Skip to content
Projects
Groups
Snippets
Help
Loading...
Help
Support
Keyboard shortcuts
?
Submit feedback
Contribute to GitLab
Sign in / Register
Toggle navigation
slapos
Project overview
Project overview
Details
Activity
Releases
Repository
Repository
Files
Commits
Branches
Tags
Contributors
Graph
Compare
Issues
0
Issues
0
List
Boards
Labels
Milestones
Merge Requests
0
Merge Requests
0
CI / CD
CI / CD
Pipelines
Jobs
Schedules
Analytics
Analytics
CI / CD
Repository
Value Stream
Wiki
Wiki
Snippets
Snippets
Members
Members
Collapse sidebar
Close sidebar
Activity
Graph
Create a new issue
Jobs
Commits
Issue Boards
Open sidebar
Léo-Paul Géneau
slapos
Commits
207570a3
Commit
207570a3
authored
Feb 18, 2014
by
Marco Mariani
Browse files
Options
Browse Files
Download
Email Patches
Plain Diff
hadoop: python demo
parent
e61c686e
Changes
9
Show whitespace changes
Inline
Side-by-side
Showing
9 changed files
with
320 additions
and
4 deletions
+320
-4
software/hadoop/deploy-config.sh.in
software/hadoop/deploy-config.sh.in
+8
-0
software/hadoop/gutenberg-mapper.py.in
software/hadoop/gutenberg-mapper.py.in
+19
-0
software/hadoop/gutenberg-reducer.py.in
software/hadoop/gutenberg-reducer.py.in
+40
-0
software/hadoop/instance.cfg.in
software/hadoop/instance.cfg.in
+40
-4
software/hadoop/put-files.sh.in
software/hadoop/put-files.sh.in
+17
-0
software/hadoop/run-demo.sh.in
software/hadoop/run-demo.sh.in
+5
-0
software/hadoop/software.cfg
software/hadoop/software.cfg
+69
-0
software/hadoop/wikipedia-mapper.py.in
software/hadoop/wikipedia-mapper.py.in
+61
-0
software/hadoop/wikipedia-reducer.py.in
software/hadoop/wikipedia-reducer.py.in
+61
-0
No files found.
software/hadoop/deploy-config.sh.in
0 → 100644
View file @
207570a3
#!/bin/bash
# exit on error
set
-e
source
environment.sh
echo cp
-a
$HADOOP_PREFIX
/etc
${
buildout
:directory
}
/
software/hadoop/gutenberg-mapper.py.in
0 → 100644
View file @
207570a3
#!/usr/bin/env python
# http://www.michael-noll.com/tutorials/writing-an-hadoop-mapreduce-program-in-python/
import
sys
# input comes from STDIN (standard input)
for
line
in
sys
.
stdin
:
# remove leading and trailing whitespace
line
=
line
.
strip
()
# split the line into words
words
=
line
.
split
()
# increase counters
for
word
in
words
:
# write the results to STDOUT (standard output);
# what we output here will be the input for the
# Reduce step, i.e. the input for reducer.py
#
# tab-delimited; the trivial word count is 1
print
'%s
\
t
%s'
%
(
word
,
1
)
software/hadoop/gutenberg-reducer.py.in
0 → 100644
View file @
207570a3
#!/usr/bin/env python
# http://www.michael-noll.com/tutorials/writing-an-hadoop-mapreduce-program-in-python/
from
operator
import
itemgetter
import
sys
current_word
=
None
current_count
=
0
word
=
None
# input comes from STDIN
for
line
in
sys
.
stdin
:
# remove leading and trailing whitespace
line
=
line
.
strip
()
# parse the input we got from mapper.py
word
,
count
=
line
.
split
(
'
\
t
'
,
1
)
# convert count (currently a string) to int
try
:
count
=
int
(
count
)
except
ValueError
:
# count was not a number, so silently
# ignore/discard this line
continue
# this IF-switch only works because Hadoop sorts map output
# by key (here: word) before it is passed to the reducer
if
current_word
==
word
:
current_count
+=
count
else
:
if
current_word
:
# write result to STDOUT
print
'%s
\
t
%s'
%
(
current_word
,
current_count
)
current_count
=
count
current_word
=
word
# do not forget to output the last word if needed!
if
current_word
==
word
:
print
'%s
\
t
%s'
%
(
current_word
,
current_count
)
software/hadoop/instance.cfg.in
View file @
207570a3
...
...
