Source code for mmlspark.featurize.text.PageSplitter

# Copyright (C) Microsoft Corporation. All rights reserved.
# Licensed under the MIT License. See LICENSE in project root for information.


import sys
if sys.version >= '3':
    basestring = str

from pyspark.ml.param.shared import *
from pyspark import keyword_only
from pyspark.ml.util import JavaMLReadable, JavaMLWritable
from mmlspark.core.serialize.java_params_patch import *
from pyspark.ml.wrapper import JavaTransformer, JavaEstimator, JavaModel
from pyspark.ml.common import inherit_doc
from mmlspark.core.schema.Utils import *

[docs]@inherit_doc class PageSplitter(ComplexParamsMixin, JavaMLReadable, JavaMLWritable, JavaTransformer): """ Args: boundaryRegex (str): how to split into words (default: \s) inputCol (str): The name of the input column maximumPageLength (int): the maximum number of characters to be in a page (default: 5000) minimumPageLength (int): the the minimum number of characters to have on a page in order to preserve work boundaries (default: 4500) outputCol (str): The name of the output column (default: [self.uid]_output) """ @keyword_only def __init__(self, boundaryRegex="\s", inputCol=None, maximumPageLength=5000, minimumPageLength=4500, outputCol=None): super(PageSplitter, self).__init__() self._java_obj = self._new_java_obj("com.microsoft.ml.spark.featurize.text.PageSplitter") self.boundaryRegex = Param(self, "boundaryRegex", "boundaryRegex: how to split into words (default: \s)") self._setDefault(boundaryRegex="\s") self.inputCol = Param(self, "inputCol", "inputCol: The name of the input column") self.maximumPageLength = Param(self, "maximumPageLength", "maximumPageLength: the maximum number of characters to be in a page (default: 5000)") self._setDefault(maximumPageLength=5000) self.minimumPageLength = Param(self, "minimumPageLength", "minimumPageLength: the the minimum number of characters to have on a page in order to preserve work boundaries (default: 4500)") self._setDefault(minimumPageLength=4500) self.outputCol = Param(self, "outputCol", "outputCol: The name of the output column (default: [self.uid]_output)") self._setDefault(outputCol=self.uid + "_output") if hasattr(self, "_input_kwargs"): kwargs = self._input_kwargs else: kwargs = self.__init__._input_kwargs self.setParams(**kwargs)
[docs] @keyword_only def setParams(self, boundaryRegex="\s", inputCol=None, maximumPageLength=5000, minimumPageLength=4500, outputCol=None): """ Set the (keyword only) parameters Args: boundaryRegex (str): how to split into words (default: \s) inputCol (str): The name of the input column maximumPageLength (int): the maximum number of characters to be in a page (default: 5000) minimumPageLength (int): the the minimum number of characters to have on a page in order to preserve work boundaries (default: 4500) outputCol (str): The name of the output column (default: [self.uid]_output) """ if hasattr(self, "_input_kwargs"): kwargs = self._input_kwargs else: kwargs = self.__init__._input_kwargs return self._set(**kwargs)
[docs] def setBoundaryRegex(self, value): """ Args: boundaryRegex (str): how to split into words (default: \s) """ self._set(boundaryRegex=value) return self
[docs] def getBoundaryRegex(self): """ Returns: str: how to split into words (default: \s) """ return self.getOrDefault(self.boundaryRegex)
[docs] def setInputCol(self, value): """ Args: inputCol (str): The name of the input column """ self._set(inputCol=value) return self
[docs] def getInputCol(self): """ Returns: str: The name of the input column """ return self.getOrDefault(self.inputCol)
[docs] def setMaximumPageLength(self, value): """ Args: maximumPageLength (int): the maximum number of characters to be in a page (default: 5000) """ self._set(maximumPageLength=value) return self
[docs] def getMaximumPageLength(self): """ Returns: int: the maximum number of characters to be in a page (default: 5000) """ return self.getOrDefault(self.maximumPageLength)
[docs] def setMinimumPageLength(self, value): """ Args: minimumPageLength (int): the the minimum number of characters to have on a page in order to preserve work boundaries (default: 4500) """ self._set(minimumPageLength=value) return self
[docs] def getMinimumPageLength(self): """ Returns: int: the the minimum number of characters to have on a page in order to preserve work boundaries (default: 4500) """ return self.getOrDefault(self.minimumPageLength)
[docs] def setOutputCol(self, value): """ Args: outputCol (str): The name of the output column (default: [self.uid]_output) """ self._set(outputCol=value) return self
[docs] def getOutputCol(self): """ Returns: str: The name of the output column (default: [self.uid]_output) """ return self.getOrDefault(self.outputCol)
[docs] @classmethod def read(cls): """ Returns an MLReader instance for this class. """ return JavaMMLReader(cls)
[docs] @staticmethod def getJavaPackage(): """ Returns package name String. """ return "com.microsoft.ml.spark.featurize.text.PageSplitter"
@staticmethod def _from_java(java_stage): module_name=PageSplitter.__module__ module_name=module_name.rsplit(".", 1)[0] + ".PageSplitter" return from_java(java_stage, module_name)