Twoway External Merge Sort

1
Two-way External Merge Sort
2
Outline

• Assumptions
• Two-way external merge sort algorithm
• Input and output file format
• File manipulation in JAVA
• Code skeleton

3
Assumptions

• The main memory is split into three fixed-size
  blocks
• Two input blocks hold the records to be sorted
• One output block hold the records that have been
  sorted
• A record has fixed size
• A file comprises 2k blocks of records

input block 1
input block 2
output block
main memory
a file ( k 2)
a record
4
Two-way External Merge Sort Algorithm
1
11
7
12
input
9
8
2
4
3
5
6
10
5
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
1
3
5
4
6
6
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
1
3
5
2
4
6
7
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
1
3
5
2
4
6
3
8
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
1
3
5
2
4
6
3
9
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
1
1
3
5
2
2
4
6
3
3
10
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
1
3
5
2
4
6
3
11
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
1
2
4
1
3
5
2
4
6
3
12
Two-way External Merge Sort Algorithm (Cont)
input block 1
input block 2
output block
output file
9
2
4
1
10
5
2
12
6
3
13
Input and Output File Format

• Input/output file consists of records
• They are not text files
• Student record structure
• sid int (4 bytes), PRIMARY KEY
• fName String (16 bytes)
• lName String (16 bytes)

14
File Manipulation in JAVA

• Create a file
• RandomAccessFile f new RandomAccessFile
  (filename, mode)
• Open a file
• RandomAccessFile f new RandomAccessFile
  (filename, r)
• Seek
• f.seek(pos)
• Read
• f.sead(buffer, offset, length)
• f.read(buffer)
• Write
• f.write(buffer, offset, length)
• Close a file
• f.close()

15
Code Skeleton

• A JAVA console application
• Three classes
• MyRecord
• MyBlockMgr
• Sort

16
MyRecord Class

• public class MyRecord implements Comparable
• / Record structure
• sid int (4 bytes) -- search key
• fName String (16 characters - 16
  bytes )
• lName String (16 characters - 16
  bytes )
• /
• public static int length 36
• private byte buf
• public MyRecord()
• buf new bytelength
• public MyRecord(byte b)
• buf b

17
MyRecord Class (Cont)

• public int getSid()
• int value 0
• for(int i0 ilt3 i)
• value value (bufi 0x00FF)
• value value ltlt 8
• value value (buf3 0x00FF)
• return value
• public void setSid(int id) throws
  IOException
• buf0 (byte)(((id gtgtgt 24) 0x00FF))
• buf1 (byte)(((id gtgtgt 16) 0x00FF))
• buf2 (byte)(((id gtgtgt 8) 0x00FF))
• buf3 (byte)(((id gtgtgt 0) 0x00FF))

18
MyRecord Class (Cont)

• public String getFirstName()
• String fn new String()
• for(int ii4iilt20ii)
• fn (char)bufii
• return fn.trim()
• public String getLastName()
• String ln new String()
• for(int j20jlt36j)
• ln (char)bufj
• return ln.trim()

19
MyRecord Class (Cont)

• // Get byte-array representation of a record
• public byte getBytes()
• return buf
• //Import a record from its byte-array
  representation.
• //"buffer" indicates the byte array
  containing the record content
• //"offset" indicates where the record content
  starts
• public void importRecord(byte buffer,int
  offset)
• for(int i0iltlengthi)
• bufi bufferoffset
• public int compareTo(Object o)
• return this.getSid() - ((MyRecord)(o)).getSi
  d()

20
MyBlockMgr Class

• Manage the three memory blocks
• Provide methods to read/write a block from/to a
  relation
• You implement the following methods
• int getNumOfBlocks(FileStream f)
• Return the number of blocks in a file
• void readBlock(FileStream fin, int blockIndex,
  int inBlock)
• Read the block specified by blockIndex from
  FileStream fin into inB1 (if inBlock 1) or
  inB2(if inBlock 2).
• It first seeks to the position
  blockIndexblockSize, and then read blockSize
  bytes into inB1 or inB2.
• void writeBlock(FileStream fout)
• Append the content of outB (blockSize bytes)
  into file fout.
• No seek is performed in this routine.

21
Sort Class

• Sort a given file on attribute sid in ascending
  order
• Execute method is the entry of your code
• You are free to develop additional classes.

22
Sort Class (Cont.)

• public class Sort
• RandomAccessFile fin, fout
• String inputFile
• MyBlockMgr bm
• int blockNum
• int recNumPerBlock
• public Sort(String inputFile)
• this.inputFile inputFile
• private String getOutputFileName(int round)
• return inputFile ".out" round

23
public void execute()throws IOException
try fin new RandomAccessFile(inputFi
le, "r") catch (IOException e)
System.out.println("Error openning the input
File! Program aborted") System.exit(1)
bm new MyBlockMgr() blockNum
bm.getNumberOfBlocks(fin) recNumPerBlock
bm.getInBlock1().length / MyRecord.length
internalSortBlocks() fin.close()
externalSortBlocks()
24

• public void printFile(String fileName)throws
  IOException
• RandomAccessFile fin new RandomAccessFile(fi
  leName,"r")
• byte b new byte36
• for(int i0iltfin.length()/MyRecord.lengthii
  1)
• fin.read(b)
• MyRecord m new MyRecord(b)
• System.out.println(i" "m.getSid()"
  "m.getFirstName()" "m.getLastName())
• fin.close()