Skip to content

Latest commit

 

History

History
144 lines (102 loc) · 4.75 KB

README.md

File metadata and controls

144 lines (102 loc) · 4.75 KB
Raw

Conda build for Soar

This repository contains recipes and instructions for building a conda package for the Soar cognitive architecture.

There are currently two recipes in this repository:

  • from-dist-conda.recipe - builds from official Soar binary distribution This currently includes VisualSoar and the TankSoar and Eaters tutorial programs. This should be changed to produce a soar package and add-on packages for VisualSoar and tutorial programs.
  • from-source.recipe - builds a local Soar source tree. This does not include VisualSoar or the tutorial programs.

(We should probably have the first recipe produce two or more packages.)

from-dist-conda.recipe

The recipe is based on the binary distribution of Soar available from the Soar releases on GitHub.

Currently this specifically is for MacOS but probably would extend fairly easily to Linux.

The recipe contains the following dependencies:

    - python 3.12
    - graphviz
    - tk 8.6.13
    - openjdk >=19.0.0

The recipe currently hard codes the version of Soar and must be updated manually for new releases.

from-source.recipe

This recipe builds Soar from a local Soar source tree.

In order to use this recipe, you must create a symbolic link in this directory named local-soar-root that points to the root of the of a local Soar source tree.

How to install

You can install into a conda environment on a supported platform using either conda or mamba:

conda install -c https://artifactory.analog.com/artifactory/garage-conda-local soar

If you have not already done so, we recommend that you configure garage-conda-local as a custom channel:

conda config --set custom_channels.garage-conda-local https://artifactory.analog.com/artifactory/

then you can install using the simple channel name:

conda install -c garage-conda-local soar

How to build

The version number must be adjusted manually in the meta.yaml file as needed. The build number should also be incremented if you do multiple builds of the same version for the same architeture and python version.

conda build <recipe-dir>

For from-source-conda.recipe you should add the --python argument to specify the desired target python version. For example:

conda build --python 3.12 from-source-conda.recipe

The resulting package can be found in the architecture-specific subdirectory of your local conda-bld directory, which by default is a subdirectory of your conda build install location, e.g. ~/miniforge3/conda-bld/osx-arm64.

You can install and test this locally by creating a test environment:

conda create -n soar-test --use-local soar # You can add =<version>

Upload to artifactory.analog.com

You must have been granted write permission to upload to the garage-conda-local repository in order to upload this or any other package.

Using garconda

If you have installed garconda, you can use garconda upload to upload the package to the Analog Artifactory garage-conda-local repository:

garconda upload -c garage-conda-local --latest 'soar-*'

You can add --dry-run to see what would be uploaded before actually uploading. Add --help to see information about more options.

Using the web UI

Alternatively, you can simply upload via the artifactory web interface:

  1. Open garage-conda-local in your browser
  2. Log in with your Analog credentials using the button in the upper right corner.
  3. Click on garage-conda-local in the left pane to see the list of subdirectories
  4. Select the directory appropriate to the architecture you built (e.g. osx-arm64)
  5. Press the Deploy button in the upper right corner
  6. Use Select File in the dialog select the file to upload from your local file system (in the conda-bld directory). Then press Deploy to complete the upload.

Testing

Once uploaded, you should see the package when you search using conda or mamba:

$ conda search -c garage-conda-local soar
Loading channels: done
# Name                       Version           Build  Channel             
soar                           9.6.3         py312_5  garage-conda-local  ```

and should be able to create an environment from the channel using the specified version, e.g.:

conda create -n soar-test -c garage-conda-local soar=9.6.3