int read_vocab(char* vocab_filename,
int verbosity,
struct idngram_hash_table* vocabulary,
int M
)
{
FILE *vocab_file;
int vocab_size;
char temp_word[MAX_WORD_LENGTH];
char temp_word2[MAX_WORD_LENGTH];
vocab_size = 0;
vocab_file = rr_iopen(vocab_filename);
pc_message(verbosity,2,"Reading vocabulary... \n");
while (fgets (temp_word, sizeof(temp_word),vocab_file)) {
if (strncmp(temp_word,"##",2)==0) continue;
sscanf (temp_word, "%s ",temp_word2);
/* printf("hey hey %s %d\n ", temp_word2, idngram_hash(temp_word2,M));*/
/* Check for repeated words in the vocabulary */
if (index2(vocabulary,temp_word2) != 0)
warn_on_repeated_words(temp_word2);
warn_on_wrong_vocab_comments(temp_word);
vocab_size++;
/* printf("%s %d\n ", temp_word2, idngram_hash(temp_word2,M));*/
add_to_idngram_hashtable(vocabulary,idngram_hash(temp_word2,M),temp_word2,vocab_size);
if(vocab_size == M){
quit(-1, "Number of entries reached the size of the hash. Run the program again with a larger has size -hash \n");
}
}
if (vocab_size > MAX_VOCAB_SIZE)
fprintf(stderr,"text2idngram : vocab_size %d\n is larger than %d\n",vocab_size,MAX_VOCAB_SIZE);
return 0;
}
wordid_t index2(struct idngram_hash_table *vocab,
char *word) {
unsigned long chain;
struct idngram_node *chain_pos;
chain = idngram_hash( word, vocab->size );
if ( chain >= vocab->size ) {
fprintf( stderr, "WARNING : invalid hash address\n" );
fprintf( stderr, "%s ignored\n", word );
return(0);
}
chain_pos = vocab->chain[chain];
while (chain_pos->next != NULL) {
if (strcmp(word,chain_pos->next->word) ) {
fflush(stderr);
chain_pos = chain_pos->next;
}else
return (chain_pos->next->ind);
}
return (0);
}
int main(int argc, char *argv[]) {
int verbosity;
int vocab_size;
FILE *vocab_file;
int buffer_size;
flag write_ascii;
int max_files;
int number_of_tempfiles;
char *vocab_filename;
char *idngram_filename;
char temp_word[MAX_WORD_LENGTH];
char temp_word2[MAX_WORD_LENGTH];
char temp_word3[MAX_WORD_LENGTH];
flag contains_unks;
int position_in_buffer;
FILE *outfile;
FILE *tempfile;
FILE *non_unk_fp;
ngram_rec *buffer;
flag same_ngram;
int i;
int j;
int fof_size;
int size_of_rec;
char temp_directory[1000];
char *temp_file_ext;
/* Vocab hash table things */
struct idngram_hash_table vocabulary;
unsigned long hash_size;
unsigned long M;
wordid_t *current_ngram;
int current_count;
wordid_t *sort_ngram;
int sort_count;
/* Process command line */
report_version(&argc,argv);
if (argc == 1 || pc_flagarg(&argc, argv,"-help")) {
/* Display help message */
help_message();
exit(1);
}
n = pc_intarg( &argc, argv, "-n",DEFAULT_N);
hash_size = pc_intarg( &argc, argv, "-hash",DEFAULT_HASH_SIZE);
buffer_size = pc_intarg( &argc, argv, "-buffer",STD_MEM);
write_ascii = pc_flagarg(&argc,argv,"-write_ascii");
verbosity = pc_intarg(&argc,argv,"-verbosity",DEFAULT_VERBOSITY);
max_files = pc_intarg( &argc, argv, "-files",DEFAULT_MAX_FILES);
fof_size = pc_intarg(&argc,argv,"-fof_size",10);
vocab_filename = salloc(pc_stringarg( &argc, argv, "-vocab", "" ));
idngram_filename = salloc(pc_stringarg( &argc, argv, "-idngram", "" ));
if (!strcmp("",vocab_filename))
quit(-1,"Error : Must specify a vocabulary file.\n");
if (!strcmp("",idngram_filename))
