commandRunner 0.3.5

Allows object oriented running of code/commands

commandRunner is yet another package created to handle running commands, scripts or programs on the command line. The simplest class lets you run anything locally on your machine. Later classes are targetted at Analytics and data processing platforms such as Grid Engine and HADOOP. The class attempts to run commands in a moderately thread safe way by requiring that you provide with sufficient information that it can build a uniquely labelled temp directory for all input and output files. This means that this can play nicely with things like Celery workers.

Release 0.3

This release supports running commands on localhost and DRMAA compliant grid engine installs (ogs, soge and univa). It also uses interpolation for the commands with the same syntax as python templates

Future

In the future we’ll provide classes to run commands over RServe, Hadoop, Octave, and SAS Server.

Usage

This is the basic usage:

from commandRunner import *

r = localRunner(tmp_id="ID_STRING", tmp_path=,/tmp/", out_glob=['file'],
                command="ls /tmp > $OUTPUT", input_data={DATA_DICT}
                input_string="test.file", output_string="out.file")
r.prepare()
exit_status = r.run_cmd(success_params=[0])
r.tidy()
print(r.output_data)

__init__ initalises all the class variables needed and performs the command string interpolation.

r.prepare() builds a temporary directory and makes any input file which is needed. In this instance “ID_STRING”, and a path where temporary files can be placed are used to create a tempdir called /tmp/ID_STRING/.

Next it takes input_data. This is a dict of {Filename:Data_string} values. Iterating over, it writes the data to each named file in the tempdir. So the following dict:

{ "test.file" : "THIS IS MY STRING OF DATA"}

would result in a file with the path /tmp/ID_STRING/test.file

out_glob is an array of file suffixes which we want to gather up when the command completes.

Not that only tmp_id, tmp_path and command are required. Omitting input_data or out_glob assumes that there are respectively no input files to write or output files to gather.

The line r.run_cmd(success_params=[0]) runs the command string provided.

The command string supports some limited interpolation. First anything labeled $INPUT or $OUTPUT will be replaced with the input_string and output_string. $OPTIONS will interpolate a dictionary of switches and values. $FLAGS will interpolate an array of flags.

In the given example “ls /tmp > $OUTPUT” will become “ls /tmp > out.file”. Additionally We can also provide an array of unix exits statuses we consider to be successful exists, default is [0]. Any command will be run so this is potentially very dangerous. The exit status of the command is returned.

Once complete if ou_globs have been provided and the files were output then the contents of those files can be found in the dict r.output_data. which has the same form as the input_data dict:

{ “output.file” : “THIS IS MY PROCESSED DATA”}

r.tidy() cleans up deleting any input and output files and the temporary working directory. Any data in the output file is read in to r.output_data

Grid Engine Quirks

geRunner uses python DRMAA to submit jobs. A consequence of this is that $INPUT, $OUTPUT, $FLAGS and $OPTIONS are NOT supported. These are concatenated in to an array of arguments that are passed to the command by the DRMAA layer in this order:

[$INPUT, $FLAGS, $OPTIONS]

The Options dict is flattened to a key:value list. You can include or omit as many of those as you’d like. The output_string if provided gives a file where the Grid Engine thread STDOUT will be sent.

Tests

Best to run these 1 suite at a time, geRunner tests will fail if you do not have ogs installed and DRMAA_LIBRARY_PATH set

Run tests with:

python setup.py test -s tests/test_commandRunner.py python setup.py test -s tests/test_localRunner.py python setup.py test -s tests/test_geRunner.py

TODO

1. Implement rserveRunner for running commands in r

2. Implement hadoopRunner for running command on Hadoop

3. Implement sasRunner for a SAS backend

4. Implement octaveRunner for Octave backend

5. matlab? mathematica?