CSC352 Homework 4 Solution

--D. Thiebaut 20:24, 25 April 2010 (UTC)

---

Problem #1: Timeline

Several observations were in order:

• The cluster only has 5 task-tracker nodes
• The default for Hadoop is two map tasks per task-tracker per job.
• This explains the maximum of 10 map tasks running first, then stopping, and then closing down, then 6 map tasks taking over, with an overlap of 4 map tasks.
• 4 files of approximately 180 MB comprise the input. Hadoop splits input files into splits of 64 MB.
• 180 MB will correspond to 3 or 4 splits, depending on the exact size of the file. This means between 12 to 16 splits for the 4 files.

Because we see that 16 (12+4) map tasks run altogether, we deduce that we had 16 splits total, requiring 16 tasks. 12 ran first, then were taken down while 4 more were started.

Problem #2

Below is a Map/Reduce pair of Python streaming programs for a job similar to what was required. Instead of reporting the unique categories, the program return the categories and a list of Ids of wiki pages in which they appear.

Examples of output

category:(10th century in ireland)	[10130001]category:(1976 films)	[10000001]
category:(argentine films)	[10000001]
category:(economy of côte d'ivoire)	[10160001]
category:(estuaries in england)	[100001,200002]
category:(hospitals in north carolina)	[1010001]
category:(lists of hospitals in the united states)	[1010001]

Mapper

hadoop@hadoop1:~/352/dft/hw4$ cat mapper_cat.py 
#!/usr/bin/env python
# mapper_cat.py
# D. Thiebaut
# 
# parses a wiki page that is in xml format
# and included in <xml> ... </xml> tags.
# Example:
#
#<xml>
#<title>Breydon Water</title>
#<id>100001</id>
#<contributors>
#<contrib>
#<ip>89.240.43.102</ip>
#
#<length>3119</length></contrib>
#</contributors>
#<categories>
#<cat>Estuaries in England</cat>
#<cat>Ramsar sites in England</cat>
#<cat>Royal Society for the Protection of Birds reserves in England</cat>
#<cat>Norfolk Broads</cat>
#<cat>Sites of Special Scientific Interest in Norfolk</cat>
#</categories>
#<pagelinks>
#<page>Arthur Henry Patterson</page>
#<page>Arthur Ransome</page>
#<page>Avocets</page>
#<page>Bewick's Swan</page>
#</pagelinks>
#<text>
#Image. Breydon-north.jpg thumb 250px ... Broads . Category. Sites of Special Scientific Interest in Norfolk . . . .
#</text>
#</xml>
#
# the program outputs the categories
# 
# cat 10000001.xml | ./mapper.py 
#  
#
# 1976 films 10000001
# Argentine films 10000001
# films 4 10000001
# category 3 10000001
# argentine 3 10000001
# 1976 3 10000001
# spanish 2 10000001
# language 2 10000001
# spanishlanguage 1 10000001
# que 1 10000001
# list 1 10000001
#

import sys


def grab( tag, xml ):
    """grabs the text between <tag> and </tag> in the xml text"""
    try:
        index1 = xml.find( '<%s>' % tag )
        index2 = xml.find( '</%s>' % tag, index1 )
        return xml[ index1+len( '<%s>' % tag ): index2 ]
    except:
        return ""

def grabAll( tag, xml ):
    """grabs all the strings between <tag> and </tag> in xml, and 
    returns them in a list"""
    index2 = 0
    list   = []
    while True:
        index1 = xml.find( '<%s>' % tag, index2 )
        if index1==-1: break
        index2 = xml.find( '</%s>' % tag, index1 )
        if index2==-1: break
        list.append( xml[ index1+len( '<%s>' % tag): index2 ] )
    return list
    
def processInput( debug=False ):
    #--- accumulate input lines from stdin ---
    xmlText = ""
    for line in sys.stdin:
        xmlText += line.lower()
    
    #--- grag the Id, and the categories ---
    id           = grab( "id", xmlText )
    categoryList = grabAll( "cat", xmlText )
    return ( id, categoryList )


def main( debug=False ):
    #--- get page Id, categories (if user wants them), and text ---
    id, categoryList  = processInput( debug )

    for cat in categoryList:
        print "%s\t%s" % ( cat, id )

    #--- debugging information ---
    if debug:
        print "-"*60
        print "DEBUG"
        print "-"*60
        print "id = ", id
        for cat in categoryList:
            print "cat:", cat
        print "-"*60
    
main( False )

Reducer

#!/usr/bin/env python
import sys
 
#--- get ready to read category, Id pairs ---
lastCategory = None
listOfIds = [] 

#--- input comes from STDIN ---
for line in sys.stdin:
    #--- remove leading and trailing whitespace ---
    line = line.strip()
    category, Id = line.split('\t', 1)

    #--- accumulate Ids for the same category ---
    if category==lastCategory:
        listOfIds.append( Id )
    else:
        if lastCategory is not None:
            print "category:(%s)\t[%s]" % ( lastCategory, ','.join( listOfIds ) )
        lastCategory = category
        listOfIds = [ Id ]


#--- write last category processed to stdout ---
if len( listOfIds )!= 0:
    print 'category:(%s)\t[%s]' % ( lastCategory, ','.join( listOfIds ) )