Cookies help us deliver our services. By using our services, you agree to our use of cookies. More information

Difference between revisions of "MapReduce"

From NoSQLZoo
Jump to: navigation, search
 
(27 intermediate revisions by 2 users not shown)
Line 1: Line 1:
−
<pre class=setup>
 
−
#ENCODING
 
−
import io
 
−
import sys
 
−
sys.stdout = io.TextIOWrapper(sys.stdout.buffer, encoding='utf-16')
 
−
#MONGO
 
−
from pymongo import MongoClient
 
−
client = MongoClient()
 
−
client.progzoo.authenticate('scott','tiger')
 
−
db = client['progzoo']
 
−
#PRETTY
 
−
import pprint
 
−
pp = pprint.PrettyPrinter(indent=4)
 
−
</pre>
 
 
{{TopTenTips}}
 
{{TopTenTips}}
 
<div style="min-height:25em">
 
<div style="min-height:25em">
 
==Introducing the MapReduce function==
 
==Introducing the MapReduce function==
−
The MapReduce function is an aggregate function that consists of two functions: Map and Reduce. As the name would suggest, the map is always performed before the reduce.<br/><br/>
+
The MapReduce function is an aggregate function that consists of two functions: Map and Reduce.
−
The map function takes data and breaks it down into tuples (key/value pairs) for each element in the dataset<br/>
+
 
−
The reduce function then takes the result of the map function and simply reduces it in to a smaller set of tuples by merging all values with the same key.<br/><br/>
+
The map is always performed before the reduce.
−
Map is used to deal with [https://en.wikipedia.org/wiki/Embarrassingly_parallel "embarassingly parallel problems"] where a task can be broken down into subtasks that can then be ran simultaneously without affecting each other. Instead of just processing elements one by one, all elements can all be dealt with at the same time in parallel. This allows for massively reduced processing times as well as large scalability across multiple servers, making it an attractive solution to handling [https://en.wikipedia.org/wiki/Big_data Big Data].<br/><br/>
 
−
We will be using Javascript here through the use of <code>Code</code>. Also note that the Mongo shell uses the camelCase <code>mapReduce()</code> whereas PyMongo uses the Python naming scheme <code>map_reduce()</code><br/><br/>
 
−
For this example MapReduce takes the form:<br/>
 
−
<pre>db.<collection>.map_reduce(
 
−
    map=<function>,
 
−
    reduce=<function>,
 
−
    out=<collection>
 
−
)
 
−
</pre>
 
−
</div>
 
−
<div class=q data-lang="py3">This example returns the population of all the continents.<br/>
 
−
In the <code>map</code> stage, <code>emit(k,v)</code> selects the fields to to turn into tuples, where <b>k</b> is the key and <b>v</b> is the value. The key will be the continents and the values will be the populations.<br/>
 
−
The <code>reduce</code> will sum all the values associated with each key.
 
−
Finally the <code>out</code> specifies that the output is to be <code>inline</code> rather than a collection, allowing it to printed it to screen.
 
−
<pre class=def>
 
−
from bson.code import Code
 
−
pp.pprint(
 
−
    db.world.map_reduce(
 
−
        map=Code("function(){emit(this.continent, this.population)}"),
 
−
        reduce=Code("""function(key, values){
 
−
                          return Array.sum(values)
 
−
                      }
 
−
                    """),
 
−
        out={"inline":1},
 
−
    )
 
−
)
 
−
</pre>
 
−
</div>
 
  
−
<div class=q data-lang="py3"><code>query</code> can be used to filter the <b>input</b> documents to map.<br/>
+
The map function examines every document in the collection and emits '''(key,value)''' pairs.
−
<p class="strong">Find the GDP for each continent, but only include data from countries that start with the letter A or B.</p>
 
−
<pre class=def>
 
−
from bson.code import Code
 
−
temp = db.world.map_reduce(
 
−
        query={"name": {"$regex":"^(A|B)"}},
 
−
        map=Code("function(){emit(this.continent, this.gdp)}"),
 
−
        reduce=Code("""function(key, values){
 
−
                          return Array.sum(values)
 
