int stats3_decompress_bits(range_coder *c,unsigned char m[1025],int *len_out)
{
int i;
*len_out=0;
/* Check if message is encoded naturally */
int notRawASCII=range_decode_equiprobable(c,2);
if (notRawASCII==0) {
/* raw bytes -- copy from input to output */
// printf("decoding raw bytes: bits_used=%d\n",c->bits_used);
for(i=0;c->bit_stream[i]&&i<1024&&i<(c->bit_stream_length>>3);i++) {
m[i]=c->bit_stream[i];
// printf("%d 0x%02x\n",i,c->bit_stream[i]);
}
// printf("%d 0x%02x\n",i,c->bit_stream[i]);
m[i]=0;
*len_out=i;
return 0;
}
int notPackedASCII=range_decode_symbol(c,&probPackedASCII,2);
int encodedLength=decodeLength(c);
for(i=0;i<encodedLength;i++) m[i]='?'; m[i]=0;
*len_out=encodedLength;
if (notPackedASCII==0) {
/* packed ASCII -- copy from input to output */
// printf("decoding packed ASCII\n");
decodePackedASCII(c,(char *)m,encodedLength);
return 0;
}
unsigned char nonAlphaValues[1024];
int nonAlphaPositions[1024];
int nonAlphaCount=0;
decodeNonAlpha(c,nonAlphaPositions,nonAlphaValues,&nonAlphaCount,encodedLength);
int alphaCount=(*len_out)-nonAlphaCount;
// printf("message contains %d non-alpha characters, %d alpha chars.\n",nonAlphaCount,alphaCount);
unsigned char lowerCaseAlphaChars[1025];
decodeLCAlphaSpace(c,lowerCaseAlphaChars,alphaCount);
lowerCaseAlphaChars[alphaCount]=0;
decodeCaseModel1(c,lowerCaseAlphaChars);
mungeCase((char *)lowerCaseAlphaChars);
/* reintegrate alpha and non-alpha characters */
int nonAlphaPointer=0;
int alphaPointer=0;
for(i=0;i<(*len_out);i++)
{
if (nonAlphaPointer<nonAlphaCount
&&nonAlphaPositions[nonAlphaPointer]==i) {
m[i]=nonAlphaValues[nonAlphaPointer++];
} else {
m[i]=lowerCaseAlphaChars[alphaPointer++];
}
}
m[i]=0;
return 0;
}
int FUNC(CaseModel1)(range_coder *c,unsigned short *line,int len,stats_handle *h)
{
int wordNumber=0;
int wordPosn=-1;
int lastWordInitialCase=0;
int lastWordInitialCase2=0;
int lastCase=0;
int i;
// printf("caps eligble chars: ");
for(i=0;i<len;i++) {
int wordChar=charInWord(line[i]);
if (!wordChar) {
wordPosn=-1; lastCase=0;
} else {
if (isalpha(line[i])) {
if (wordPosn<0) wordNumber++;
wordPosn++;
int upper=-1;
int caseEnd=0;
/* note if end of word (which includes end of message,
implicitly detected here by finding null at end of string */
if (!charInWord(line[i+1])) caseEnd=1;
if (wordPosn==0) {
/* first letter of word, so can only use 1st-order model */
unsigned int frequencies[1]={h->caseposn1[0][0]};
if (i==0) frequencies[0]=h->casestartofmessage[0][0];
else if (wordNumber>1&&wordPosn==0) {
/* start of word, so use model that considers initial case of
previous word */
frequencies[0]=h->casestartofword2[lastWordInitialCase][0];
if (wordNumber>2)
frequencies[0]=
h->casestartofword3[lastWordInitialCase2][lastWordInitialCase][0];
if (0)
printf("last word began with case=%d, p_lower=%f\n",
lastWordInitialCase,
(frequencies[0]*1.0)/0x1000000
);
}
if (0) printf("case of first letter of word/message @ %d: p=%f\n",
i,(frequencies[0]*1.0)/0x1000000);
#ifdef ENCODING
upper=isupper(line[i]);
range_encode_symbol(c,frequencies,2,upper);
#else
upper=range_decode_symbol(c,frequencies,2);
#endif
} else {
/* subsequent letter, so can use case of previous letter in model */
if (wordPosn>79) wordPosn=79;
if (0) {
printf("case of first letter of word/message @ %d.%d: p=%f\n",
i,wordPosn,
(h->caseposn2[lastCase][wordPosn][0]*1.0)/0x1000000);
printf(" lastCase=%d, wordPosn=%d\n",lastCase,wordPosn);
}
int pos=wordPosn;
while ((!h->caseposn2[lastCase][pos][0])&&pos) pos--;
#ifdef ENCODING
upper=isupper(line[i]);
range_encode_symbol(c,h->caseposn2[lastCase][pos],2,upper);
#else
upper=range_decode_symbol(c,h->caseposn2[lastCase][pos],2);
#endif
}
if (upper==1) line[i]=toupper(line[i]);
if (isupper(line[i])) lastCase=1; else lastCase=0;
if (wordPosn==0) {
lastWordInitialCase2=lastWordInitialCase;
lastWordInitialCase=lastCase;
}
else if (upper==-1) {
fprintf(stderr,"%s(): character processed without determining case.\n",
__FUNCTION__);
exit(-1);
}
}
}
}
return 0;
}
