/*
    mmorph, MULTEXT morphology tool
    Version 2.3, October 1995
    Copyright (c) 1994,1995 ISSCO/SUISSETRA, Geneva, Switzerland
    Dominique Petitpierre, <petitp@divsun.unige.ch>
*/
/*
    tbl.c

    procedure to lookup words in the database
    input and output are in Multext record/field format
    should be changed to use an API
*/

#include <ctype.h>
#include "user.h"

/* max number of relevant fields in a line */
#define MAX_FIELD	4

#define MAXSTRLNG 128
#define LINE_CHUNK 128
static char *line = NULL;
static int  max_line_size;

#define CLASS_CHUNK 10
static char **class_table = NULL;
static int  max_class_table_size;
static int  class_table_size;

/* save a slot for end of last field */
static char *field[MAX_FIELD + 1];
static int  field_card;
static t_boolean begin_sentence;

static void
print_tbl(field_no)
int         field_no;

{
    int         i;
    int         last_field;

    /* copy the line as is, no lookup, no annotations */
    print_out("%s", field[0]);
    last_field = MAX(field_no, field_card - 1);
    for (i = 1; i <= last_field; i++)
	print_out("%c%s", FIELD_SEP, field[i]);
    print_out("%c", RECORD_SEP);
}

/** Fonction de comparaison utilise par qsort et bsearch */
static int
#if defined(__STDC__)
class_cmp(const T_PTR class1, const T_PTR class2)
#else
class_cmp(class1, class2)
T_PTR       class1;
T_PTR       class2;

#endif

{
    return strcmp(*((T_PTR *) class1), *((T_PTR *) class2));
}


/*
    Read  a line up to a newline and divide it in MAXFIELD fields
    separated by TABs.
    return in segment the form in field form_field with its size
    in segment_size.
 */
static      t_boolean
get_tbl_segment(infile, segment, segment_size)
FILE       *infile;
char      **segment;	/* output */
int        *segment_size;	/* output */

{

    char       *s;
    char       *last_s;
    int         c;
    int         field_no;
    t_boolean   irrelevant;

    /*
       read a line and extract the word to lookup. fgets is not used because
       it takes the newline in.
    */
    begin_sentence = FALSE;	/* reset sentence boundary flag */
    do {
	s = line;
	field[0] = line;
	field_no = 1;
	last_s = line + max_line_size;
	while ((c = getc(infile)) != EOF
	       && (c != (int) RECORD_SEP)) {
	    if (c == (int) FIELD_SEP && field_no < MAX_FIELD) {
		*s++ = NUL_CHAR;
		field[field_no++] = s;
	    }
	    else
		*s++ = (char) c;
	    if (s >= last_s) {
		max_line_size += LINE_CHUNK;
		MY_REALLOC(line, max_line_size, char);
		/* line address might have changed */
		s = line + max_line_size - LINE_CHUNK;
		last_s = line + max_line_size;
	    }
	}
	*s = NUL_CHAR;
	/* fill the table, just in case */
	field_card = field_no;
	for (; field_no <= MAX_FIELD; field_no++)
	    field[field_no] = s;
	/*
	   no lookup occurs if - the form field is empty, - the morph field
	   is not empty and we are not extending nor overwriting,
	    - the class field is not in the class set
	*/
	irrelevant = (*field[morph_field] && !(extend_morph_field
					       || overwrite_morph_field))
	    || !*field[form_field]
	    || (bsearch((T_PTR) (field + class_field),
			(T_PTR) class_table,
			(size_t) class_table_size,
			sizeof(char *), class_cmp)
		== NULL);
	if (irrelevant && !(c == EOF && s == line)) {
	    begin_sentence = !strcmp(field[class_field],
				     begin_sentence_class);
	    print_tbl(0);
	}
    } while (irrelevant && c != EOF);
    if (c == EOF && s != line)
	print_warning("input file does not terminate with a newline");
    if (irrelevant)
	return (FALSE);	/* EOF */
    else {
	*segment = field[form_field];
	*segment_size = field[form_field + 1] - field[form_field];
	return (TRUE);
    }
}

/*
    copy a string, converting uppercase to lowercase.
    Use the correspondance specified by LC_CTYPE environment variable
*/
t_boolean
fold_case(source, target)
unsigned char *source;
unsigned char *target;

{
    register unsigned char *s;
    register unsigned char *t;
    t_boolean   folded;

    folded = FALSE;
    for (s = source, t = target; *s;) {
	if (isupper((int) *s)) {
	    folded = TRUE;
	    *t++ = (unsigned char) tolower((int) *s++);
	}
	else
	    *t++ = *s++;
    }
    *t = (unsigned char) NUL_CHAR;
    return (folded);
}

void
lookup_tbl(infile, lookup_classes)
FILE       *infile;
char       *lookup_classes;

{
    char       *segment;
    t_letter    surface_lex[MAXSTRLNG];
    char       *p;
    int         segment_size;
    int         i;
    t_boolean   found;
    t_boolean   folded;
    t_boolean   extend;

/* separator are comma, backslash, tab and space (and newline). Comma and
   space should not be there strictly, but are unlikely to appear in a
   class name
*/
#define SEPARATORS " |,\\\t\n"

    if (class_table == NULL) {
	max_class_table_size = CLASS_CHUNK;
	MY_CALLOC(class_table, max_class_table_size, char *);
    }
    if (line == NULL) {
	max_line_size = LINE_CHUNK;
	MY_CALLOC(line, max_line_size, char);
    }
    /* parse and store the lookup classes in a table, sorted */
    p = strtok(lookup_classes, SEPARATORS);
    class_table_size = 0;
    while (p != NULL) {
	if (class_table_size >= max_class_table_size) {
	    max_class_table_size += CLASS_CHUNK;
	    MY_REALLOC(class_table, max_class_table_size, char *);
	}
	class_table[class_table_size++] = p;
	p = strtok((char *) NULL, SEPARATORS);
    }
    if (class_table_size == 0)
	fatal_error("lookup classes are not correctly specified: \"%s\"",
		    lookup_classes);
    qsort((T_PTR) class_table, (size_t) class_table_size, sizeof(char *),
          class_cmp);
    /* check each line for lookup */
    while (get_tbl_segment(infile, &segment, &segment_size)) {
	if (segment_size >= MAXSTRLNG) {
	    fatal_error("word too long (max %d):\n %s",
			MAXSTRLNG - 1, segment);
	}
	for (i = 0; i < morph_field; i++)
	    print_out("%s%c", field[i], FIELD_SEP);
	extend = (*field[morph_field]  && ! overwrite_morph_field);
	if (extend)
	    print_out("%s", field[morph_field]);
	if (fold_case_always) {
	    folded = fold_case((unsigned char *) segment,
			       (unsigned char *) surface_lex);
	    found = map_letter((t_str) surface_lex, surface_lex,
			       Surface_Letter)
		&& db_forms_lookup_tbl(surface_lex, extend, folded);
	}
	else {
	    folded = FALSE;
	    found = map_letter(segment, surface_lex, Surface_Letter)
		&& db_forms_lookup_tbl(surface_lex, extend, folded);
	}
	if (!fold_case_always
	    && ((begin_sentence && isupper((int) ((unsigned char) *segment)))
		|| (fold_case_fallback && !found))) {
	    folded = fold_case((unsigned char *) segment,
			       (unsigned char *) surface_lex);
	    if (folded)
		found |= (map_letter((char *) surface_lex, surface_lex,
				     Surface_Letter)
			  && db_forms_lookup_tbl(surface_lex,
						 (extend || found),
						 TRUE));
	}
	if (!found && !extend && mark_unknown)
	    print_out("%s", UNKNOWN_MORPH);
	for (i++; i < field_card; i++)
	    print_out("%c%s", FIELD_SEP, field[i]);
	print_out("%c", RECORD_SEP);
	if (want_flush)
	    flush_out();
    }
}