We see some time there is requirement where you need to split a file into different individual files based on some key value. You can do it using Java/C/C++ or any other programming language where you write some dozens of lines of code which is fine if the file size is 1GB or less. What if the file size is greater than 1GB? This process runs forever to do the job.

On Hadoop system using Apache Pig you can write very simple code that will split file on the fly. You will have the flexibility to control flow of data and do manipulations (if any) and split file.

Now we see how to split file into individual files using Pig Script. Here is our sample file TEST.DEV.ENV.SAMPLE.FILE –

 HEADER1DIBBCCLY8-9568347556434756972CMMS21WUE
000001010K1DIBB  7RHUN  2100000AE            J82V  2269167AD         2002-03-079999-12-31+000000000100000    22004-02-28-20.00.13.106749
000002010K1DIBB  7RHUN  2100000AE            J82V  2269167AD         2002-03-072004-07-30+000000000100000    32004-02-28-20.00.13.106749
9TRAILR1DIBBCCLY8-95683475564347560000084
 HEADERVV9IHFYSKN-4654178251104433898CMMS21ANI
0000030108VV9IH  AKB7L  3300000AE            XMV9  2269167AD         2002-03-059999-12-31+000000000100000    22004-02-28-20.00.13.106749
0000040108VV9IH  AKB7L  3300000SE            XMV9  2269167AD         2002-03-052004-07-30+000000000100000    32004-02-28-20.00.13.106749
9TRAILRVV9IHFYSKN-46541782511044330000510
 HEADERLY674FBNR4-9375012333185998800CMMS21AIZ
000005010ULY674  5XWJR  0100000SE            XMV9  2269167AD         2002-03-059999-12-31+000000000100000    22004-02-28-20.00.13.106749
000006010ULY674  5XWJR  0100000AE            XMV9  2269167AD         2002-03-052004-07-30+000000000100000    32004-02-28-20.00.13.106749
9TRAILRLY674FBNR4-93750123331859980000150
 HEADERT0X36Q6YVQ-5632769394873798290CMMS21WLO
000007010QT0X36  RJPWK  5500000AE            J82V  2269167AD         2002-10-229999-12-31+000000000100000    22004-02-28-20.00.13.106749
000008010QT0X36  RJPWK  5500000AE            J82V  2269167AD         2002-10-222004-07-30+000000000100000    32004-02-28-20.00.13.106749
9TRAILRT0X36Q6YVQ-56327693948737980000642
 HEADER8LIKAC67U9-2737265552238819829CMMS21HMV
000009010S8LIKA  07L1X  4400000BE            J82V  2269167AD         2002-03-079999-12-31+000000000100000    22004-02-28-20.00.13.106749
000010010S8LIKA  07L1X  4400000BE            J82V  2269167AD         2002-03-072004-07-30+000000000100000    32004-02-28-20.00.13.106749
9TRAILR8LIKAC67U9-27372655522388190000412
 HEADERAS7G3QPIUC-8825934656338659366CMMS21BQA
000011010QAS7G3  CO46Q  8500000BE            RI12  2269167AD         2002-03-059999-12-31+000000000100000    22004-02-28-20.00.13.106749
000012010QAS7G3  CO46Q  8500000BE            RI12  2269167AD         2002-03-052004-07-30+000000000100000    32004-02-28-20.00.13.106749
9TRAILRAS7G3QPIUC-88259346563386590000865

In the sample file if you see there are lot of headers and trailers and some data between them. Our requirement is to split each set of data with HEADER, TRAILER and DETAIL DATA into individual files. For our sample it should generate 6 different files.

We will split file using key values in the file. Here we use positions 11-15 (5 characters) in DETAIL DATA and positions 8-12 in HEADER and TRAILER data. We must maintain the consistency of the data as Header row, Details rows and Trailer row. This makes the complete structure of file and keeps all data together.

Now we write Pig Script to split the file –

REGISTER /home/jars/pig/piggybank.jar;

A  = LOAD  '/path/to/input/file/TEST.DEV.ENV.SAMPLE.FILE'
     USING PigStorage('\t') AS (line:chararray);

B  = FILTER A BY SUBSTRING(line, 1, 7) != 'HEADER' AND SUBSTRING(line, 0, 7) != '9TRAILR';

C  = FILTER A BY SUBSTRING(line, 1, 7) == 'HEADER' OR SUBSTRING(line, 0, 7) == '9TRAILR';

-- Extract data based on key value from Header, Details and Trailer rows 
D  = GROUP B BY SUBSTRING($0, 10, 15);
E  = GROUP C BY SUBSTRING($0, 7, 12);

F  = UNION D, E;

G  = FOREACH F GENERATE FLATTEN($0), FLATTEN($1);

SPLIT G INTO H IF SIZE($0) > 0, X IF SIZE($0) <= 0;

J  = ORDER H BY $1;

STORE J INTO '/path/to/output/directory'
        -- Stores using \t as the input separator 
        USING org.apache.pig.piggybank.storage.MultiStorage('/path/to/output/directory', '0');

File is split on the value we taken as key value. Here our key value from all rows (Header, Details and Trailer) is 5 charcters specified using function substring() at transformtions ‘D’ and ‘E’.

Now we will see output directory for files –

I have given you the idea, rest is your imagination.