def pfaValueType(self, dataType):
"""Create a PFA type schema representing the comparison values."""
if dataType.get("type", None) != "record":
raise TypeError("dataType must be a record")
dataTypeName = dataType["name"]
dataFieldNames = [x["name"] for x in dataType["fields"]]
if not set(self.dataset.names).issubset(set(dataFieldNames)):
raise TypeError("dataType must be a record with at least as many fields as the ones used to train the tree")
dataFieldTypes = []
for field in dataType["fields"]:
try:
fieldIndex = self.dataset.names.index(field["name"])
except ValueError:
pass
else:
if self.dataset.fields[fieldIndex].tpe == numbers.Real:
if field["type"] not in ("int", "long", "float", "double"):
raise TypeError("dataType field \"{0}\" must be a numeric type, since this was a numeric type in the dataset training".format(field["name"]))
dataFieldTypes.append(field["type"])
elif self.dataset.fields[fieldIndex].tpe == basestring:
if field["type"] != "string":
raise TypeError("dataType field \"{0}\" must be a string, since this was a string in the dataset training".format(field["name"]))
if self.maxSubsetSize == 1:
dataFieldTypes.append(field["type"])
else:
dataFieldTypes.append({"type": "array", "items": field["type"]})
asjson = [json.dumps(x) for x in dataFieldTypes]
astypes = ForwardDeclarationParser().parse(asjson)
return LabelData.broadestType(astypes.values())
def pfaValueType(self, dataType):
"""Create an Avro schema representing the comparison value type.
:type dataType: Pythonized JSON
:param dataType: Avro record schema of the input data
:rtype: Pythonized JSON
:return: value type (``value`` field of the PFA ``TreeNode``)
"""
if dataType.get("type", None) != "record":
raise TypeError("dataType must be a record")
dataTypeName = dataType["name"]
dataFieldNames = [x["name"] for x in dataType["fields"]]
if not set(self.dataset.names).issubset(set(dataFieldNames)):
raise TypeError(
"dataType must be a record with at least as many fields as the ones used to train the tree"
)
dataFieldTypes = []
for field in dataType["fields"]:
try:
fieldIndex = self.dataset.names.index(field["name"])
except ValueError:
pass
else:
if self.dataset.fields[fieldIndex].tpe == numbers.Real:
if field["type"] not in ("int", "long", "float", "double"):
raise TypeError(
"dataType field \"{0}\" must be a numeric type, since this was a numeric type in the dataset training"
.format(field["name"]))
dataFieldTypes.append(field["type"])
elif self.dataset.fields[fieldIndex].tpe == str:
if field["type"] != "string":
raise TypeError(
"dataType field \"{0}\" must be a string, since this was a string in the dataset training"
.format(field["name"]))
if self.maxSubsetSize == 1:
dataFieldTypes.append(field["type"])
else:
dataFieldTypes.append({
"type": "array",
"items": field["type"]
})
asjson = [json.dumps(x) for x in dataFieldTypes]
astypes = ForwardDeclarationParser().parse(asjson)
return LabelData.broadestType(astypes.values())
def broadestType(self, types):
return LabelData.broadestType(types)