quit(-1,"text2idngram : Error : Must specify idngram file.\n");
if (pc_flagarg(&argc,argv,"-compress"))
temp_file_ext = salloc(".Z");
else {
if (pc_flagarg(&argc,argv,"-gzip"))
temp_file_ext = salloc(".gz");
else
temp_file_ext = salloc("");
}
strcpy(temp_directory, "cmuclmtk-XXXXXX");
if (mkdtemp(temp_directory) == NULL) {
quit(-1, "Failed to create temporary folder: %s\n", strerror(errno));
}
pc_report_unk_args(&argc,argv,verbosity);
outfile = rr_fopen(idngram_filename,"wb");
pc_message(verbosity,2,"Vocab : %s\n",vocab_filename);
pc_message(verbosity,2,"Output idngram : %s\n",idngram_filename);
pc_message(verbosity,2,"Buffer size : %d\n",buffer_size);
pc_message(verbosity,2,"Hash table size : %d\n",hash_size);
pc_message(verbosity,2,"Max open files : %d\n",max_files);
pc_message(verbosity,2,"n : %d\n",n);
pc_message(verbosity,2,"FOF size : %d\n",fof_size);
size_of_rec = (sizeof(wordid_t) * n) + 16 - (( n* sizeof(wordid_t)) % 16);
buffer_size *= (1000000/((sizeof(ngram_rec) + size_of_rec)));
fprintf(stderr,"buffer size = %d\n",buffer_size);
/* Allocate memory for hash table */
fprintf(stderr,"Initialising hash table...\n");
M = nearest_prime(hash_size);
new_idngram_hashtable(&vocabulary,M);
/* Read in the vocabulary */
vocab_size = 0;
vocab_file = rr_iopen(vocab_filename);
pc_message(verbosity,2,"Reading vocabulary...\n");
while (fgets (temp_word, sizeof(temp_word),vocab_file)) {
if (strncmp(temp_word,"##",2)==0) continue;
sscanf (temp_word, "%s ",temp_word2);
/* Check for vocabulary order */
if (vocab_size > 0 && strcmp(temp_word2,temp_word3)<0)
quit(-1,"wngram2idngram : Error : Vocabulary is not alphabetically ordered.\n");
/* Check for repeated words in the vocabulary */
if (index2(&vocabulary,temp_word2) != 0)
warn_on_repeated_words(temp_word);
warn_on_wrong_vocab_comments(temp_word);
vocab_size++;
add_to_idngram_hashtable(&vocabulary,idngram_hash(temp_word2,M),temp_word2,vocab_size);
strcpy(temp_word3,temp_word2);
}
if (vocab_size > MAX_VOCAB_SIZE)
quit(-1,"Error : Vocabulary size exceeds maximum.\n");
pc_message(verbosity,2,"Allocating memory for the buffer...\n");
buffer=(ngram_rec *) rr_malloc((buffer_size+1)*sizeof(ngram_rec));
for (i=0;i<=buffer_size;i++)
buffer[i].word = (wordid_t *) rr_malloc(n*sizeof(wordid_t));
/* Open the "non-OOV" tempfile */
sprintf(temp_word, "%s/1%s", temp_directory, temp_file_ext);
non_unk_fp = rr_fopen(temp_word,"w");
pc_message(verbosity,2,"Writing non-OOV counts to temporary file %s\n",
temp_word);
number_of_tempfiles = 1;
current_ngram = (wordid_t *) rr_malloc(n*sizeof(wordid_t));
sort_ngram = (wordid_t *) rr_malloc(n*sizeof(wordid_t));
/* Read text into buffer */
position_in_buffer = 0;
while (!rr_feof(stdin)) {
for (i=0;i<=n-1;i++) {
get_word(stdin,temp_word);
current_ngram[i]=index2(&vocabulary,temp_word);
}
if (scanf("%d",¤t_count) != 1)
if (!rr_feof(stdin))
quit(-1,"Error reading n-gram count from stdin.\n");
if (!rr_feof(stdin)) {
contains_unks = 0;
for (i=0;i<=n-1;i++) {
if (!current_ngram[i])
contains_unks = 1;
}
if (contains_unks) {
/* Write to buffer */
position_in_buffer++;
if (position_in_buffer >= buffer_size) {
/* Sort buffer */
pc_message(verbosity,2,
