Hash連線,如記憶體足夠，首先遍歷內表建立Hash表,然後遍歷外表,對連線鍵計算HashCode,如一致,則遍歷Hash表中具有同一HashCode的連結串列，值一致，則返回該值。
如記憶體不夠，可遍歷兩張表，使用同樣的Hash函式把表拆分為N個Hash“分割槽”，遍歷內表每一個Hash分割槽和外表相應的Hash分割槽，如找到與連線鍵值一致的資料，則返回該值。

詳見程式碼註釋.

```C

#include <stdio.h>
#include <stdlib.h>
#include "hash_join.h"

#define MAX_ELEMENTS 1024

//生成hash code
static int generate_hashcode(int n)
{
  return n % HASH_BUCKET;
}

//生成hash桶(寫入到檔案中,以檔案的方式模擬)
static int generate_bucket(FILE *file,char *tag)
{
  printf("----------- generate_bucket ---------- \n");
  //陣列
  char buf[MAX_BYTES];
  FILE *fd = NULL;
  for(;!feof(file);)
  {
    int x = read_int(file,buf);
    if(x == 0)
      break;
    int hashcode = generate_hashcode(x);
    char filename[30];
    sprintf(filename,"/cygdrive/d/tmp/hash/%s_%d.csv",tag,hashcode);
    //printf("Hash code is %d,Bucket filename is %s.\n",hashcode,filename);
    fd = fopen(filename,"a");
    if(fd == NULL)
    {
      printf("Can not open file %s.\n",filename);
      return 0;
    }  
    //寫入檔案中
    write_int(fd,x);
    fclose(fd);
  }
  return 1;
}

//把hash表載入到記憶體中,適用於記憶體足夠的情況
//使用二維陣列模擬Hash表,D1 : hash桶,D2 : 桶中的資料
static int load_hashtable(int ht[][MAX_ELEMENTS])
{
  printf("----------- load_hashtable ---------- \n");
  for(int i=0;i < HASH_BUCKET;i++)
  {
    //迴圈桶號
    char filename[MAX_BYTES];
    //讀檔案
    sprintf(filename,"/cygdrive/d/tmp/hash/inner_%d.csv",i);
    FILE *fd = fopen(filename,"r");
    if(fd == NULL){
      //printf("Can not open file : %s\n",filename);
      continue;
    }
    int j=0;
    char buf[MAX_BYTES];
    for(;!feof(fd) && j < MAX_ELEMENTS;)
    {
      //把檔案內容放到陣列中
      int x = read_int(fd,buf);
      ht[i][j++] = x;
    }
    fclose(fd);
  }
  return 1;
}

//使用記憶體建立hash表進行hash連線
static void hash_join_onmemory(FILE *outerfile,FILE *innerfile)
{
  printf("----------- hash_join_onmemory ---------- \n");
  int ht[HASH_BUCKET][MAX_ELEMENTS];
  char buffer[MAX_BYTES];
  int flag = 0;
 
  //建立hash bucket檔案
  flag = generate_bucket(innerfile,"inner");
  if(!flag)
  {
    printf("Can not generate bucket file!\n");
    return;
  }
  //載入到hash表中(二維陣列模擬)
  flag = load_hashtable(ht);
  if(!flag)
  {
    printf("Can not load hash table!\n");
    return;
  }
  
  //遍歷第二個檔案,執行JOIN
  for(;!feof(outerfile);)
  {
    //讀第二個檔案,執行join
    int outer = read_int(outerfile,buffer);
    //計算hashcode
    int hashcode = generate_hashcode(outer);
    for(int i=0;i < MAX_ELEMENTS;i++)
    {
      //遍歷hash桶中的資料,找到對應的資料
      if(ht[hashcode][i] == outer)
      {
        printf("Found one,hash bucket is %d,value is : %d.\n",hashcode,outer);
      }
    }
  }
}

//使用磁碟快取進行hash連線
static void hash_join_ondisk(FILE *outerfile,FILE *innerfile)
{
  printf("----------- hash_join_ondisk ---------- \n");
  char buffer[MAX_BYTES];
  int flag = 0;
 
  //建立hash"桶"檔案
  flag = generate_bucket(innerfile,"inner");
  if(!flag)
  {
    printf("Can not generate inner bucket file!\n");
    return;
  }
  flag = generate_bucket(outerfile,"outer");
  if(!flag)
  {
    printf("Can not generate outer bucket file!\n");
    return;
  }
  //遍歷hash值相同的檔案,執行連線
  for(int i=0;i < HASH_BUCKET;i++)
  {
    //從0號桶開始
    char innerfname[MAX_BYTES];
    char outerfname[MAX_BYTES];
    //讀檔案
    sprintf(innerfname,"/cygdrive/d/tmp/hash/%s_%d.csv","inner",i);
    sprintf(outerfname,"/cygdrive/d/tmp/hash/%s_%d.csv","outer",i);
    FILE *fd_inner = fopen(innerfname,"r");
    if(fd_inner == NULL){
      //printf("Can not open file : %s\n",filename);
      continue;
    }
    FILE *fd_outer = fopen(outerfname,"r");
    if(fd_outer == NULL)
    {
      continue;
    }
    for(;!feof(fd_outer);)
    {
      int v_out = read_int(fd_outer,buffer);
      if(v_out == 0)
        continue;
      for(;!feof(fd_inner);)
      {
        int v_in = read_int(fd_inner,buffer);
        if(v_in == 0)
          continue;
        if(v_out == v_in)
        {
          printf("Found one,hash bucket is %d,value is : %d.\n",i,v_out);        
        }
      }
      rewind(fd_inner);
    }
  }
}

//執行Hash連線
void hash_join(char *file1,char * file2,char *flag)
{
  printf("----------- hash join ---------- \n");
  FILE *outerfile = fopen(file1,"r");
  if(outerfile == NULL)
  {
    printf("Can not open file %s.\n",file1);
    return;
  }
  //開啟第二個檔案
  FILE *innerfile = fopen(file2,"r");
  if(innerfile == NULL)
  {
    printf("Can not open file %s.\n",file2);
    return;
  }
  //執行JOIN
  if(strcmp(flag,"memory") == 0)
    hash_join_onmemory(outerfile,innerfile);
  else
    hash_join_ondisk(outerfile,innerfile);
  //關閉
  fclose(outerfile);
  fclose(innerfile);
}

```

執行輸出
```bash
$ cat file1.csv
1
2
3
4
5
1
234
2939
9002
20
$ cat file2.csv
11
20
3
40
55
50
234
33
90
1

$ /cygdrive/d/tmp/test.exe file1.csv file2.csv
------------- use memory ------------------
----------- hash join ----------
----------- hash_join_onmemory ----------
----------- generate_bucket ----------
----------- load_hashtable ----------
Found one,hash bucket is 1,value is : 1.
Found one,hash bucket is 3,value is : 3.
Found one,hash bucket is 1,value is : 1.
Found one,hash bucket is 106,value is : 234.
Found one,hash bucket is 20,value is : 20.
------------- use disk ------------------
----------- hash join ----------
----------- hash_join_ondisk ----------
----------- generate_bucket ----------
----------- generate_bucket ----------
Found one,hash bucket is 1,value is : 1.
Found one,hash bucket is 1,value is : 1.
Found one,hash bucket is 3,value is : 3.
Found one,hash bucket is 20,value is : 20.
Found one,hash bucket is 106,value is : 234.

```</stdlib.h></stdio.h>

資料庫實現原理#4（Hash Join）

相關文章