Closed acl33 closed 3 years ago
I think we need a high-level docstring that describes the design of PathElement
or ExprWithPath
.
It would be also nice to have an example of SerializedPath
in the source code as comment / docstring. I imagine that it will be much larger than a single integer we use at the moment. What are the pros and cons of using a path?
I think we need a high-level docstring that describes the design of
PathElement
orExprWithPath
.It would be also nice to have an example of
SerializedPath
in the source code as comment / docstring. I imagine that it will be much larger than a single integer we use at the moment. What are the pros and cons of using a path?
Docstring + example, seems good (I think best as a test, to verify the doc stays correct - do we want to bring in doctest? Probably yes :), but now?).
"Pros and cons of using a path" I think @awf has argued the path will be more readable and debuggable and that the additional length does not matter.
I think we need a high-level docstring that describes the design of
PathElement
orExprWithPath
.
Let's try to do that in #923 . I've rebased this to come after that PR, as this now implements the following suggestion:
It would be also nice to have an example of
SerializedPath
in the source code as comment / docstring.
Hang on, that wasn't supposed to close - just trying to retarget this to go into master
This address AB#20021 for the evaluation pipeline
rewrite_seq_to_exprs
to rewrites.py, somewhat modelled on the RLO version.__str__
to be more consistent ("Let_rhs"
or"Call_args[0]"
) I made the latter take serialized paths, but we could rely on the caller to deserialize first? This way the test does show what some serialized paths look like.