1 /********************************************************************
2 * *
3 * THIS FILE IS PART OF THE OggVorbis SOFTWARE CODEC SOURCE CODE. *
4 * USE, DISTRIBUTION AND REPRODUCTION OF THIS LIBRARY SOURCE IS *
5 * GOVERNED BY A BSD-STYLE SOURCE LICENSE INCLUDED WITH THIS SOURCE *
6 * IN 'COPYING'. PLEASE READ THESE TERMS BEFORE DISTRIBUTING. *
7 * *
8 * THE OggVorbis SOURCE CODE IS (C) COPYRIGHT 1994-2001 *
9 * by the Xiph.Org Foundation http://www.xiph.org/ *
10 * *
11 ********************************************************************
12
13 function: utility for finding the distribution in a data set
14 last mod: $Id: distribution.c 16037 2009-05-26 21:10:58Z xiphmont $
15
16 ********************************************************************/
17
18 #include <stdlib.h>
19 #include <stdio.h>
20 #include <math.h>
21 #include <string.h>
22 #include <errno.h>
23 #include "bookutil.h"
24
25 /* command line:
26 distribution file.vqd
27 */
28
ascend(const void * a,const void * b)29 int ascend(const void *a,const void *b){
30 return(**((long **)a)-**((long **)b));
31 }
32
main(int argc,char * argv[])33 int main(int argc,char *argv[]){
34 FILE *in;
35 long lines=0;
36 float min;
37 float max;
38 long bins=-1;
39 int flag=0;
40 long *countarray;
41 long total=0;
42 char *line;
43
44 if(argv[1]==NULL){
45 fprintf(stderr,"Usage: distribution {data.vqd [bins]| book.vqh} \n\n");
46 exit(1);
47 }
48 if(argv[2]!=NULL)
49 bins=atoi(argv[2])-1;
50
51 in=fopen(argv[1],"r");
52 if(!in){
53 fprintf(stderr,"Could not open input file %s\n",argv[1]);
54 exit(1);
55 }
56
57 if(strrchr(argv[1],'.') && strcmp(strrchr(argv[1],'.'),".vqh")==0){
58 /* load/decode a book */
59
60 codebook *b=codebook_load(argv[1]);
61 static_codebook *c=(static_codebook *)(b->c);
62 float delta;
63 int i;
64 fclose(in);
65
66 switch(c->maptype){
67 case 0:
68 printf("entropy codebook only; no mappings\n");
69 exit(0);
70 break;
71 case 1:
72 bins=_book_maptype1_quantvals(c);
73 break;
74 case 2:
75 bins=c->entries*c->dim;
76 break;
77 }
78
79 max=min=_float32_unpack(c->q_min);
80 delta=_float32_unpack(c->q_delta);
81
82 for(i=0;i<bins;i++){
83 float val=c->quantlist[i]*delta+min;
84 if(val>max)max=val;
85 }
86
87 printf("Minimum scalar value: %f\n",min);
88 printf("Maximum scalar value: %f\n",max);
89
90 switch(c->maptype){
91 case 1:
92 {
93 /* lattice codebook. dump it. */
94 int j,k;
95 long maxcount=0;
96 long **sort=calloc(bins,sizeof(long *));
97 long base=c->lengthlist[0];
98 countarray=calloc(bins,sizeof(long));
99
100 for(i=0;i<bins;i++)sort[i]=c->quantlist+i;
101 qsort(sort,bins,sizeof(long *),ascend);
102
103 for(i=0;i<b->entries;i++)
104 if(c->lengthlist[i]>base)base=c->lengthlist[i];
105
106 /* dump a full, correlated count */
107 for(j=0;j<b->entries;j++){
108 if(c->lengthlist[j]){
109 int indexdiv=1;
110 printf("%4d: ",j);
111 for(k=0;k<b->dim;k++){
112 int index= (j/indexdiv)%bins;
