pyspark.RDD.saveAsTextFile#

RDD.saveAsTextFile(path, compressionCodecClass=None)[source]#

Save this RDD as a text file, using string representations of elements.

Added in version 0.7.0.

Parameters:

pathstr: path to text file
compressionCodecClassstr, optional: fully qualified classname of the compression codec class i.e. “org.apache.hadoop.io.compress.GzipCodec” (None by default)

See also

SparkContext.textFile()
SparkContext.wholeTextFiles()

Examples

>>> import os
>>> import tempfile
>>> from fileinput import input
>>> from glob import glob
>>> with tempfile.TemporaryDirectory(prefix="saveAsTextFile1") as d1:
...     path1 = os.path.join(d1, "text_file1")
...
...     # Write a temporary text file
...     sc.parallelize(range(10)).saveAsTextFile(path1)
...
...     # Load text file as an RDD
...     ''.join(sorted(input(glob(path1 + "/part-0000*"))))
'0\n1\n2\n3\n4\n5\n6\n7\n8\n9\n'

Empty lines are tolerated when saving to text files.

>>> with tempfile.TemporaryDirectory(prefix="saveAsTextFile2") as d2:
...     path2 = os.path.join(d2, "text2_file2")
...
...     # Write another temporary text file
...     sc.parallelize(['', 'foo', '', 'bar', '']).saveAsTextFile(path2)
...
...     # Load text file as an RDD
...     ''.join(sorted(input(glob(path2 + "/part-0000*"))))
'\n\n\nbar\nfoo\n'

Using compressionCodecClass

>>> from fileinput import input, hook_compressed
>>> with tempfile.TemporaryDirectory(prefix="saveAsTextFile3") as d3:
...     path3 = os.path.join(d3, "text3")
...     codec = "org.apache.hadoop.io.compress.GzipCodec"
...
...     # Write another temporary text file with specified codec
...     sc.parallelize(['foo', 'bar']).saveAsTextFile(path3, codec)
...
...     # Load text file as an RDD
...     result = sorted(input(glob(path3 + "/part*.gz"), openhook=hook_compressed))
...     ''.join([r.decode('utf-8') if isinstance(r, bytes) else r for r in result])
'bar\nfoo\n'