Chris@1: /********************************************************************
Chris@1:  *                                                                  *
Chris@1:  * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE.   *
Chris@1:  * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS     *
Chris@1:  * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
Chris@1:  * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING.       *
Chris@1:  *                                                                  *
Chris@1:  * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2001             *
Chris@1:  * by the Xiph.Org Foundation http://www.xiph.org/                  *
Chris@1:  *                                                                  *
Chris@1:  ********************************************************************
Chris@1: 
Chris@1:  function: utility for finding the distribution in a data set
Chris@1:  last mod: $Id: distribution.c 16037 2009-05-26 21:10:58Z xiphmont $
Chris@1: 
Chris@1:  ********************************************************************/
Chris@1: 
Chris@1: #include <stdlib.h>
Chris@1: #include <stdio.h>
Chris@1: #include <math.h>
Chris@1: #include <string.h>
Chris@1: #include <errno.h>
Chris@1: #include "bookutil.h"
Chris@1: 
Chris@1: /* command line:
Chris@1:    distribution file.vqd
Chris@1: */
Chris@1: 
Chris@1: int ascend(const void *a,const void *b){
Chris@1:   return(**((long **)a)-**((long **)b));
Chris@1: }
Chris@1: 
Chris@1: int main(int argc,char *argv[]){
Chris@1:   FILE *in;
Chris@1:   long lines=0;
Chris@1:   float min;
Chris@1:   float max;
Chris@1:   long bins=-1;
Chris@1:   int flag=0;
Chris@1:   long *countarray;
Chris@1:   long total=0;
Chris@1:   char *line;
Chris@1: 
Chris@1:   if(argv[1]==NULL){
Chris@1:     fprintf(stderr,"Usage: distribution {data.vqd [bins]| book.vqh} \n\n");
Chris@1:     exit(1);
Chris@1:   }
Chris@1:   if(argv[2]!=NULL)
Chris@1:     bins=atoi(argv[2])-1;
Chris@1: 
Chris@1:   in=fopen(argv[1],"r");
Chris@1:   if(!in){
Chris@1:     fprintf(stderr,"Could not open input file %s\n",argv[1]);
Chris@1:     exit(1);
Chris@1:   }
Chris@1: 
Chris@1:   if(strrchr(argv[1],'.') && strcmp(strrchr(argv[1],'.'),".vqh")==0){
Chris@1:     /* load/decode a book */
Chris@1: 
Chris@1:     codebook *b=codebook_load(argv[1]);
Chris@1:     static_codebook *c=(static_codebook *)(b->c);
Chris@1:     float delta;
Chris@1:     int i;
Chris@1:     fclose(in);
Chris@1: 
Chris@1:     switch(c->maptype){
Chris@1:     case 0:
Chris@1:       printf("entropy codebook only; no mappings\n");
Chris@1:       exit(0);
Chris@1:       break;
Chris@1:     case 1:
Chris@1:       bins=_book_maptype1_quantvals(c);
Chris@1:       break;
Chris@1:     case 2:
Chris@1:       bins=c->entries*c->dim;
Chris@1:       break;
Chris@1:     }
Chris@1: 
Chris@1:     max=min=_float32_unpack(c->q_min);
Chris@1:     delta=_float32_unpack(c->q_delta);
Chris@1: 
Chris@1:     for(i=0;i<bins;i++){
Chris@1:       float val=c->quantlist[i]*delta+min;
Chris@1:       if(val>max)max=val;
Chris@1:     }
Chris@1: 
Chris@1:     printf("Minimum scalar value: %f\n",min);
Chris@1:     printf("Maximum scalar value: %f\n",max);
Chris@1: 
Chris@1:     switch(c->maptype){
Chris@1:     case 1:
Chris@1:       {
Chris@1:         /* lattice codebook.  dump it. */
Chris@1:         int j,k;
Chris@1:         long maxcount=0;
Chris@1:         long **sort=calloc(bins,sizeof(long *));
Chris@1:         long base=c->lengthlist[0];
Chris@1:         countarray=calloc(bins,sizeof(long));
Chris@1: 
Chris@1:         for(i=0;i<bins;i++)sort[i]=c->quantlist+i;
Chris@1:         qsort(sort,bins,sizeof(long *),ascend);
Chris@1: 
Chris@1:         for(i=0;i<b->entries;i++)
Chris@1:           if(c->lengthlist[i]>base)base=c->lengthlist[i];
Chris@1: 
Chris@1:         /* dump a full, correlated count */
Chris@1:         for(j=0;j<b->entries;j++){
Chris@1:           if(c->lengthlist[j]){
Chris@1:             int indexdiv=1;
Chris@1:             printf("%4d: ",j);
Chris@1:             for(k=0;k<b->dim;k++){        
Chris@1:               int index= (j/indexdiv)%bins;
Chris@1:               printf("%+3.1f,", c->quantlist[index]*_float32_unpack(c->q_delta)+
Chris@1:                      _float32_unpack(c->q_min));
Chris@1:               indexdiv*=bins;
Chris@1:             }
Chris@1:             printf("\t|");
Chris@1:             for(k=0;k<base-c->lengthlist[j];k++)printf("*");
