Define augmentaitons and their names once, and then the whole package is made aware of available transforms etc
basically:
-
avoid name clash from transform vs. config name of that transform
-
easy validation that transform exists and config exists for it
-
easy sanity validation of a config before even running anything (undefined error crash, this is just silent behavior right now)
-
crash if arguments do not exist for a transform (silenty doing nothing)
-
either with inheritance or simple tagging system define for each augmentaiton whether it is a GEO, TA, or GE. then the codebase can automatically based on this warn if the order of transforms is unusual, ...
Define augmentaitons and their names once, and then the whole package is made aware of available transforms etc
basically:
avoid name clash from transform vs. config name of that transform
easy validation that transform exists and config exists for it
easy sanity validation of a config before even running anything (undefined error crash, this is just silent behavior right now)
crash if arguments do not exist for a transform (silenty doing nothing)
either with inheritance or simple tagging system define for each augmentaiton whether it is a GEO, TA, or GE. then the codebase can automatically based on this warn if the order of transforms is unusual, ...