Switch branches/tags
Nothing to show
Find file History
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
..
Failed to load latest commit information.
releases/org/nts/pigutils/udfs
src
README.markdown
pom.xml
redeploy.sh

README.markdown

#Pig Utils

This library is freely distributed on the Apache2 Licence.

Its a collection of Pig UDFS, Loaders and Stores providing functionality on top of the current pig libraries.

The latest install package (linux) is provided at : https://github.com/gerritjvv/repo/tree/master/rpms

#UDFS

  • TO_MAP org.nts.pigutils.udfs.TO_MAP(input)

    Converts a DataBag or Tuple into a HashMap

    The Tuple or Tuples in the Bag is expected to have between one and two keys.

    If the Tuple has:
    One entry, the entry is placed as a key and the value as null.
    Two entries, the first is taken as the key and the second as the value.
    Three or more, the first is taken as the key and the rest placed as a Tuple and taken as the value.

    Important, keys are always converted to Strings.

  • FIND_MAP and FIND_MAP_MAP

    Does a regex pattern find on each key, if true the value for that key is returned.

    The FIND_MAP_MAP UDF expects the value to be a Map and casts it.

  DEFINE FIND org.nts.pigutils.udfs.MAP_FIND('.*(info)+');
  DEFINE FIND_MAP org.nts.pigutils.udfs.MAP_FIND_MAP('.*(-csv)+');

  r = foreach a generate FIND(mymap), FIND_MAP(mymap)#'size';

  • DateFormat org.nts.pigutils.udfs.DateFormat(tsMillis, 'formatstring')

    The formatstring is the same as for Java's SimpleDateFormat e.g. 'yyyy-MM'dd' will format a timestamp to year month and day

  • PurifyInt org.nts.udfs.PurifyInt(strInteger)

    Takes a String, removes any non integer characters and returns an integer

#LZO GPB Loader

  • LZO Protobuf Base64 Loader

org.nts.pigutils.proto.LzoProtobuffB64LinePigStore('gpbkey')

The gpbkey should be configured in your pig.properties file as a key and the value should point to the actual protobuff java class

JSON Loader

This json loader loads records of type Map into a java HashMap using the Jackson ObjectMapper. All types are correctly translated into their java types i.e. {"age":1} will result in a hash map with key=String("age") and value=Integer(1)

Note:

All lists and Object arrays are converted to Tuples

To use

  l = load 'myfile.json.gz' using org.nts.pigutils.udfs.JSONLoader();
  r = foreach l generate m#'age', m#'name';

SOLR Store

  • Store data to a SOLR Cloud Server

    Example

  l = load '/opt/solr/exampledocs/pigbooks.csv' using PigStorage(',') as 
	(id:chararray,cat:chararray,name:chararray,price:double,
	inStock:chararray,author:chararray,series_t:chararray,sequence_i:chararray,
	genre_s:chararray);

	
  store l into '/tmp/abc' using org.nts.pigutils.lucene.SolrCloudStore('localhost:9983', 'collection1');	 

Lucene Index Create Store

  • Create Lucene Indexes using Pig

    Exmaple

   r = load 'luceneinput.csv' as (lbl:chararray,desc:chararray,score:int);
   store r into 'target/luceneindex' using org.nts.pigutils.lucene.LuceneStore();