This is already the case for bash transformations that are synthesized by the cmd module, since run_transformation can accept them natively. However, all non-Python transformers in workflows are translated to low-level Python transformers. What is needed:
-
Add an (inert) language attribute to low-level transformers.
-
Improve TransformationCache.build_transformation to gain the powers of cmd/bash_transformation/prepare_bash_transformation,
perhaps integrate them into the same machinery.
-
Rip translate_bashX_transformer (maintain them as stdgraph for unbashify), and rename translate_py_transformer to translate_transformer
-
Expand the unbashify/run_transformation machinery to include (the executor part of) compiled transformers. Fold translate_compiled_transformer (maintain them as stdgraph) into translate_transformer.
-
Try to include dynamic language support (ipybridges etc.) as well.
This is already the case for bash transformations that are synthesized by the cmd module, since run_transformation can accept them natively. However, all non-Python transformers in workflows are translated to low-level Python transformers. What is needed:
Add an (inert) language attribute to low-level transformers.
Improve
TransformationCache.build_transformationto gain the powers ofcmd/bash_transformation/prepare_bash_transformation,perhaps integrate them into the same machinery.
Rip
translate_bashX_transformer(maintain them as stdgraph for unbashify), and renametranslate_py_transformertotranslate_transformerExpand the unbashify/run_transformation machinery to include (the executor part of) compiled transformers. Fold
translate_compiled_transformer(maintain them as stdgraph) intotranslate_transformer.Try to include dynamic language support (ipybridges etc.) as well.