如何让scala字符串拆分以匹配python

Question

I am using spark-shell and pyspark to do word count on one article. scala flatmap on line.split(" ") and python split() get different word counts (scala has more). I tried split(" +") and split("\\W+") in the scala code, but can not get the count to come down to the same as the python one.

Anyone knows what pattern would match python exactly?

Answer 1

Python's str.split() has some special behaviour for default separator:

runs of consecutive whitespace are regarded as a single separator, and the result will contain no empty strings at the start or end if the string has leading or trailing whitespace. Consequently, splitting an empty string or a string consisting of just whitespace with a None separator returns [] .

For example, ' 1 2 3 '.split() returns ['1', '2', '3']

The easiest way to fully match this in Scala is probably like this:

scala> """\S+""".r.findAllIn(" 1  2   3  ").toList
res0: List[String] = List(1, 2, 3)

scala> """\S+""".r.findAllIn("   ").toList
res1: List[String] = List()

scala> """\S+""".r.findAllIn("").toList
res2: List[String] = List()

Another way is to trim() the string beforehand:

scala> " 1  2   3  ".trim().split("""\s+""")
res3: Array[String] = Array(1, 2, 3)

But that doesn't have the same behaviour as Python for empty strings:

scala> "".trim().split("""\s+""")
res4: Array[String] = Array("")

In Scala split() of an empty string returns an array with one element, but in Python the result is a list with zero elements.