−
                      }
 
−
                  """),
 
−
        out={"inline":1},
 
−
)
 
  
−
pp.pprint(
+
The map function takes no input however the current document can be accessed as '''this'''
−
  temp["results"]
 
−
)
 
−
</pre>
 
−
<div class="ans">
 
−
from bson.code import Code;temp = db.world.map_reduce(query={"name": {"$regex":"^(A|B)"}},map=Code("function(){emit(this.continent, this.gdp)}"),reduce=Code("function(key, values){return Array.sum(values)}"),out={"inline":1});import operator;pp.pprint(temp["results"])
 
−
</div>
 
−
</div>
 
  
 +
The reduce function has two inputs, for every distinct key emitted by map the reduce function is called with a list of the corresponding values.
  
−
<div class=q data-lang="py3"><code>scope</code> takes in a <b>document</b>:<code>{}</code> and lets you create global variables.<br/>
+
==Population of each continent==
−
It's syntax is: <code>scope={}</code>.<br/>
+
<div class=q data-lang="mongo">
−
<p class="strong">Using <code>scope</code>, list all the countries with a higher population than Mexico.</p>
+
Here the map function emits the continent and the population for each country.
−
<pre class=def>
 
−
mexico_data = db.world.find_one({"name":"Mexico"})
 
−
pp.pprint(mexico_data)
 
  
−
from bson.code import Code
+
The reduce function uses the JavaScript function <code>Array.sum</code> to add the populations.
−
temp = db.world.map_reduce(
+
<pre class="def"><nowiki>
−
        scope = {"MEXICO":mexico_data},
+
db.world.mapReduce(
−
        map = Code("""function(){
+
  function () {emit(this.continent, this.population);},  
−
                        if (this.population > MEXICO.population) emit(this.name, this.population)
+
  function (k, v) { return Array.sum(v); },
−
                      }
+
  {out: {inline: 1}}
−
                  """),
+
);</nowiki></pre>
−
        reduce=Code("function(key, values){return values}"),
 
−
        out={"inline":1},
 
−
)
 
−
pp.pprint(
 
−
  temp["results"]
 
−
)
 
−
</pre>
 
−
<div class="ans">
 
−
mexico_data = db.world.find_one({"name":"Mexico"}); pp.pprint(mexico_data);  from bson.code import Code; temp = db.world.map_reduce( scope={"MEXICO":mexico_data}, map=Code("function(){if (this.population > MEXICO.population) emit(this.name, this.population)}"), reduce=Code("function(key, values){return values}"), out={"inline":1});pp.pprint(temp['results'])
 
 
</div>
 
</div>
 +
 +
==Number of countries in each continent==
 +
<div class=q data-lang="mongo">
 +
Instead of sending populations you can send a list one 1s to the reduce function.
 +
 +
The reduce function will now create a count of the number of countries in each continent.
 +
<pre class="def"><nowiki>
 +
db.world.mapReduce(
 +
  function () {emit(this.continent, 1);},
 +
  function (k, v) { return Array.sum(v); },
 +
  {out: {inline: 1}}
 +
);</nowiki></pre>
 
</div>
 
</div>
  
−
<div class=q data-lang="py3"><code>sort</code> and <code>limit</code><br/>
+
==Count only some countries==
−
Sort allows us to sort the <b>input</b> documents that are passed to <b>map</b><br/>Limit is self explanatory and also applies to the <b>input</b> documents that are passed to <b>map</b>
+
<div class=q data-lang="mongo">
−
<p class="strong">Get the five countries with the highest GDPs</p>
+
The map function does not need to emit once for every entry.
−
<pre class=def>
 
−
from bson.code import Code
 
−
temp = db.world.map_reduce(
 
−
        query={"gdp":{"$ne":None}},
 
−
        sort={"gdp":-1},
 
−
        limit=5,
 
−
        map=Code("function(){emit(this.name, this.gdp)}"),
 
−
        reduce=Code("function(key, values){return values}"),
 
−
        out={"inline":1},
 
−
)
 
