The user should be able to initiate container-based execution without the need to explicitly specify singularity run / singularity exec. Instead datalad make should use information (metadata) attached to the dataset that tracks the container and its preconfigured container calls.
This would be a result of a sufficient metadata description. In such a description there would be some level that specifies something like {"use_container": true, "container_runner": "singularity"}. That would enclose any computation in the specified container, without the need to know this beforehand.
The user should be able to initiate container-based execution without the need to explicitly specify
singularity run/singularity exec. Insteaddatalad makeshould use information (metadata) attached to the dataset that tracks the container and its preconfigured container calls.This would be a result of a sufficient metadata description. In such a description there would be some level that specifies something like
{"use_container": true, "container_runner": "singularity"}. That would enclose any computation in the specified container, without the need to know this beforehand.