/*****************************************************************************
* Reads frequency attributes into the pre-allocated freqs array.
****************************************************************************/
static void parse_freq_attrs(PS_T *ps, const char* tag, const xmlChar **attrs) {
int i, ncore, seen, *idx;
char *end_ptr;
double value, sum;
RBNODE_T *node;
bool seen_bad;
ncore = rbtree_size(ps->alph_ids);
// initilize the freqs array
if (ps->freqs == NULL) ps->freqs = mm_malloc(sizeof(double) * ncore);
// reset freqs array;
for (i = 0; i < ncore; i++) ps->freqs[i] = -1;
seen = 0;
seen_bad = false;
sum = 0.0;
// iterate over attributes
for (i = 0; attrs[i] != NULL; i += 2) {
idx = (int*)rbtree_get(ps->alph_ids, attrs[i]);
if (idx != NULL) {
assert(*idx < ncore);
if (ps->freqs[*idx] != -1) {
dreme_attr_parse_error(ps, PARSE_ATTR_DUPLICATE, tag, (const char*)attrs[i], NULL);
continue;
}
seen++;
errno = 0; // reset because we're about to check it
value = strtod((const char*)attrs[i+1], &end_ptr);
// allow out of range values, mainly because freqs can be very close to zero
if (end_ptr == (const char*)attrs[i+1] || (errno && errno != ERANGE) || value < 0 || value > 1) {
dreme_attr_parse_error(ps, PARSE_ATTR_BAD_VALUE, tag, (const char*)attrs[i], (const char*)attrs[i+1]);
ps->freqs[*idx] = 0; // mark frequence as seen, even though it's bad
seen_bad = true;
continue;
}
ps->freqs[*idx] = value;
sum += value;
}
}
// check we got everthing
if (seen < ncore) {
// identify what we're missing
for (node = rbtree_first(ps->alph_ids); node != NULL; node = rbtree_next(node)) {
idx = (int*)rbtree_value(node);
if (ps->freqs[*idx] == -1) {
dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, tag, (char*)rbtree_key(node), NULL);
}
}
} else if (!seen_bad) {
// check the frequencies sum to 1
double delta = sum - 1;
delta = (delta < 0 ? -delta : delta);
if (delta > (0.001 * ncore)) {
// dreme writes background probabilities to 3 decimal places so assuming
// the error on each is at maximum 0.001 then the total error for the
// sum must be less than or equal to 0.004
error(ps, "Probabilities of %s do not sum to 1, got %g .\n", tag, sum);
}
}
}
/*****************************************************************************
* dreme > model > negatives
*
* name the name of the negative dataset
* count the number of sequences in the negative dataset
* from the source of the negative dataset (eg shuffled positives)
* file the file containing the negative dataset (optional)
* last_mod_date the last modified date of the file (optional)
****************************************************************************/
static void start_ele_negatives(PS_T *ps, const xmlChar **attrs) {
char *name, *file, *lastmod;
long count;
int from;
file = NULL;
lastmod = NULL;
char* from_options[2] = {"file", "shuffled"};
int from_values[2] = {DREME_NEG_FILE, DREME_NEG_SHUFFLED};
MULTI_T from_multi = {.count = 2, .options = from_options, .outputs = from_values, .target = &(from)};
char* names[5] = {"count", "file", "from", "last_mod_date", "name"};
int (*parsers[5])(char*, void*) = {ld_long, ld_str, ld_multi, ld_str, ld_str};
void *data[5] = {&count, &file, &from_multi, &lastmod, &name};
BOOLEAN_T required[5] = {TRUE, FALSE, TRUE, FALSE, TRUE};
BOOLEAN_T done[5];
parse_attributes(dreme_attr_parse_error, ps, "negatives", attrs, 5, names, parsers, data, required, done);
if (ps->state != PS_ERROR && from == DREME_NEG_FILE) {
if (file == NULL) {
dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "negatives", "file", NULL);
}
if (lastmod == NULL) {
dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "negatives", "last_mod_date", NULL);
}
}
if (ps->callbacks->handle_negatives && ps->state != PS_ERROR) {
ps->callbacks->handle_negatives(ps->user_data, name, count, (DREME_NEG_EN)from, file, lastmod);
}
}
/*****************************************************************************