  
−
pp.pprint(
+
In this example we are only counting the countries that have a large population.
−
  temp["results"]
+
<pre class="def"><nowiki>
−
)
+
db.world.mapReduce(
−
</pre>
+
  function () {
−
<div class="ans">
+
    if (this.population > 100000000)
−
from bson.code import Code; temp = db.world.map_reduce( query={"gdp":{"$ne":None}}, sort={"gdp":-1}, limit=5, map=Code("function(){emit(this.name, this.gdp)}"), reduce=Code("function(key, values){return values}"), out={"inline":1}, );pp.pprint(temp["results"])
+
    {
 +
      emit(this.continent, 1);
 +
    }
 +
  },
 +
  function (k, v) { return Array.sum(v); },
 +
  {out: {"inline": 1}}
 +
);</nowiki></pre>
 
</div>
 
</div>
 +
 +
==Examine the reduce function==
 +
<div class=q data-lang="mongo">
 +
<p class="strong">Examine the reduce function.</p>
 +
 +
Here we emit the continent and the name, and in the reduce function we <code>return v.join(',')</code> to see a comma separated list of the values in the list.
 +
<pre class="def"><nowiki>
 +
db.world.mapReduce(
 +
  function () {
 +
    if (this.population > 100000000) {
 +
      emit(this.continent, this.name);
 +
    }
 +
  },
 +
  function (k, v) { return v.join(','); },
 +
  {out: {"inline": 1}}
 +
);</nowiki></pre>
 
</div>
 
</div>
  
−
<div class=q data-lang="py3"><code>finalize</code> is an optional additional step that allows you to modify the data produce by <code>reduce</code><br/>
+
==Reduce to a single value==
−
<p class="strong">Show the top 15 countries by population, then show their population as a percentage of Mexico's population.</p>
+
<div class=q data-lang="mongo">
−
<pre class=def>
+
If you emit the same key every time you will get exactly one result from your query.
−
mexico_data = db.world.find_one({"name":"Mexico"})
 
−
 
 
−
from bson.code import Code
 
−
temp = db.world.map_reduce(
 
−
        scope = {"MEXICO":mexico_data},
 
−
        query={"population":{"$ne":None}},
 
−
        sort={"population":-1},
 
−
        limit=15,
 
−
        map=Code("function(){emit(this.name, this.population)}"),
 
−
        reduce=Code("function(key, values){return values}"),
 
−
        out={"inline":1},
 
−
        finalize=Code("""function(key, values){
 
−
                            return 100*(values/MEXICO.population)+"%"
 
−
                        }
 
−
                      """)
 
−
)
 
  
−
pp.pprint(
+
Here we emit the value 1 as the key and 1 as the value. The reduce function sums those 1s to get a count of the total number of countries.
−
  temp["results"]
+
<pre class="def"><nowiki>
−
)
+
db.world.mapReduce(
−
</pre>
+
  function () {
−
<div class="ans">
+
    emit(1, 1);
−
mexico_data = db.world.find_one({"name":"Mexico"});from bson.code import Code; temp = db.world.map_reduce( scope = {"MEXICO":mexico_data}, query={"population":{"$ne":None}}, sort={"population":-1}, limit=15, map=Code("function(){emit(this.name, this.population)}"), reduce=Code("function(key, values){return values}"), out={"inline":1}, finalize=Code("""function(key, values){return 100*(values/MEXICO.population)+"%"} """) );pp.pprint(temp["results"] );
+
  },
−
</div>
+
  function (k, v) { return Array.sum(v); },
 +
  {out: {"inline": 1}}
 +
);</nowiki></pre>
 
</div>
 
</div>
  
−
<div class=q data-lang="py3">Rounding can also be performed by using JavaScript.<br/>
+
==Emit a name==
−
<p class="strong">Show the top 15 countries by population, then show their population as a whole number percentage of Mexico's population.</p>
+
<div class=q data-lang="mongo">
−
<pre class=def>
+
You can use the list given in the reduce function.
−
mexico_data = db.world.find_one({"name":"Mexico"})
 
