You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Aug 29, 2023. It is now read-only.
@forman: I like the diff op, but I still disagree with the arithmetic op. We should really move to an op, where users can specify a normal expression which we either
eval without globals but with a safe locals environment, see cate.util.safe_eval()
and/or use numexpr for boosting performance.
The text was updated successfully, but these errors were encountered:
Idea: use new new_expression_op() function and register it as operation compute with expression as only input.
Problems:
This leads to a workflow step
with no other step inputs than expression, therefore it appears as it wouldn't have any dependencies to other steps. But actually, the expression may refer to any other step output by referencing their generated resources.
with a single return value of unknown data type, because an expression may evaluate to any Python data type.
The latter problem may be avoided by offering dedicated compute_<type> operations.
The first may be avoided by treating steps containing a compute operation (or one containing general Python expressions evaluated in the context of the workflow) as being dependent on all other steps. This way they would be re-invoked if any other step input changes.
@forman: I like the diff op, but I still disagree with the arithmetic op. We should really move to an op, where users can specify a normal expression which we either
cate.util.safe_eval()
The text was updated successfully, but these errors were encountered: