[jira] [Created] (SPARK-24698) In Pyspark's ML, an Identifiable's UID has 20 random characters rather than the 12 mentioned in the documentation.

["Thomas Dunne (JIRA)" <ji...@apache.org>]

Thomas Dunne created SPARK-24698:
------------------------------------

             Summary: In Pyspark's ML, an Identifiable's UID has 20 random characters rather than the 12 mentioned in the documentation.
                 Key: SPARK-24698
                 URL: https://issues.apache.org/jira/browse/SPARK-24698
             Project: Spark
          Issue Type: Bug
          Components: ML
    Affects Versions: 2.3.1
            Reporter: Thomas Dunne


Hi.

In pyspark, an Identifiable object has a random ID assigned to help distinguish instances from each other. This ID is made by concatenating the name of the class with part of a Python's built-in UUID.

The docstring of the method (__randomUID()_) that generates this ID says that 12 random characters are used from the Python UUID, but the code actually skips the first 12 characters. The hex representation of the UUID is 32 characters, so the last 20 characters are used.

Code can be found [here|https://github.com/apache/spark/blob/master/python/pyspark/ml/util.py#L66], and also copied here for your viewing pleasure:
{code}
@classmethod
def _randomUID(cls):
    """
    Generate a unique unicode id for the object. The default implementation
    concatenates the class name, "_", and 12 random hex chars.
    """
    return unicode(cls.__name__ + "_" + uuid.uuid4().hex[12:])
{code}



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org