  
−
from bson.code import Code
+
Here we emit the key '''this.continent''' and the value '''this.name'''.
−
temp = db.world.map_reduce(
+
The reduce function returns the first element of the collected list.
−
        scope = {"MEXICO":mexico_data},
+
<pre class="def"><nowiki>
−
        query={"population":{"$ne":None}},
+
db.world.mapReduce(
−
        sort={"population":-1},
+
  function () {
−
        limit=15,
+
    emit(this.continent, this.name);
−
        map=Code("function(){emit(this.name, this.population)}"),
+
  },
−
        reduce=Code("function(key, values){return values}"),
+
  function (k, v) { return v[0]; },
−
        out={"inline":1},
+
  {out: {"inline": 1}}
−
        finalize=Code("""function(key, values){
+
);</nowiki></pre>
−
                            return Math.round(100*(values/MEXICO.population))+"%"
 
−
                        }
 
−
                      """)
 
−
)
 
−
 
 
−
pp.pprint(
 
−
  temp["results"]
 
−
)
 
−
</pre>
 
−
<div class="ans">
 
−
mexico_data = db.world.find_one({"name":"Mexico"});from bson.code import Code;temp=db.world.map_reduce(scope ={"MEXICO":mexico_data},query={"population":{"$ne":None}},sort={"population":-1},limit=15,map=Code("function(){emit(this.name,this.population)}"),reduce=Code("function(key, values){return values}"), out={"inline":1},finalize=Code("function(key,values){return Math.round(100*(values/MEXICO.population))+'%'}"));pp.pprint(temp["results"])
 
−
</div>
 
 
</div>
 
</div>

Latest revision as of 08:47, 26 June 2018

Top Tips

Introducing the MapReduce function

The MapReduce function is an aggregate function that consists of two functions: Map and Reduce.

The map is always performed before the reduce.

The map function examines every document in the collection and emits (key,value) pairs.

The map function takes no input however the current document can be accessed as this

The reduce function has two inputs, for every distinct key emitted by map the reduce function is called with a list of the corresponding values.

Population of each continent

Here the map function emits the continent and the population for each country.

The reduce function uses the JavaScript function Array.sum to add the populations. 

db.world.mapReduce(
  function () {emit(this.continent, this.population);}, 
  function (k, v) { return Array.sum(v); },
  {out: {inline: 1}}
);

Number of countries in each continent

Instead of sending populations you can send a list one 1s to the reduce function.

The reduce function will now create a count of the number of countries in each continent. 

db.world.mapReduce(
  function () {emit(this.continent, 1);}, 
  function (k, v) { return Array.sum(v); },
  {out: {inline: 1}}
);

Count only some countries

The map function does not need to emit once for every entry.

In this example we are only counting the countries that have a large population. 

db.world.mapReduce(
  function () {
    if (this.population > 100000000)
    {
      emit(this.continent, 1);
    }
  },
  function (k, v) { return Array.sum(v); },
  {out: {"inline": 1}}
);

Examine the reduce function

Examine the reduce function.

Here we emit the continent and the name, and in the reduce function we return v.join(',') to see a comma separated list of the values in the list. 

db.world.mapReduce(
  function () {
    if (this.population > 100000000) {
      emit(this.continent, this.name);
    }
  },
  function (k, v) { return v.join(','); },
  {out: {"inline": 1}}
);

Reduce to a single value

If you emit the same key every time you will get exactly one result from your query.

Here we emit the value 1 as the key and 1 as the value. The reduce function sums those 1s to get a count of the total number of countries. 

db.world.mapReduce(
  function () {
    emit(1, 1);
  },
  function (k, v) { return Array.sum(v); },
  {out: {"inline": 1}}
);

Emit a name

You can use the list given in the reduce function.

Here we emit the key this.continent and the value this.name. The reduce function returns the first element of the collected list. 

db.world.mapReduce(
  function () {
    emit(this.continent, this.name);
  },
  function (k, v) { return v[0]; },
  {out: {"inline": 1}}
);
Retrieved from ‘http://nosqlzoo.net/mw/index.php?title=MapReduce&oldid=1219’