* dreme > model > background
*
* type is the alphabet DNA or RNA
* A frequency of A
* C frequency of C
* G frequency of G
* T/U frequency of T or U depending on type
* from from the negative dataset or a background file
* file the background file (optional)
* last_mod_date the last modified date of the background file (optional)
****************************************************************************/
static void start_ele_background(PS_T *ps, const xmlChar **attrs) {
int type, from;
double A, C, G, T;
char *file, *lastmod;
// set reasonable defaults
type = DREME_BG_FROM_DATASET;
from = DREME_ALPH_DNA;
file = NULL;
lastmod = NULL;
char* type_options[2] = {"dna", "rna"};
int type_values[2] = {DREME_ALPH_DNA, DREME_ALPH_RNA};
MULTI_T type_multi = {.count = 2, .options = type_options,
.outputs = type_values, .target = &(type)};
char* from_options[2] = {"dataset", "file"};
int from_values[2] = {DREME_BG_FROM_DATASET, DREME_BG_FROM_FILE};
MULTI_T from_multi = {.count = 2, .options = from_options,
.outputs = from_values, .target = &(from)};
char* names[9] = {"A", "C", "G", "T", "U", "file", "from",
"last_mod_date", "type"};
int (*parsers[9])(char*, void*) = {ld_pvalue, ld_pvalue, ld_pvalue,
ld_pvalue, ld_pvalue, ld_str, ld_multi, ld_str, ld_multi};
void *data[9] = {&A, &C, &G, &T, &T, &file, &from_multi, &lastmod, &type_multi};
BOOLEAN_T required[9] = {TRUE, TRUE, TRUE, FALSE, FALSE, FALSE, TRUE, FALSE, TRUE};
BOOLEAN_T done[9];
parse_attributes(dreme_attr_parse_error, ps, "background", attrs, 9, names, parsers, data, required, done);
if (type == DREME_ALPH_DNA) {
if (!done[3]) dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "background", "T", NULL);
if (done[4]) error(ps, "Attribute background::U should not exist because alphabet is DNA.\n");
} else {
if (!done[4]) dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "background", "U", NULL);
if (done[3]) error(ps, "Attribute background::T should not exist because alphabet is RNA.\n");
}
if (from == DREME_BG_FROM_FILE) {
if (!done[5]) dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "background", "file", NULL);
if (!done[7]) dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "background", "last_mod_date", NULL);
}
if (ps->state != PS_ERROR) {
// check sums to 1
double delta = A + C + G + T - 1;
delta = (delta < 0 ? -delta : delta);
if (delta > 0.004) {
// dreme writes background probabilities to 3 decimal places so assuming
// the error on each is at maximum 0.001 then the total error for the
// sum must be less than or equal to 0.004
error(ps, "Probabilities of background do not sum to 1, got %g .\n",
A + C + G + T);
}
}
if (ps->state != PS_ERROR) {
ps->alphabet = (DREME_ALPH_EN)type;
}
if (ps->callbacks->handle_background && ps->state != PS_ERROR) {
ps->callbacks->handle_background(ps->user_data, (DREME_ALPH_EN)type, A, C, G, T, from, file, lastmod);
}
}
/*****************************************************************************
* dreme > model > stop
*
* evalue the stopping evalue (returned as log10).
* count the stopping count.
* time the stopping time.
****************************************************************************/
static void start_ele_stop(PS_T *ps, const xmlChar **attrs) {
int count, time;
double log10evalue;
char* names[3] = {"count", "evalue", "time"};
int (*parsers[3])(char*, void*) = {ld_int, ld_log10_ev, ld_int};
void *data[3] = {&count, &log10evalue, &time};
BOOLEAN_T required[3] = {FALSE, FALSE, FALSE};
BOOLEAN_T done[3];
parse_attributes(dreme_attr_parse_error, ps, "stop", attrs, 3, names, parsers, data, required, done);
if (ps->callbacks->handle_stop && ps->state != PS_ERROR) {
ps->callbacks->handle_stop(ps->user_data, &log10evalue, &count, &time);
}
}
/*****************************************************************************
* dreme > model > /ngen
* the number of generations to check (or something like that).
