Hi Evan and Brian! Here is a PR for adding some changes to Skipper main branch. Biggest changes are the installation process, modifying the Skipper.py to handle both running on containers or locally, and modifying the Skipper_config.py into sections.
One issue with UMICollapse is that the conda installed UMICollapse has difficulties with parsing in java virtual machine parameters. So we still have to download the repo, which leads to discrepancies on paths to umicollapse.jar file (the containerize environment has different paths, compared to users' computers). We thought about going back to use UMICOLLAPSE_DIR and let users define the path to the umicollapse.jar file. But we also want to try to provide the bash script to install UMICollapse. Therefore, let's try if it's ok to just let users to install the UMICollapse into the installation folder and then we take care from there. Sadly hard coding is involved here :( so let me know your thoughts after you review the code and I can make adjustments.
We include this note below in the protocol as well:
Create the conda environment using the skipper_env.yaml. Conda-installed UMICollapse is unable to process large datasets, therefore install UMICollapse v1.0.0 and dependencies into the installation folder for the pipeline to detect UMICollapse files, using the installation bash scripts provided. Refer to umicollapse.sh script for details.
Let me know your thoughts on these changes. Thank you!
Hi Evan and Brian! Here is a PR for adding some changes to Skipper main branch. Biggest changes are the installation process, modifying the Skipper.py to handle both running on containers or locally, and modifying the Skipper_config.py into sections.
One issue with UMICollapse is that the conda installed UMICollapse has difficulties with parsing in java virtual machine parameters. So we still have to download the repo, which leads to discrepancies on paths to umicollapse.jar file (the containerize environment has different paths, compared to users' computers). We thought about going back to use UMICOLLAPSE_DIR and let users define the path to the umicollapse.jar file. But we also want to try to provide the bash script to install UMICollapse. Therefore, let's try if it's ok to just let users to install the UMICollapse into the installation folder and then we take care from there. Sadly hard coding is involved here :( so let me know your thoughts after you review the code and I can make adjustments.
We include this note below in the protocol as well:
Create the conda environment using the skipper_env.yaml. Conda-installed UMICollapse is unable to process large datasets, therefore install UMICollapse v1.0.0 and dependencies into the installation folder for the pipeline to detect UMICollapse files, using the installation bash scripts provided. Refer to umicollapse.sh script for details.
Let me know your thoughts on these changes. Thank you!