Currently, DRYML's compute context system relies on developer 'hints' to prepare an appropriate compute context. This is fraught with problems namely that its currently impossible to get more specific in the hints than which frameworks are needed. This requires user intervention to specify appropriate compute environments through context requirements when compute methods are launched (either in the main process or a subprocess).
One idea for a solution: Dynamic code analysis at runtime finding context_check calls. The developer can insert context_check calls as a way to signpost what resources are required. Then a code analysis can find and collate these requirements so DRYML can produce the necessary environment.
The major problem with this is AST analysis is fraught with errors and very challenging. May require partial code execution for it to be good enough.
Currently, DRYML's compute context system relies on developer 'hints' to prepare an appropriate compute context. This is fraught with problems namely that its currently impossible to get more specific in the hints than which frameworks are needed. This requires user intervention to specify appropriate compute environments through context requirements when compute methods are launched (either in the main process or a subprocess).
One idea for a solution: Dynamic code analysis at runtime finding
context_checkcalls. The developer can insertcontext_checkcalls as a way to signpost what resources are required. Then a code analysis can find and collate these requirements so DRYML can produce the necessary environment.The major problem with this is AST analysis is fraught with errors and very challenging. May require partial code execution for it to be good enough.