****************************************************************************/
static void end_ele_ngen(PS_T *ps) {
int ngen;
if (ld_int(ps->characters.buffer, &ngen)) {
error(ps, "Bad value \"%s\" for ngen.\n", ps->characters.buffer);
}
if (ps->callbacks->handle_ngen && ps->state != PS_ERROR) {
ps->callbacks->handle_ngen(ps->user_data, ngen);
}
}
/*****************************************************************************
* dreme > model > background
*
* type is the alphabet DNA or RNA (optional when custom alphabet specified)
* <symbol> frequency of <symbol> from core alphabet
* from from the negative dataset or a background file
* file the background file (optional)
* last_mod_date the last modified date of the background file (optional)
****************************************************************************/
static void start_ele_background(PS_T *ps, const xmlChar **attrs) {
int type, from;
char *file, *lastmod;
// set reasonable defaults
type = DREME_BG_FROM_DATASET;
from = DREME_ALPH_DNA;
file = NULL;
lastmod = NULL;
char* type_options[2] = {"dna", "rna"};
int type_values[2] = {DREME_ALPH_DNA, DREME_ALPH_RNA};
MULTI_T type_multi = {.count = 2, .options = type_options,
.outputs = type_values, .target = &(type)};
char* from_options[2] = {"dataset", "file"};
int from_values[2] = {DREME_BG_FROM_DATASET, DREME_BG_FROM_FILE};
MULTI_T from_multi = {.count = 2, .options = from_options,
.outputs = from_values, .target = &(from)};
char* names[4] = {"file", "from", "last_mod_date", "type"};
int (*parsers[4])(char*, void*) = {ld_str, ld_multi, ld_str, ld_multi};
void *data[4] = {&file, &from_multi, &lastmod, &type_multi};
BOOLEAN_T required[4] = {FALSE, TRUE, FALSE, FALSE};
BOOLEAN_T done[4];
required[3] = !ps->seen_alphabet;
parse_attributes(dreme_attr_parse_error, ps, "background", attrs, 4, names, parsers, data, required, done);
if (from == DREME_BG_FROM_FILE) {
if (!done[0]) dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "background", "file", NULL);
if (!done[2]) dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "background", "last_mod_date", NULL);
}
// if we haven't seen the alphabet then we must define it from the type
if (!ps->seen_alphabet) {
int idx = 0;
rbtree_make(ps->alph_ids, "A", &idx); idx++;
rbtree_make(ps->alph_ids, "C", &idx); idx++;
rbtree_make(ps->alph_ids, "G", &idx); idx++;
rbtree_make(ps->alph_ids, (type == DREME_ALPH_DNA ? "T" : "U"), &idx);
}
parse_freq_attrs(ps, "background", attrs);
if (ps->callbacks->handle_background && ps->state != PS_ERROR) {
ps->callbacks->handle_background(ps->user_data, rbtree_size(ps->alph_ids), ps->freqs, from, file, lastmod);
}
}
/*****************************************************************************
* dreme > model > stop
*
* evalue the stopping evalue (returned as log10).
* count the stopping count.
* time the stopping time.
****************************************************************************/
static void start_ele_stop(PS_T *ps, const xmlChar **attrs) {
int count, time;
double log10evalue;
char* names[3] = {"count", "evalue", "time"};
int (*parsers[3])(char*, void*) = {ld_int, ld_log10_ev, ld_int};
void *data[3] = {&count, &log10evalue, &time};
BOOLEAN_T required[3] = {FALSE, FALSE, FALSE};
BOOLEAN_T done[3];
parse_attributes(dreme_attr_parse_error, ps, "stop", attrs, 3, names, parsers, data, required, done);
if (ps->callbacks->handle_stop && ps->state != PS_ERROR) {
ps->callbacks->handle_stop(ps->user_data, &log10evalue, &count, &time);
}
}
/*****************************************************************************
* dreme > model > /ngen
* the number of generations to check (or something like that).