113 printf("%+3.1f,", c->quantlist[index]*_float32_unpack(c->q_delta)+
114 _float32_unpack(c->q_min));
115 indexdiv*=bins;
116 }
117 printf("\t|");
118 for(k=0;k<base-c->lengthlist[j];k++)printf("*");
119 printf("\n");
120 }
121 }
122
123 /* do a rough count */
124 for(j=0;j<b->entries;j++){
125 int indexdiv=1;
126 for(k=0;k<b->dim;k++){
127 if(c->lengthlist[j]){
128 int index= (j/indexdiv)%bins;
129 countarray[index]+=(1<<(base-c->lengthlist[j]));
130 indexdiv*=bins;
131 }
132 }
133 }
134
135 /* dump the count */
136
137 {
138 long maxcount=0,i,j;
139 for(i=0;i<bins;i++)
140 if(countarray[i]>maxcount)maxcount=countarray[i];
141
142 for(i=0;i<bins;i++){
143 int ptr=sort[i]-c->quantlist;
144 int stars=rint(50./maxcount*countarray[ptr]);
145 printf("%+08f (%8ld) |",c->quantlist[ptr]*delta+min,countarray[ptr]);
146 for(j=0;j<stars;j++)printf("*");
147 printf("\n");
148 }
149 }
150 }
151 break;
152 case 2:
153 {
154 /* trained, full mapping codebook. */
155 printf("Can't do probability dump of a trained [type 2] codebook (yet)\n");
156 }
157 break;
158 }
159 }else{
160 /* load/count a data file */
161
162 /* do it the simple way; two pass. */
163 line=setup_line(in);
164 while(line){
165 float code;
166 char buf[80];
167 lines++;
168
169 sprintf(buf,"getting min/max (%.2f::%.2f). lines...",min,max);
170 if(!(lines&0xff))spinnit(buf,lines);
171
172 while(!flag && sscanf(line,"%f",&code)==1){
173 line=strchr(line,',');
174 min=max=code;
175 flag=1;
176 }
177
178 while(line && sscanf(line,"%f",&code)==1){
179 line=strchr(line,',');
180 if(line)line++;
181 if(code<min)min=code;
182 if(code>max)max=code;
183 }
184
185 line=setup_line(in);
186 }
187
188 if(bins<1){
189 if((int)(max-min)==min-max){
190 bins=max-min;
191 }else{
192 bins=25;
193 }
194 }
195
196 printf("\r \r");
197 printf("Minimum scalar value: %f\n",min);
198 printf("Maximum scalar value: %f\n",max);
199
200 if(argv[2]){
201
202 printf("\n counting hits into %ld bins...\n",bins+1);
203 countarray=calloc(bins+1,sizeof(long));
204
205 rewind(in);
206 line=setup_line(in);
207 while(line){
208 float code;
209 lines--;
210 if(!(lines&0xff))spinnit("counting distribution. lines so far...",lines);
211
212 while(line && sscanf(line,"%f",&code)==1){
213 line=strchr(line,',');
214 if(line)line++;
215
216 code-=min;
217 code/=(max-min);
218 code*=bins;
219 countarray[(int)rint(code)]++;
220 total++;
221 }
222
223 line=setup_line(in);
224 }
225
226 /* make a pretty graph */
227 {
228 long maxcount=0,i,j;
229 for(i=0;i<bins+1;i++)
230 if(countarray[i]>maxcount)maxcount=countarray[i];
231
232 printf("\r \r");
233 printf("Total scalars: %ld\n",total);
234 for(i=0;i<bins+1;i++){
235 int stars=rint(50./maxcount*countarray[i]);
236 printf("%08f (%8ld) |",(max-min)/bins*i+min,countarray[i]);
237 for(j=0;j<stars;j++)printf("*");
238 printf("\n");
239 }
240 }
241 }
242
243 fclose(in);
244
245 }
246 printf("\nDone.\n");
247 exit(0);
248 }
249