@@ -3,6 +3,10 @@
parts =
deploy-config
sh-environment
put-files
mapper
reducer
run-demo
eggs-directory = ${buildout:eggs-directory}
develop-eggs-directory = ${buildout:develop-eggs-directory}
...
...
@@ -15,6 +19,7 @@ output = $${buildout:directory}/environment.sh
input = inline:
export JAVA_HOME="${java:location}"
export HADOOP_PREFIX="${hadoop:location}"
export PATH=$PATH:$HADOOP_PREFIX/bin
export HADOOP_HOME="${hadoop:location} "
export HADOOP_COMMON_HOME="${hadoop:location}"
export HADOOP_CONF_DIR="$${buildout:directory}/etc/hadoop"
...
...
@@ -25,11 +30,42 @@ input = inline:
[deploy-config]
recipe = cp.recipe.cmd
update_cmd = /bin/true
install_cmd =
cp -a ${hadoop:location}/etc $${buildout:directory}/
recipe = slapos.recipe.template
url = ${:_profile_base_location_}/deploy-config.sh.in
output = $${buildout:directory}/deploy-config.sh
# md5sum =
mode = 0755
[put-files]
recipe = slapos.recipe.template
url = ${:_profile_base_location_}/put-files.sh.in
output = $${buildout:directory}/put-files.sh
# md5sum =
mode = 0755
# http://www.michael-noll.com/tutorials/writing-an-hadoop-mapreduce-program-in-python/
[mapper]
recipe = slapos.recipe.template
url = ${:_profile_base_location_}/gutenberg-mapper.py.in
output = $${buildout:directory}/gutenberg-mapper.py
# md5sum =
mode = 0755
[reducer]
recipe = slapos.recipe.template
url = ${:_profile_base_location_}/gutenberg-reducer.py.in
output = $${buildout:directory}/gutenberg-reducer.py
# md5sum =
mode = 0755
[run-demo]
recipe = slapos.recipe.template
url = ${:_profile_base_location_}/run-demo.sh.in
output = $${buildout:directory}/run-demo.sh
# md5sum =
mode = 0755
software/hadoop/put-files.sh.in
0 → 100644
View file @
207570a3
#!/bin/bash
# exit on error
# set -e
source
environment.sh
hdfs dfs
-mkdir
gutenberg
RAW_DATA
=
${
buildout
:directory
}
/software_release/gutenberg
for
file
in
`
ls
$RAW_DATA
`
;
do
hdfs dfs
-put
$RAW_DATA
/
$file
gutenberg/
done
software/hadoop/run-demo.sh.in
0 → 100644
View file @
207570a3
#!/bin/bash
.
environment.sh
hadoop jar software_release/parts/hadoop-streaming/
*
jar
-mapper
gutenberg-mapper.py
-reducer
gutenberg-reducer.py
-input
gutenberg/
*
-output
gutenberg-output
software/hadoop/software.cfg
View file @
207570a3
...
...
@@ -9,6 +9,8 @@ parts =
eggs
java
hadoop
hadoop-streaming
gutenberg-dataset
instance
...
...
@@ -18,6 +20,7 @@ eggs =
slapos.cookbook
collective.recipe.template
cp.recipe.cmd
plone.recipe.command
[hadoop]
...
...
@@ -27,6 +30,15 @@ md5sum = 25f27eb0b5617e47c032319c0bfd9962
strip-top-level-dir = true
[hadoop-streaming]
recipe = hexagonit.recipe.download
url = http://repo1.maven.org/maven2/org/apache/hadoop/hadoop-streaming/0.20.203.0/hadoop-streaming-0.20.203.0.jar
download-only = true
#md5sum =
[instance]
recipe = slapos.recipe.template
url = ${:_profile_base_location_}/instance.cfg.in
...
...
