. A "editable" install puts a links intot he source code, rather than copying it into the Python install, so that changes in the python code are immediately available in your python environment without re-installing.
For this, run the following:
> cd <your_pygnome_git_repo>/py_gnome > python -m pip install --no-build-isolation --editable ./
If you would like or need to uninstall the package, run the following:
> python -m pip uninstall gnome
We have an extensive set of unit and functional tests to make sure that PyGNOME is working properly.
To run the tests:
> cd <your_pygnome_git_repo>/py_gnome/tests/unit_tests > pytest
and if those pass, you can run:
> pytest --runslow
which will run some more tests, some of which take a while to run.
Note that the tests will try to auto-download some data files. If you are not on the internet, this will fail. And of course if you have a slow connection, these files could take a while to download. Once the tests are run once, the downloaded files are cached for future test runs.
We do our best to keep all tests passing on release versions of the package. But sometimes tests will fail due to the setup of the machine they are being run on -- package versions, etc. So the first thing to do is to make sure you have installed the dependencies as specified.
But gnome is large package -- hardly anyone is going to use all of it.
So while we'd like all tests to pass, a given test failure may not be an issue
for any given use case. It's a bit hard to know whether a given test failure
will affect your use case, but if you look at the name of the tests that fail,
you might get a hint. For example, if any of the tests fail under
test_weathering, and you are not doing any oil weathering modeling,
you don't need to worry about it.
In any case, you can try to run your use case, and see what happens.
Please report any unresolved test failures as an Issue on the gitHub project.
There are a number of scripts in the scripts directory.
In example_scripts you will find examples of using the gnome package
for various tasks.
In testing_scripts you will find scripts that have been developed to
test various features of the model. There are many more of these, so do look
to see if they have what you need. But they are generally written in a
less compact way as they are designed to exercise particular features.
You should be able to run these scripts in the same way as any Python script (with an IDE such as Spyder or PyCharm, or at the command line).
To run a script on the command line:
> cd py_gnome/scripts/example_scripts
If you are using a conda environment:
> conda activate gnome
Run the script:
> python example_script.py
Each of the scripts exercises different features of PyGNOME -- they are hopefully well commented to see how they work.
In the testing_scripts dir, there is a run_all.py script that will
run all the testing scripts -- primarily to make sure they all can still run
as we update the model.
For further documentation of PyGNOME, see: