vibhatha commented on code in PR #12590:
URL: https://github.com/apache/arrow/pull/12590#discussion_r842634846
##########
python/pyarrow/_compute.pyx:
##########
@@ -2251,3 +2325,144 @@ cdef CExpression _bind(Expression filter, Schema
schema) except *:
return GetResultValue(filter.unwrap().Bind(
deref(pyarrow_unwrap_schema(schema).get())))
+
+
+cdef CFunctionDoc _make_function_doc(dict func_doc):
+ """
+ Helper function to generate the FunctionDoc
+ """
+ cdef:
+ CFunctionDoc f_doc
+ vector[c_string] c_arg_names
+ c_bool c_options_required
+
+ f_doc.summary = tobytes(func_doc["summary"])
+ f_doc.description = tobytes(func_doc["description"])
+ for arg_name in func_doc["arg_names"]:
+ c_arg_names.push_back(tobytes(arg_name))
+ f_doc.arg_names = c_arg_names
+ # UDFOptions integration:
+ # TODO: https://issues.apache.org/jira/browse/ARROW-16041
+ f_doc.options_class = tobytes("None")
+ c_options_required = False
+ f_doc.options_required = c_options_required
+ return f_doc
+
+
+def register_function(func_name, num_args, function_doc, in_types,
+ out_type, callback):
+ """
+ Register a user-defined-function.
+
+ Parameters
+ ----------
+
+ func_name : str
+ function name
+ num_args : int
+ number of arguments in the function
+ function_doc : dict
+ a dictionary object with keys "summary" (str),
+ "description" (str), and "arg_names" (list of str).
+ in_types : List[InputType]
+ list of InputType objects which defines the input
+ types for the function
+ out_type : DataType
+ output type of the function
+ callback : callable
+ user defined function
+ function includes arguments equal to the number
+ of input_types defined. The return type of the
+ function is of the type defined as output_type.
+ The output should be an Array, Scalar, ChunkedArray,
+ Table, or RecordBatch based on the out_type.
+
+ Example
+ -------
+
+ >>> from pyarrow import compute as pc
+ >>> from pyarrow.compute import register_function
+ >>> from pyarrow.compute import InputType
+ >>>
+ >>> func_doc = {}
+ >>> func_doc["summary"] = "simple udf"
+ >>> func_doc["description"] = "add a constant to a scalar"
+ >>> func_doc["arg_names"] = ["x"]
+ >>>
+ >>> def add_constant(array):
+ ... return pc.call_function("add", [array, 1])
+ ...
+ >>>
+ >>> func_name = "py_add_func"
+ >>> arity = 1
+ >>> in_types = [InputType.array(pa.int64())]
+ >>> out_type = pa.int64()
+ >>> register_function(func_name, arity, func_doc,
+ ... in_types, out_type, add_constant)
+ >>>
+ >>> func = pc.get_function(func_name)
+ >>> func.name
+ 'py_add_func'
+ >>> ans = pc.call_function(func_name, [pa.array([20])])
+ >>> ans
+ <pyarrow.lib.Int64Array object at 0x10c22e700>
+ [
+ 21
+ ]
+ """
+ cdef:
+ c_string c_func_name
+ CArity c_arity
+ CFunctionDoc c_func_doc
+ CInputType in_tmp
+ vector[CInputType] c_in_types
+ PyObject* c_callback
+ shared_ptr[CDataType] c_type
+ COutputType* c_out_type
+ CScalarUdfBuilder* c_sc_builder
+ CStatus st
+ CScalarUdfOptions* c_options
+ object obj
+
+ c_func_name = tobytes(func_name)
+
+ if num_args <= 0:
+ raise ValueError("number of arguments must be >= 0")
+ if num_args == 0:
+ c_arity = CArity.Nullary()
+ elif num_args == 1:
+ c_arity = CArity.Unary()
+ elif num_args == 2:
+ c_arity = CArity.Binary()
+ elif num_args == 3:
+ c_arity = CArity.Ternary()
+ elif num_args > 3:
+ c_arity = CArity.VarArgs(num_args)
+
+ c_func_doc = _make_function_doc(function_doc)
Review Comment:
Yes, I thought about this too. I modified the code where I previously raised
custom exceptions after catching each unfilled field, but it is not nice
either. As you're suggesting we could make it optional. But I favor in function
docs to be provided by the user. May be it is better to add a single validation
for function doc and raise an exception?
--
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: github-unsubscr...@arrow.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