@@ -36,3 +48,60 @@ mode = 0644
[gutenberg-dataset]
recipe = cp.recipe.cmd
update_cmd = /bin/true
install_cmd =
mkdir -p ${buildout:directory}/gutenberg
cd ${buildout:directory}/gutenberg
wget -c http://www.gutenberg.org/cache/epub/103/pg103.txt
wget -c http://www.gutenberg.org/cache/epub/18857/pg18857.txt
wget -c http://www.gutenberg.org/cache/epub/2488/pg2488.txt
wget -c http://www.gutenberg.org/cache/epub/164/pg164.txt
wget -c http://www.gutenberg.org/cache/epub/1268/pg1268.txt
wget -c http://www.gutenberg.org/cache/epub/800/pg800.txt
wget -c http://www.gutenberg.org/cache/epub/4791/pg4791.txt
wget -c http://www.gutenberg.org/cache/epub/3526/pg3526.txt
wget -c http://www.gutenberg.org/cache/epub/2083/pg2083.txt
#[wikipedia-dataset]
#recipe = cp.recipe.cmd
#update_cmd = /bin/true
##update_cmd = ${:install_cmd}
#install_cmd =
# mkdir -p ${buildout:directory}/raw-data
# cd ${buildout:directory}/raw-data
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current1.xml-p000000010p000010000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current2.xml-p000010001p000025000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current3.xml-p000025001p000055000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current4.xml-p000055002p000104998.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current5.xml-p000105001p000184999.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current6.xml-p000185003p000305000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current7.xml-p000305002p000464997.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current8.xml-p000465001p000665000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current9.xml-p000665001p000925000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current10.xml-p000925001p001325000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current11.xml-p001325001p001825000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current12.xml-p001825001p002425000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current13.xml-p002425001p003124998.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current14.xml-p003125001p003924999.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current15.xml-p003925001p004825000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current16.xml-p004825002p006025000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current17.xml-p006025001p007524997.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current18.xml-p007525002p009225000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current19.xml-p009225001p011125000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current20.xml-p011125001p013324998.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current21.xml-p013325001p015725000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current22.xml-p015725003p018225000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current23.xml-p018225001p020925000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current24.xml-p020925002p023725000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current25.xml-p023725001p026624999.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current26.xml-p026625002p029625000.bz2
# wget -c http://dumps.wikimedia.org/enwiki/20140203/enwiki-20140203-pages-meta-current27.xml-p029625001p041836446.bz2
software/hadoop/wikipedia-mapper.py.in
0 → 100644
View file @
207570a3
import bz2
import os
import sys
import xml.sax
class WikipediaTitleHandler(xml.sax.ContentHandler):
def startElement(self, name, attrs):
self.chars = []
self.tag = name
def characters(self, content):
if self.tag == 'title':
self.chars.append(content)
def endElement(self, name):
if self.tag == 'title':
title = ''.join(self.chars)
if title.startswith('Talk:'):
return
if title.startswith('User talk:'):
return
if title.startswith('Wikipedia:'):
return
if title.startswith('Wikipedia talk:'):
return
if title.startswith('User:'):
return
print title.encode('utf8')
def process_xml(input):
parser = xml.sax.make_parser()
parser.setContentHandler(WikipediaTitleHandler())
parser.parse(input)
if __name__ == '__main__':
input = bz2.BZ2File('/dev/fd/0')
process_xml(input)
# dirname = '/srv/slapgrid/slappart20/srv/runner/instance/slappart0/software_release/raw-data/'
# filenames = os.listdir(dirname)
# # ['enwiki-20140203-pages-meta-current1.xml-p000000010p000010000.bz2']
# for fname in filenames:
# process_xml(os.path.join(dirname, fname))
# input = bz2.BZ2File(process_xml(os.path.join(dirname, fname)))
software/hadoop/wikipedia-reducer.py.in
0 → 100644
View file @
207570a3
import bz2
import os
import sys
import xml.sax
class WikipediaTitleHandler(xml.sax.ContentHandler):
def startElement(self, name, attrs):
self.chars = []
self.tag = name
def characters(self, content):
if self.tag == 'title':
self.chars.append(content)
def endElement(self, name):
if self.tag == 'title':
title = ''.join(self.chars)
if title.startswith('Talk:'):
return
if title.startswith('User talk:'):
return
if title.startswith('Wikipedia:'):
return
if title.startswith('Wikipedia talk:'):
return
if title.startswith('User:'):
return
print title.encode('utf8')
def process_xml(input):
parser = xml.sax.make_parser()
parser.setContentHandler(WikipediaTitleHandler())
parser.parse(input)
if __name__ == '__main__':
input = bz2.BZ2File('/dev/fd/0')
process_xml(input)
# dirname = '/srv/slapgrid/slappart20/srv/runner/instance/slappart0/software_release/raw-data/'
# filenames = os.listdir(dirname)
# # ['enwiki-20140203-pages-meta-current1.xml-p000000010p000010000.bz2']
# for fname in filenames:
# process_xml(os.path.join(dirname, fname))
# input = bz2.BZ2File(process_xml(os.path.join(dirname, fname)))
Write
Preview
Markdown
is supported
0%
Try again
or
attach a new file
Attach a file
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Cancel
Please
register
or
sign in
to comment