Installation

It is recommended to install fort-myrmidon, py-fort-myrmidon and r-fort-myrmidon via conda or mamba and to install fort-studio using snap.

fort-studio through snap

fort-studio is available on the snap store with the eponymous snap:

sudo snap install fort-studio

Warning

As for most other snaps, fort-studio uses strict confinement. It means that by default, it will only be allowed to access files on the current user’s home. However, with the FORT system, experiment tends to be quite large and stored on external hard drives. To enable fort-studio to access these files, you must grant the snap right to access external hard drives by allowing the removable-media connection:

sudo snap connect fort-studio:removable-media :removable-media

Note

For systems with auxiliary disk, it is recommended to mount the disk meant to be used with fort-studio under /media, as otherwise they won’t be accessible through the confinement.

The edge channel of this snap is reserved for continuous deployment and is automatically updated with any code pushed on the master branch.

fort-myrmidon and bindings through conda

A complete distribution of libfort-myrmidon (C++ library) py-fort-myrmidon (python bindings) and r-fort-myrmidon (R bindings) are available on the formicidae-tracker channel.

Tips on an efficient conda usage

conda can be a bit daunting to master at first. The ecosystem of tools has grown quite large, and it is difficult to get all essential information in one place. You can find some pointers on essentials items regarding fort-myrmidon.

Note

More general, but detailed information can be found in the “Introduction to Conda for (Data) Scientist”.

  1. Install miniconda instead of anaconda. The latter comes with a very large number of preinstalled packages, and most of them are not required to run fort-myrmidon. This could lead to quite long solving environment computation time, which are for some users quite irritating.

  2. conda setup: Since libfort-myrmidon depends on package from anaconda and conda-forge a careful installation as per conda-forge documentation is strongly recommended. In summary, one must perform once:

    conda config --add channels conda-forge
conda config --set channel_priority strict

  3. Use mamba to install conda package. This is a tool that is a replacement for the conda command which is an order of magnitude faster with dependencies and environment solving. fort-myrmidon packages are now built and tested with mamba, and therefore is supported.

  4. Do not ever use the base environment. This is a good practice rule to avoid running into a lot of dependency problems when you may want to be able to use different fort-myrmidon version on the same system. Let says you are running a long experiment for over a year and do not want to lose time rewriting all your script because of a major update of fort-myrmidon but you still want for a new project to start to use the latest version of fort-myrmidon. With the power of conda virtual environment, you can easily install the two versions, and choose when you run your script which version to use. However, if you install everything in the base environment, you would certainly lose the ability to do that as it will certainly lead to a lot of conflict between your base and environment where you want another fort-myrmidon version installed. Therefore, always make sure:

    1. To activate the environment you want to modify:

      conda activate my-env-name

      You can create an environment with:

      conda create -n my-env-name

    2. The right environment name is displayed in your prompt, and not base.

Including fort-studio to your project

Python

After installing the py-fort-myrmidon package in an environment, the py_fort_myrmidon module is available to python in that environment.

Note

The examples in this documentation assumes this module to be included as :

import py_fort_myrmidon as fm

R

After installing the r-fort-myrmidon package in an environment, the FortMyrmidon package is available to the R version packaged with this environment. The system packaged R will not be able to see nor to use without crashing the FortMyrmidon package.

C++

After installing the libfort-myrmidon package, libraries and header files for fort-myrmidon will be available to the targeted conda environment. The recommanded way is to use CMake to build your project and link with fort-myrmidon.

find_package(FortMyrmidon REQUIRED)
include_directories(FORT_MYRMIDON_INCLUDE_DIRS)
# fort-myrmidon requires at least the c++17 standard
set(CMAKE_CXX_STANDARD 17)

# once my-target is created
target_link_libraries(my-target ${FORT_MYRMIDON_LIBRARIES})

A pkg-config configuration file is also available, but it is not nicely integrated with conda. Its utilisation should be avoided.

Using rstudio with conda

If you are an rstudio user, you will remark that it will always use the system’s R version instead of the one you installed with conda, which means that it will never be able to find r-fort-myrmidon. Indeed, rstudio will use the first R version found in the PATH, and if you launch it from the system’s menu, it cannot find the version you installed as a user (it is not conda-aware), and defaults to the system’s one. There are two means to circumvent that:

  1. Runs rstudio from a terminal within the right environment (recommended). One can simply use the following commands to activate my-env and run rstudio within it:

    conda activate my-env-name
rstudio

  2. Installs and runs rstudio from conda. It is not recommended as it is a more heavy solution and will make your environments much more complex and time-consuming to solve.