Mapper and reducer classes

As explained in the short tutorial, Dumbo provides the ability to use a class as mapper and/or reducer (and also as combiner, of course, but a combiner is really just a reducer with a slightly different purpose). Up until now, the main benefit of this was the possibility that it creates to use the constructor for initializations like, e.g., loading the contents of a file into memory. Version 0.20.27 introduces another reason to use classes instead of functions, however, as illustrated by the following extended version of the sampling program from my previous post:

class Mapper:
    def __init__(self):
        self.status = "Initialization started"
        from random import Random
        self.randgen = Random()
        self.cutoff = float(self.params["percentage"]) / 100
        self.samplesize = self.counters["Sample size"]
        self.status = "Initialization done"
    def __call__(self, key, value):
        if self.randgen.random() < self.cutoff:
            self.samplesize += 1
            yield key, value

if __name__ == "__main__":
    from dumbo import run

The things to note in this code are the class variables params, counters, and status, which seem to come out of nowhere. From version 0.20.27 onwards, Dumbo will instantiate a dynamically generated class that inherits from both the class supplied by the programmer and dumbo.MapRedBase, resulting in the seemingly magical addition of the following fields:

  • params: A dictionary that contains the parameters specified using -param <key>=<value> options.
  • counters: You can think of this field as a defaultdict containing dumbo.Counter objects. When a given key has no corresponding counter yet, a new counter is created using the key as name for it. You can still change the name afterwards by assigning a different string to the counter’s name field, but by using a suitable key you can put out two candles with one blow.
  • status: Strings assigned to this field will show up as status message for the task in Hadoop’s web interface.

And as if this is not enough, you can now also use the

counter += amount

syntax to increment counters, instead of the less fancy (and harder to remember)


method call.

2 Responses to Mapper and reducer classes

  1. […] that, this program works exactly as explained in the lecture. When running it, you have to use the parameter option -param doccount=<number of documents> to specify the total number of documents (which could […]

  2. […] like the original interface, the alternative one also works for mapper and reducer classes. Adapting the first example above such that a class is used for both the mapper and reducer results […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: