/**
* Append an unmarked (i.e. without INFIXMARK) morpheme to join_buff.
* join_buff is a zeroed-out buffer which has enough room for morpheme to be
* added + terminating NUL.
* Note that MT_PREFIX or MT_SUFFIX can be without an INFIX_MARK, in case
* INFIX_MARK is not defined. XXX: What about MT_MIDDLE? (not in use yet).
*
* FIXME Combining contracted words is not handled yet, because combining
* morphemes which have non-LL links to other words is not yet implemented.
*/
static void add_morpheme_unmarked(Sentence sent, char *join_buff,
const char *wm, Morpheme_type mt)
{
const char infix_mark = INFIX_MARK(sent->dict->affix_table);
const char *sm = strrchr(wm, SUBSCRIPT_MARK);
if (NULL == sm) sm = (char *)wm + strlen(wm);
if ((MT_PREFIX == mt) && (infix_mark == sm[-INFIX_MARK_L]))
strncat(join_buff, wm, sm-wm-INFIX_MARK_L);
else if ((MT_SUFFIX == mt) && (infix_mark == wm[0]))
strncat(join_buff, INFIX_MARK_L+wm, sm-wm-INFIX_MARK_L);
else if ((MT_MIDDLE == mt))
strncat(join_buff, INFIX_MARK_L+wm, sm-wm-2*INFIX_MARK_L);
else
strncat(join_buff, wm, sm-wm);
}
/**
* Split randomly.
* Return true on success.
* Return false when:
* - disabled (i.e. when doing regular language processing).
* - an error occurs (the behavior then is undefined).
* Such an error has not been observed yet.
*/
bool anysplit(Sentence sent, const char *word)
{
Dictionary afdict = sent->dict->affix_table;
anysplit_params *as;
Afdict_class * stemsubscr;
size_t stemsubscr_len;
size_t l = strlen(word);
p_list pl;
size_t pos;
int p;
int sample_point;
size_t nsplits;
size_t rndtried = 0;
size_t rndissued = 0;
size_t i;
unsigned int seed = 0;
char *prefix_string = alloca(l+2+1); /* word + ".=" + NUL */
char *suffix_string = alloca(l+1); /* word + NUL */
bool use_sampling = true;
const char infix_mark = INFIX_MARK(afdict);
if (NULL == afdict) return false;
as = afdict->anysplit;
if ((NULL == as) || (0 == as->nparts)) return false; /* Anysplit disabled */
if (0 == l)
{
prt_error("Warning: anysplit(): word length 0\n");
return false;
}
stemsubscr = AFCLASS(afdict, AFDICT_STEMSUBSCR);
stemsubscr_len = (NULL == stemsubscr->string[0]) ? 0 :
strlen(stemsubscr->string[0]);
/* Don't split morphemes again. If INFIXMARK and/or SUBSCRMARK are
* not defined in the affix file, then morphemes may get split again unless
* restricted by REGPRE/REGMID/REGSUF. */
if (word[0] == infix_mark) return true;
if ((l > stemsubscr_len) &&
(0 == strcmp(word+l-stemsubscr_len, stemsubscr->string[0])))
return true;
// seed = time(NULL)+(unsigned int)(long)&seed;
#if DEBUG_ANYSPLIT
gw = word;
#endif
nsplits = split(l, as->nparts, &as->scl[l]);
if (0 == nsplits)
{
prt_error("Warning: anysplit(): split() failed (shouldn't happen)\n");
return false;
}
if (as->altsmax >= nsplits)
{
/* Issue everything */
sample_point = -1;
use_sampling = false;
}
lgdebug(+2, "Start%s sampling: word=%s, nsplits=%zu, maxsplits=%d, "
"as->altsmin=%zu, as->altsmax=%zu\n", use_sampling ? "" : " no",
word, nsplits, as->nparts, as->altsmin, as->altsmax);
while (rndtried < nsplits && (!use_sampling || (rndissued < as->altsmin)))
{
if (use_sampling)
{
sample_point = rng_uniform(&seed, nsplits);
if (sample_point < 0) /* Cannot happen with rand_r() */
{
prt_error("Error: rng: %s\n", strerror(errno));
return false;
}
}
else
{
sample_point++;
}
lgdebug(2, "Sample: %d ", sample_point);
if (as->scl[l].p_tried[sample_point])
{
lgdebug(4, "(repeated)\n");
continue;
}
lgdebug(4, "(new)");
rndtried++;
as->scl[l].p_tried[sample_point] = true;
if (morpheme_match(sent, word, l, &as->scl[l].sp[sample_point*as->nparts]))
{
as->scl[l].p_selected[sample_point] = true;
rndissued++;
}
else
{
lgdebug(2, "\n");
}
}
lgdebug(2, "Results: word '%s' (length=%zu): %zu/%zu:\n", word, l, rndissued, nsplits);
for (i = 0; i < nsplits; i++)
{
const char **suffixes = NULL;
int num_suffixes = 0;
if (!as->scl[l].p_selected[i]) continue;
pl = &as->scl[l].sp[i*as->nparts];
pos = 0;
for (p = 0; p < as->nparts; p++)
{
if (pl[0] == (int)l) /* This is the whole word */
{
strncpy(prefix_string, &word[pos], pl[p]-pos);
prefix_string[pl[p]-pos] = '\0';
}
else
if (0 == pos) /* The first but not the only morpheme */
{
strncpy(prefix_string, &word[pos], pl[p]-pos);
prefix_string[pl[p]-pos] = '\0';
if (0 != stemsubscr->length)
strcat(prefix_string, stemsubscr->string[0]);
}
else /* 2nd and on morphemes */
{
strncpy(suffix_string, &word[pos], pl[p]-pos);
suffix_string[pl[p]-pos] = '\0';
altappend(sent, &suffixes, suffix_string);
num_suffixes++;
}
pos = pl[p];
if (pos == l) break;
}
/* Here a leading INFIX_MARK is added to the suffixes if needed. */
add_alternative(sent,
0,NULL, 1,(const char **)&prefix_string, num_suffixes,suffixes);
free(suffixes);
}
return true;
}