"Sorting n-grams which include an OOV word...\n");
qsort((void*) buffer,(size_t) position_in_buffer,
sizeof(ngram_rec),compare_ngrams2);
pc_message(verbosity,2,"Done.\n");
/* Write buffer to temporary file */
number_of_tempfiles++;
sprintf(temp_word,"%s/%hu%s", temp_directory,
number_of_tempfiles,temp_file_ext);
pc_message(verbosity,2,
"Writing sorted OOV-counts buffer to temporary file %s\n",
temp_word);
tempfile = rr_fopen(temp_word,"w");
for (i=0;i<=n-1;i++)
sort_ngram[i] = buffer[0].word[i];
sort_count = buffer[0].count;
for (i=0;i<=position_in_buffer-2;i++) {
same_ngram = 1;
for (j=n-1;j>=0;j--) {
if (buffer[i].word[j] != sort_ngram[j]) {
same_ngram = 0;
j = -1;
}
}
if (same_ngram)
sort_count += buffer[i].count;
else {
for (j=0;j<=n-1;j++) {
rr_fwrite((char*)&sort_ngram[j],sizeof(wordid_t),1,
tempfile,"temporary n-gram ids");
sort_ngram[j] = buffer[i].word[j];
}
rr_fwrite((char*)&sort_count,sizeof(int),1,tempfile,
"temporary n-gram counts");
sort_count = buffer[i].count;
}
}
for (j=0;j<=n-1;j++)
rr_fwrite((char*)&sort_ngram[j],sizeof(wordid_t),1,
tempfile,"temporary n-gram ids");
rr_fwrite((char*)&sort_count,sizeof(int),1,tempfile,
"temporary n-gram counts");
rr_oclose(tempfile);
position_in_buffer = 1;
}
for (i=0;i<=n-1;i++)
buffer[position_in_buffer-1].word[i] = current_ngram[i];
buffer[position_in_buffer-1].count = current_count;
}else {
/* Write to temporary file */
for (i=0;i<=n-1;i++)
rr_fwrite((char*)¤t_ngram[i],sizeof(wordid_t),1,
non_unk_fp,"temporary n-gram ids");
rr_fwrite((char*)¤t_count,sizeof(int),1,non_unk_fp,
"temporary n-gram counts");
}
}
}
if (position_in_buffer > 0) {
/* Only do this bit if we have actually seen some OOVs */
/* Sort final buffer */
pc_message(verbosity,2,"Sorting final buffer...\n");
qsort((void*) buffer,(size_t) position_in_buffer,
sizeof(ngram_rec),compare_ngrams2);
/* Write final buffer */
number_of_tempfiles++;
sprintf(temp_word,"%s/%hu%s", temp_directory,
number_of_tempfiles,temp_file_ext);
pc_message(verbosity,2,"Writing sorted buffer to temporary file %s\n", temp_word);
tempfile = rr_fopen(temp_word,"w");
for (i=0;i<=n-1;i++)
sort_ngram[i] = buffer[0].word[i];
sort_count = buffer[0].count;
for (i=1;i<=position_in_buffer-1;i++) {
same_ngram = 1;
for (j=n-1;j>=0;j--) {
if (buffer[i].word[j] != sort_ngram[j]) {
same_ngram = 0;
j = -1;
}
}
if (same_ngram)
sort_count += buffer[i].count;
else {
for (j=0;j<=n-1;j++) {
rr_fwrite((char*)&sort_ngram[j],sizeof(wordid_t),1,
tempfile,"temporary n-gram ids");
sort_ngram[j] = buffer[i].word[j];
}
rr_fwrite((char*)&sort_count,sizeof(int),1,tempfile,
"temporary n-gram counts");
sort_count = buffer[i].count;
}
}
for (j=0;j<=n-1;j++)
rr_fwrite((char*)&sort_ngram[j],sizeof(wordid_t),1,
tempfile,"temporary n-gram ids");
rr_fwrite((char*)&sort_count,sizeof(int),1,tempfile,
"temporary n-gram counts");
fclose(tempfile);
}
/* Merge the temporary files, and output the result */
fclose(non_unk_fp);
pc_message(verbosity,2,"Merging temporary files...\n");
merge_idngramfiles(1,
number_of_tempfiles,
temp_directory,
temp_file_ext,
max_files,
outfile,
write_ascii,
fof_size,
n);
fclose(outfile);
rmdir(temp_directory);
pc_message(verbosity,0,"wngram2idngram : Done.\n");
return 0;
}
