/***********************************************************************
* Read and set the alphabet from a parsed XML document.
***********************************************************************/
ALPH_T read_alphabet_from_xml(xmlXPathContextPtr xpath_ctxt) {
ALPH_T alph;
xmlXPathObjectPtr xpathObj = NULL;
xmlChar* property = NULL;
xpathObj = xpath_query(xpath_ctxt, "//*/alphabet");
property = read_xml_node_property(
xpathObj->nodesetval->nodeTab[0],
"length"
);
int alph_size = atoi((char *) property);
xmlFree(property);
xmlXPathFreeObject(xpathObj);
xpathObj = xpath_query(xpath_ctxt, "//*/alphabet/letter");
// The stated size of the alphabet had better match the
// number of letter elements in the alphabet.
assert(alph_size == xpathObj->nodesetval->nodeNr);
char* buffer = mm_calloc(sizeof(char), alph_size + 1);
int i = 0;
xmlNodePtr currLetterNode = NULL;
for (i = 0; i < alph_size; i++) {
currLetterNode = xpathObj->nodesetval->nodeTab[i];
if (currLetterNode == NULL) {
die("Error: missing letter %d in alphabet.\n", i);
}
// Get the letter symbol attribute
property = read_xml_node_property(currLetterNode, "symbol");
buffer[i] = *property;
xmlFree(property);
}
buffer[i] = 0;
alph = alph_type(buffer, alph_size+1);
myfree(buffer);
xmlXPathFreeObject(xpathObj);
return alph;
}
/*
* plib_name name of prior library file
* desired_beta > 0, scale \beta_{i,j} so
* \sum_{i=0}^L \lambda_i \sum_{j=1}^20 \beta_{i,j}
* has this value
* == 0, don't scale prior
* < 0, just get alphabet
*/
PriorLib *read_PriorLib(char *plib_name, double desired_beta, ALPH_T *custom_alph)
{
int i,j, line=0;
int l;
PriorLib *temp;
char input[MAXS], foo[MAXS], alphabet[MAXALPH+1], checkstr[81], *token;
double x;
FILE *fp;
ALPH_T *alph;
// tlb
fp = fopen(plib_name, "r");
if (!fp) {
fprintf(stderr, "Can't find prior library %s\n", plib_name);
exit(1);
}
token = "Alphabet="; line++;
test_value(fscanf(fp,"%s %s\n", checkstr, alphabet) == 2);
if (strcmp(checkstr, token)) {
fprintf(stderr, "Line %d of prior library file \n %s \n"
"should start with \"%s\" "
"but it starts with \"%s\".\n", line, plib_name, token, checkstr);
exit(1);
}
// determine alphabet
if (custom_alph == NULL) {
alph = alph_type(alphabet, 30);
if (alph == NULL) {
fprintf(stderr, "The partial alphabet specified in the prior library file"
" does not match a built-in alphabet and no complete alphabet was specified.\n");
exit(1);
}
} else {
int alen_core;
alen_core = strlen(alphabet);
i = 0;
if (alen_core == alph_size_core(custom_alph)) {
for (i = 0; i < alen_core; i++) {
if (!alph_is_concrete(custom_alph, alphabet[i])) break;
}
}
if (i == 0 || i < alen_core) {
fprintf(stderr, "The partial alphabet specified in the prior library file"
" does not match the complete alphabet specified.\n");
exit(1);
}
alph = alph_hold(custom_alph);
}
token = "NumDistr="; line++;
test_value(fscanf(fp,"%s %d\n", checkstr, &l) == 2);
if (strcmp(checkstr, token)) {
fprintf(stderr, "Line %d of prior library file \n %s \n"
"should start with \"%s\" "
"but it starts with \"%s\"\n.", line, plib_name, token, checkstr);
exit(1);
}
temp = alloc_PriorLib(l, alph);
if (desired_beta < 0) {
fclose(fp);
return(temp);
}
for (i = 0; i < temp->L; i++) {
// Get rid of number=
ignore_value(fscanf(fp,"%*s %*s\n"));
// Mixture
ignore_value(fscanf(fp,"%*s"));
test_value(fscanf(fp,"%lf\n", &x) == 1);
temp->Mix[i] = x;
// B (strength)
ignore_value(fscanf(fp,"%*s"));
test_value(fscanf(fp,"%lf\n", &x) == 1);
temp->B[i] = x;
// Alpha
temp->Distr[i][0] = temp->B[i];
ignore_value(fscanf(fp,"%*s"));
for (j = 1; j < alph_size_wild(alph); j++) {
test_value(fscanf(fp,"%lg", &x) == 1);
temp->Distr[i][j] = x * temp->B[i];
}
// FullUpdate
ignore_value(fscanf(fp,"%*s"));
test_value(fscanf(fp,"%d\n", &(temp->FullUpdate[i])) == 1);
// QUpdate
ignore_value(fscanf(fp,"%*s"));
test_value(fscanf(fp,"%d\n", &(temp->QUpdate[i])) == 1);
// StructID
test_value(fgets(input, MAXS, fp) != NULL);
test_value(sscanf(input,"%s",foo) == 1);
input[strlen(input)-1] = '\0';
strcpy( (temp->StructID[i]), (input + strlen(foo)) );
// Comments
test_value(fgets(input, MAXS, fp) != NULL);
test_value(sscanf(input,"%s",foo) == 1);
strcpy( (temp->Comment[i]), (input + strlen(foo)) );
}
// tlb; scale beta to desired value
if (desired_beta > 0) {
int i, j;
double beta = 0;
double scale;
for (i=0; i<temp->L; i++) {
beta += temp->Mix[i] * temp->B[i];
}
/*printf("beta = %10.6f\n", beta);*/
scale = desired_beta/beta;
for (i=0; i<temp->L; i++) {
for (j=0; j < alph_size_wild(alph); j++) {
temp->Distr[i][j] *= scale;
}
}
}
fclose(fp);
alph_release(alph);
return(temp);
}
