Single definition point for Hadoop properties

[simleo]

We have several explicit, duplicate references to Hadoop property keys, things like conf["mapreduce.job.reduces"] = "2". We should define them once and for all in some common module and use the corresponding variables everywhere else.

For instance:

NUM_REDUCES = "mapreduce.job.reduces"
[...]

[pydoop/prop.py]

And then:

conf = {}
import pydoop.prop as prop
conf[prop.NUM_REDUCES] = "2"

To simplify the process, we can use the same names as in Hadoop itself. In this case (from MRJobConfig.java):

public static final String NUM_REDUCES = "mapreduce.job.reduces";