****************************************************************************/
static void end_ele_ngen(PS_T *ps) {
int ngen;
if (ld_int(ps->characters.buffer, &ngen)) {
error(ps, "Bad value \"%s\" for ngen.\n", ps->characters.buffer);
}
if (ps->callbacks->handle_ngen && ps->state != PS_ERROR) {
ps->callbacks->handle_ngen(ps->user_data, ngen);
}
}
/*****************************************************************************
* dreme > model > negatives
*
* name the name of the negative dataset
* count the number of sequences in the negative dataset
* from the source of the negative dataset (eg shuffled positives)
* file the file containing the negative dataset (optional)
* last_mod_date the last modified date of the file (optional)
****************************************************************************/
static void start_ele_negatives(PS_T *ps, const xmlChar **attrs) {
char *name, *file, *lastmod;
long count;
int from;
file = NULL;
lastmod = NULL;
char* from_options[2] = {"file", "shuffled"};
int from_values[2] = {DREME_NEG_FILE, DREME_NEG_SHUFFLED};
MULTI_T from_multi = {.count = 2, .options = from_options, .outputs = from_values, .target = &(from)};
char* names[5] = {"count", "file", "from", "last_mod_date", "name"};
int (*parsers[5])(char*, void*) = {ld_long, ld_str, ld_multi, ld_str, ld_str};
void *data[5] = {&count, &file, &from_multi, &lastmod, &name};
BOOLEAN_T required[5] = {TRUE, FALSE, TRUE, FALSE, TRUE};
BOOLEAN_T done[5];
parse_attributes(dreme_attr_parse_error, ps, "negatives", attrs, 5, names, parsers, data, required, done);
if (ps->state != PS_ERROR && from == DREME_NEG_FILE) {
if (file == NULL) {
dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "negatives", "file", NULL);
}
if (lastmod == NULL) {
dreme_attr_parse_error(ps, PARSE_ATTR_MISSING, "negatives", "last_mod_date", NULL);
}
}
if (ps->callbacks->handle_negatives && ps->state != PS_ERROR) {
ps->callbacks->handle_negatives(ps->user_data, name, count, (DREME_NEG_EN)from, file, lastmod);
}
}
/*****************************************************************************
* DREME > model > alphabet
****************************************************************************/
static void start_ele_alphabet(PS_T *ps, const xmlChar **attrs) {
char *name;
int extends;
char* extends_options[3] = {"dna", "protein", "rna"};
int extends_values[3] = {ALPH_FLAG_EXTENDS_DNA, ALPH_FLAG_EXTENDS_PROTEIN, ALPH_FLAG_EXTENDS_RNA};
MULTI_T extends_multi = {.count = 3, .options = extends_options,
.outputs = extends_values, .target = &(extends)};
char* names[2] = {"like", "name"};
int (*parsers[2])(char*, void*) = {ld_multi, ld_str};
void *data[2] = {&extends_multi, &name};
BOOLEAN_T required[2] = {FALSE, FALSE};
BOOLEAN_T done[2];
// just so we know later on when reading the background which used to set the alphabet
ps->seen_alphabet = true;
// defaults
name = NULL;
extends = 0;
parse_attributes(dreme_attr_parse_error, ps, "alphabet", attrs, 2, names, parsers, data, required, done);
if (ps->callbacks->start_alphabet && ps->state != PS_ERROR) {
ps->callbacks->start_alphabet(ps->user_data, name, extends);
}
dreme_push_es(ps, PS_IN_ALPHABET_LETTER, ES_ONE_OR_MORE);
}
/*****************************************************************************
* DREME > model > /alphabet
****************************************************************************/
static void end_ele_alphabet(PS_T *ps) {
if (ps->callbacks->end_alphabet && ps->state != PS_ERROR) {
ps->callbacks->end_alphabet(ps->user_data);
}
}
/*****************************************************************************
* DREME > model > alphabet > letter
****************************************************************************/
static void start_ele_alphabet_letter(PS_T *ps, const xmlChar **attrs) {
char *aliases, *id, *name, *equals, symbol, complement;
int colour, idx;
char* names[7] = {"aliases", "colour", "complement", "equals", "id", "name", "symbol"};
int (*parsers[7])(char*, void*) = {ld_str, ld_hex, ld_char, ld_str, ld_str, ld_str, ld_char};
void *data[7] = {&aliases, &colour, &complement, &equals, &id, &name, &symbol};
BOOLEAN_T required[7] = {FALSE, FALSE, FALSE, FALSE, TRUE, FALSE, TRUE};
BOOLEAN_T done[7];
aliases = NULL;
name = NULL;
equals = NULL;
complement = '\0';
colour = -1;
parse_attributes(dreme_attr_parse_error, ps, "letter", attrs, 7, names, parsers, data, required, done);
if (ps->seen_ambig) {
if (equals == NULL) {
error(ps, "All core symbols must appear before any ambigous symbols.\n");
}
} else if (equals == NULL) {
idx = rbtree_size(ps->alph_ids);
rbtree_make(ps->alph_ids, id, &idx);
} else {
ps->seen_ambig = true;
}
if (ps->callbacks->handle_alphabet_letter && ps->state != PS_ERROR) {
ps->callbacks->handle_alphabet_letter(ps->user_data, id, symbol, aliases, complement, equals, name, colour);
}
}
