mrjob.util - general utility functions

Utility functions for MRJob

mrjob.util.args_for_opt_dest_subset(option_parser, args, dests=None)

For the given OptionParser and list of command line arguments args, yield values in args that correspond to option destinations in the set of strings dests. If dests is None, return args as parsed by OptionParser.

Deprecated since version 0.5.8.


Escape single quotes in a shell command string and wrap it with bash -c '<string>'.

This low-tech replacement works because we control the surrounding string and single quotes are the only character in a single-quote string that needs escaping.

Deprecated since version 0.5.8.


Alias for to_lines(). Will be removed in v0.6.0.

Deprecated since version 0.5.0.


build a command line that works in a shell.


Resolve ~ (home dir) and environment variables in path.

If path is None, return None.


return the file extension, including the .

>>> file_ext('foo.tar.gz')

Set up a null handler for the given stream, to suppress “no handlers could be found” warnings.

mrjob.util.log_to_stream(name=None, stream=None, format=None, level=None, debug=False)

Set up logging.

  • name (str) – name of the logger, or None for the root logger
  • stderr (file object) – stream to log to (default is sys.stderr)
  • format (str) – log message format (default is ‘%(message)s’)
  • level – log level to use
  • debug (bool) – quick way of setting the log level: if true, use logging.DEBUG, otherwise use logging.INFO
mrjob.util.parse_and_save_options(option_parser, args)

Return a map from option name (dest) to a list of the arguments in args that correspond to that dest.

This won’t modify option_parser.

mrjob.util.populate_option_groups_with_options(assignments, indexed_options)

Given a dictionary mapping OptionGroup and OptionParser objects to a list of strings represention option dests, populate the objects with options from indexed_options (generated by scrape_options_and_index_by_dest()) in alphabetical order by long option name. This function primarily exists to serve scrape_options_into_new_groups().

  • assignments (dict of the form {my_option_parser: ('verbose', 'help', ...), my_option_group: (...)}) – specification of which parsers/groups should get which options
  • indexed_options (dict generated by util.scrape_options_and_index_by_dest()) – options to use when populating the parsers/groups

A random 16-digit hex string.

mrjob.util.read_file(path, fileobj=None, yields_lines=True, cleanup=None)

Yields lines from a file, possibly decompressing it based on file extension.

Currently we handle compressed files with the extensions .gz and .bz2.

  • path (string) – file path. Need not be a path on the local filesystem (URIs are okay) as long as you specify fileobj too.
  • fileobj – file object to read from. Need not be seekable. If this is omitted, we open(path).
  • yields_lines – Does iterating over fileobj yield lines (like file objects are supposed to)? If not, set this to False (useful for boto.s3.Key)
  • cleanup – Optional callback to call with no arguments when EOF is reached or an exception is thrown.
mrjob.util.read_input(path, stdin=None)

Stream input the way Hadoop would.

  • Resolve globs (foo_*.gz).
  • Decompress .gz and .bz2 files.
  • If path is '-', read from stdin
  • If path is a directory, recursively read its contents.

You can redefine stdin for ease of testing. stdin can actually be any iterable that yields lines (e.g. a list).

mrjob.util.safeeval(expr, globals=None, locals=None)

Like eval, but with nearly everything in the environment blanked out, so that it’s difficult to cause mischief.

globals and locals are optional dictionaries mapping names to values for those names (just like in eval()).

mrjob.util.save_current_environment(*args, **kwds)

Context manager that saves os.environ and loads it back again after execution

mrjob.util.save_cwd(*args, **kwds)

Context manager that saves the current working directory, and chdir’s back to it after execution.


Scrapes optparse options from OptionParser and OptionGroup objects and builds a dictionary of dest_var: [option1, option2, ...]. This function primarily exists to serve scrape_options_into_new_groups().

An example return value: {'verbose': [<verbose_on_option>, <verbose_off_option>], 'files': [<file_append_option>]}

Parameters:parsers_and_groups (OptionParser or OptionGroup) – Parsers and groups to scrape option objects from
Returns:dict of the form {dest_var: [option1, option2, ...], ...}
mrjob.util.scrape_options_into_new_groups(source_groups, assignments)

Puts options from the OptionParser and OptionGroup objects in source_groups into the keys of assignments according to the values of assignments. An example:

  • source_groups (list of OptionParser and OptionGroup objects) – parsers/groups to scrape options from
  • assignments (dict with keys that are OptionParser and OptionGroup objects and values that are lists of strings) – map empty parsers/groups to lists of destination names that they should contain options for

Wrapper around shlex.split(), but convert to str if Python version < 2.7.3 when unicode support was added.


Return the given datetime.timedelta, without microseconds.

Useful for printing datetime.timedelta objects.

mrjob.util.tar_and_gzip(dir, out_path, filter=None, prefix='')

Tar and gzip the given dir to a tarball at out_path.

If we encounter symlinks, include the actual file, not the symlink.

  • dir (str) – dir to tar up
  • out_path (str) – where to write the tarball too
  • filter – if defined, a function that takes paths (relative to dir and returns True if we should keep them
  • prefix (str) – subdirectory inside the tarball to put everything into (e.g. 'mrjob')

Take in data as a sequence of bytes, and yield it, one line at a time.

Only breaks lines on \n (not \r), and does not add a trailing newline.

Optimizes for:

  • chunks bigger than lines (e.g. reading test files)
  • chunks that are lines (idempotency)
mrjob.util.unarchive(archive_path, dest)

Extract the contents of a tar or zip file at archive_path into the directory dest.

  • archive_path (str) – path to archive file
  • dest (str) – path to directory where archive will be extracted

dest will be created if it doesn’t already exist.

tar files can be gzip compressed, bzip2 compressed, or uncompressed. Files within zip files can be deflated or stored.


Yield items from item in order, skipping duplicates.

mrjob.util.which(cmd, path=None)

Like the UNIX which command: search in path for the executable named cmd. path defaults to PATH. Returns None if no such executable found.

This is basically shutil.which() (which was introduced in Python 3.3) without the mode argument. Best practice is to always specify path as a keyword argument.

mrjob.util.zip_dir(dir, out_path, filter=None, prefix='')

Compress the given dir into a zip file at out_path.

If we encounter symlinks, include the actual file, not the symlink.

  • dir (str) – dir to tar up
  • out_path (str) – where to write the tarball too
  • filter – if defined, a function that takes paths (relative to dir and returns True if we should keep them
  • prefix (str) – subdirectory inside the tarball to put everything into (e.g. 'mrjob')