Skip to content

[SPARK-4398][PySpark] specialize sc.parallelize(xrange) #3264

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
mengxr wants to merge 2 commits into from
Closed

[SPARK-4398][PySpark] specialize sc.parallelize(xrange) #3264

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
cbd58e3
specialize sc.parallelize(xrange)
mengxr Nov 14, 2014
8953c41
follow davies' suggestion
mengxr Nov 14, 2014
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
25 changes: 21 additions & 4 deletions python/pyspark/context.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -289,12 +289,29 @@ def stop(self):

def parallelize(self, c, numSlices=None):
"""
Distribute a local Python collection to form an RDD.
Distribute a local Python collection to form an RDD. Using xrange
is recommended if the input represents a range for performance.

>>> sc.parallelize(range(5), 5).glom().collect()
[[0], [1], [2], [3], [4]]
>>> sc.parallelize([0, 2, 3, 4, 6], 5).glom().collect()
[[0], [2], [3], [4], [6]]
>>> sc.parallelize(xrange(0, 6, 2), 5).glom().collect()
[[], [0], [], [2], [4]]
"""
numSlices = numSlices or self.defaultParallelism
numSlices = int(numSlices) if numSlices is not None else self.defaultParallelism
if isinstance(c, xrange):
size = len(c)
if size == 0:
return self.parallelize([], numSlices)
step = c[1] - c[0] if size > 1 else 1
start0 = c[0]

Copy link
Contributor

@davies davies Nov 14, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

How about pre-calculate all the boundaries for all the partitions?

Copy link
Contributor Author

@mengxr mengxr Nov 14, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This only serializes an xrange object. If we pre-calculate the boundaries, the cost is O(p).

Copy link
Contributor

@davies davies Nov 14, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, but the size + 1 is tricky, how about this one:

start = c[0]
def getStart(split):
      return start + size * split / numSlices * step
def f(split, iterator):
      return xrange(getStart(split), getStart(split+1), step)

Copy link
Contributor Author

@mengxr mengxr Nov 14, 2014

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, this is better!

def getStart(split):
return start0 + (split * size / numSlices) * step

def f(split, iterator):
return xrange(getStart(split), getStart(split + 1), step)

return self.parallelize([], numSlices).mapPartitionsWithIndex(f)
# Calling the Java parallelize() method with an ArrayList is too slow,
# because it sends O(n) Py4J commands. As an alternative, serialized
# objects are written to a file and loaded through textFile().
Expand Down