Chris@1:             printf("\n");
Chris@1:           }
Chris@1:         }
Chris@1: 
Chris@1:         /* do a rough count */
Chris@1:         for(j=0;j<b->entries;j++){
Chris@1:           int indexdiv=1;
Chris@1:           for(k=0;k<b->dim;k++){
Chris@1:             if(c->lengthlist[j]){
Chris@1:               int index= (j/indexdiv)%bins;
Chris@1:               countarray[index]+=(1<<(base-c->lengthlist[j]));
Chris@1:               indexdiv*=bins;
Chris@1:             }
Chris@1:           }
Chris@1:         }
Chris@1: 
Chris@1:         /* dump the count */
Chris@1: 
Chris@1:         {
Chris@1:           long maxcount=0,i,j;
Chris@1:           for(i=0;i<bins;i++)
Chris@1:             if(countarray[i]>maxcount)maxcount=countarray[i];
Chris@1:       
Chris@1:           for(i=0;i<bins;i++){
Chris@1:             int ptr=sort[i]-c->quantlist;
Chris@1:             int stars=rint(50./maxcount*countarray[ptr]);
Chris@1:             printf("%+08f (%8ld) |",c->quantlist[ptr]*delta+min,countarray[ptr]);
Chris@1:             for(j=0;j<stars;j++)printf("*");
Chris@1:             printf("\n");
Chris@1:           }
Chris@1:         }
Chris@1:       }
Chris@1:       break;
Chris@1:     case 2:
Chris@1:       {
Chris@1:         /* trained, full mapping codebook. */
Chris@1:         printf("Can't do probability dump of a trained [type 2] codebook (yet)\n");
Chris@1:       }
Chris@1:       break;
Chris@1:     }
Chris@1:   }else{
Chris@1:     /* load/count a data file */
Chris@1: 
Chris@1:     /* do it the simple way; two pass. */
Chris@1:     line=setup_line(in);
Chris@1:     while(line){      
Chris@1:       float code;
Chris@1:       char buf[80];
Chris@1:       lines++;
Chris@1: 
Chris@1:       sprintf(buf,"getting min/max (%.2f::%.2f). lines...",min,max);
Chris@1:       if(!(lines&0xff))spinnit(buf,lines);
Chris@1:       
Chris@1:       while(!flag && sscanf(line,"%f",&code)==1){
Chris@1:         line=strchr(line,',');
Chris@1:         min=max=code;
Chris@1:         flag=1;
Chris@1:       }
Chris@1:       
Chris@1:       while(line && sscanf(line,"%f",&code)==1){
Chris@1:         line=strchr(line,',');
Chris@1:         if(line)line++;
Chris@1:         if(code<min)min=code;
Chris@1:         if(code>max)max=code;
Chris@1:       }
Chris@1:       
Chris@1:       line=setup_line(in);
Chris@1:     }
Chris@1:     
Chris@1:     if(bins<1){
Chris@1:       if((int)(max-min)==min-max){
Chris@1:         bins=max-min;
Chris@1:       }else{
Chris@1:         bins=25;
Chris@1:       }
Chris@1:     }
Chris@1:     
Chris@1:     printf("\r                                                     \r");
Chris@1:     printf("Minimum scalar value: %f\n",min);
Chris@1:     printf("Maximum scalar value: %f\n",max);
Chris@1: 
Chris@1:     if(argv[2]){
Chris@1:       
Chris@1:       printf("\n counting hits into %ld bins...\n",bins+1);
Chris@1:       countarray=calloc(bins+1,sizeof(long));
Chris@1:       
Chris@1:       rewind(in);
Chris@1:       line=setup_line(in);
Chris@1:       while(line){      
Chris@1:         float code;
Chris@1:         lines--;
Chris@1:         if(!(lines&0xff))spinnit("counting distribution. lines so far...",lines);
Chris@1:         
Chris@1:         while(line && sscanf(line,"%f",&code)==1){
Chris@1:           line=strchr(line,',');
Chris@1:           if(line)line++;
Chris@1:           
Chris@1:           code-=min;
Chris@1:           code/=(max-min);
Chris@1:           code*=bins;
Chris@1:           countarray[(int)rint(code)]++;
Chris@1:           total++;
Chris@1:         }
Chris@1:         
Chris@1:         line=setup_line(in);
Chris@1:       }
Chris@1:     
Chris@1:       /* make a pretty graph */
Chris@1:       {
Chris@1:         long maxcount=0,i,j;
Chris@1:         for(i=0;i<bins+1;i++)
Chris@1:           if(countarray[i]>maxcount)maxcount=countarray[i];
Chris@1:         
Chris@1:         printf("\r                                                     \r");
Chris@1:         printf("Total scalars: %ld\n",total);
Chris@1:         for(i=0;i<bins+1;i++){
Chris@1:           int stars=rint(50./maxcount*countarray[i]);
Chris@1:           printf("%08f (%8ld) |",(max-min)/bins*i+min,countarray[i]);
Chris@1:           for(j=0;j<stars;j++)printf("*");
Chris@1:           printf("\n");
Chris@1:         }
Chris@1:       }
Chris@1:     }
Chris@1: 
Chris@1:     fclose(in);
Chris@1: 
Chris@1:   }
Chris@1:   printf("\nDone.\n");
Chris@1:   exit(0);
Chris@1: }