rtpsw commented on code in PR #14682:
URL: https://github.com/apache/arrow/pull/14682#discussion_r1041929947
##########
python/pyarrow/_compute.pyx:
##########
@@ -2629,14 +2718,98 @@ def register_scalar_function(func, function_name,
function_doc, in_types,
21
]
"""
+ return register_scalar_like_function(GetRegisterScalarFunction(),
+ func, function_name, function_doc,
in_types,
+ out_type, func_registry)
+
+
+def register_tabular_function(func, function_name, function_doc, in_types,
out_type,
+ func_registry=None):
+ """
+ Register a user-defined tabular function.
+
+ A tabular function is one accepting a context argument of type
+ ScalarUdfContext and returning a generator of struct arrays.
+ The in_types argument must be empty and the out_type argument
+ specifies a schema. Each struct array must have field types
+ correspoding to the schema.
+
+ Parameters
+ ----------
+ func : callable
+ A callable implementing the user-defined function.
+ The only argument is the context argument of type
+ ScalarUdfContext. It must return a callable that
+ returns on each invocation a StructArray matching
+ the out_type, where an empty array indicates end.
+ function_name : str
+ Name of the function. This name must be globally unique.
+ function_doc : dict
+ A dictionary object with keys "summary" (str),
+ and "description" (str).
+ in_types : Dict[str, DataType]
+ Must be an empty dictionary.
+ out_type : DataType
+ Output type of the function.
+ func_registry : FunctionRegistry
+ Optional function registry to use instead of the default global one.
+ """
+ return register_scalar_like_function(GetRegisterTabularFunction(),
+ func, function_name, function_doc,
in_types,
+ out_type, func_registry)
+
+
+def register_scalar_like_function(register_func, func, function_name,
function_doc, in_types,
+ out_type, func_registry=None):
+ """
+ Register a user-defined scalar-like function.
+
+ A scalar-like function is a callable accepting a first
+ context argument of type ScalarUdfContext as well as
+ possibly additional Arrow arguments, and returning a
+ an Arrow result appropriate for the kind of function.
+ A scalar function and a tabular function are examples
+ for scalar-like functions.
+ This function is normally not called directly but via
+ register_scalar_function or register_tabular_function.
Review Comment:
I made it internal.
> They could create a new kind of UDX but is there any point if there is no
node that consumes those types of functions?
Note that a new kind of UDX could introduce a different way of configuration
(e.g., using a callable returning a SQL script) while still outputting in a
standard way (e.g., tabular data), so existing nodes could consume its output.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
