Open
Description
Is your feature request related to a problem? Please describe.
We had an error where the user ran chain
[n({'n': '0'}), e_forward(), n()]
And got not hits, b/c it was a type error, should have been:
[n({node_id: 0}), e_forward(), n()]
This could have warned or raised ahead of time with a clear message
Describe the solution you'd like
This query does not type check --- '0' : str
will never match g._nodes.n.dtype.name == 'int64'
So a few things:
g.chain()
: By default, we can run & raise constant predictate type mismatchesg.chain_remote()
: Return a client error (422) on errors here- Add optional arg
validation='skip'|'warn'|'error'
Describe alternatives you've considered
I was trying to think of a way we can do python-typed predicate functions, not just primitve value checking, but was not clear
We should probably also support both pandas 1 (obj = str or obj) and pandas 2 (str dtype)