SparkContext.hadoopFile(path: str, inputFormatClass: str, keyClass: str, valueClass: str, keyConverter: Optional[str] = None, valueConverter: Optional[str] = None, conf: Optional[Dict[str, str]] = None, batchSize: int = 0) → pyspark.rdd.RDD[Tuple[T, U]][source]

Read an ‘old’ Hadoop InputFormat with arbitrary key and value class from HDFS, a local file system (available on all nodes), or any Hadoop-supported file system URI. The mechanism is the same as for meth:SparkContext.sequenceFile.

New in version 1.1.0.

A Hadoop configuration can be passed in as a Python dict. This will be converted into a Configuration in Java.


path to Hadoop file


fully qualified classname of Hadoop InputFormat (e.g. “org.apache.hadoop.mapreduce.lib.input.TextInputFormat”)


fully qualified classname of key Writable class (e.g. “”)


fully qualified classname of value Writable class (e.g. “”)

keyConverterstr, optional

fully qualified name of a function returning key WritableConverter

valueConverterstr, optional

fully qualified name of a function returning value WritableConverter

confdict, optional

Hadoop configuration, passed in as a dict

batchSizeint, optional, default 0

The number of Python objects represented as a single Java object. (default 0, choose batchSize automatically)


RDD of tuples of key and corresponding value


>>> import os
>>> import tempfile

Set the related classes

>>> output_format_class = "org.apache.hadoop.mapred.TextOutputFormat"
>>> input_format_class = "org.apache.hadoop.mapred.TextInputFormat"
>>> key_class = ""
>>> value_class = ""
>>> with tempfile.TemporaryDirectory() as d:
...     path = os.path.join(d, "old_hadoop_file")
...     # Write a temporary Hadoop file
...     rdd = sc.parallelize([(1, ""), (1, "a"), (3, "x")])
...     rdd.saveAsHadoopFile(path, output_format_class, key_class, value_class)
...     loaded = sc.hadoopFile(path, input_format_class, key_class, value_class)
...     collected = sorted(loaded.collect())
>>> collected
[(0, '1\t'), (0, '1\ta'), (0, '3\tx')]