source: trunk/third/ispell/makedent.c @ 10334

#ifndef lint
static char Rcs_Id[] =
        "$Id: makedent.c,v 1.1.1.1 1997-09-03 21:08:12 ghudson Exp $";
#endif

/*
 * Copyright 1988, 1989, 1992, 1993, Geoff Kuenning, Granada Hills, CA
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All modifications to the source code must be clearly marked as
 *    such.  Binary redistributions based on modified source code
 *    must be clearly marked as modified versions in the documentation
 *    and/or other materials provided with the distribution.
 * 4. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgment:
 *      This product includes software developed by Geoff Kuenning and
 *      other unpaid contributors.
 * 5. The name of Geoff Kuenning may not be used to endorse or promote
 *    products derived from this software without specific prior
 *    written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY GEOFF KUENNING AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL GEOFF KUENNING OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

/*
 * $Log: not supported by cvs2svn $
 * Revision 1.45  1994/12/27  23:08:52  geoff
 * Add code to makedent to reject words that contain non-word characters.
 * This helps protect people who use ISO 8-bit characters when ispell
 * isn't configured for that option.
 *
 * Revision 1.44  1994/10/25  05:46:20  geoff
 * Fix some incorrect declarations in the lint versions of some routines.
 *
 * Revision 1.43  1994/09/16  03:32:34  geoff
 * Issue an error message for bad affix flags
 *
 * Revision 1.42  1994/02/07  04:23:43  geoff
 * Correctly identify the deformatter when changing file types
 *
 * Revision 1.41  1994/01/25  07:11:55  geoff
 * Get rid of all old RCS log lines in preparation for the 3.1 release.
 *
 */

#include "config.h"
#include "ispell.h"
#include "proto.h"
#include "msgs.h"

int             makedent P ((char * lbuf, int lbuflen, struct dent * ent));
#ifndef NO_CAPITALIZATION_SUPPORT
long            whatcap P ((ichar_t * word));
#endif
int             addvheader P ((struct dent * ent));
int             combinecaps P ((struct dent * hdr, struct dent * newent));
#ifndef NO_CAPITALIZATION_SUPPORT
static void     forcevheader P ((struct dent * hdrp, struct dent * oldp,
                  struct dent * newp));
#endif /* NO_CAPITALIZATION_SUPPORT */
static int      combine_two_entries P ((struct dent * hdrp,
                  struct dent * oldp, struct dent * newp));
static int      acoversb P ((struct dent * enta, struct dent * entb));
void            upcase P ((ichar_t * string));
void            lowcase P ((ichar_t * string));
void            chupcase P ((char * s));
static int      issubset P ((struct dent * ent1, struct dent * ent2));
static void     combineaffixes P ((struct dent * ent1, struct dent * ent2));
void            toutent P ((FILE * outfile, struct dent * hent,
                  int onlykeep));
static void     toutword P ((FILE * outfile, char * word,
                  struct dent * cent));
static void     flagout P ((FILE * outfile, int flag));
int             stringcharlen P ((char * bufp, int canonical));
int             strtoichar P ((ichar_t * out, char * in, int outlen,
                  int canonical));
int             ichartostr P ((char * out, ichar_t * in, int outlen,
                  int canonical));
ichar_t *       strtosichar P ((char * in, int canonical));
char *          ichartosstr P ((ichar_t * in, int canonical));
char *          printichar P ((int in));
#ifndef ICHAR_IS_CHAR
ichar_t *       icharcpy P ((ichar_t * out, ichar_t * in));
int             icharlen P ((ichar_t * str));
int             icharcmp P ((ichar_t * s1, ichar_t * s2));
int             icharncmp P ((ichar_t * s1, ichar_t * s2, int n));
#endif /* ICHAR_IS_CHAR */
int             findfiletype P ((char * name, int searchnames,
                  int * deformatter));

static int      has_marker;

/*
 * Fill in a directory entry, including setting the capitalization flags, and
 * allocate and initialize memory for the d->word field.  Returns -1
 * if there was trouble.  The input word must be in canonical form.
 */

int makedent (lbuf, lbuflen, d)
    char *              lbuf;
    int                 lbuflen;
    struct dent *       d;
    {
    ichar_t             ibuf[INPUTWORDLEN + MAXAFFIXLEN];
    ichar_t *           ip;
    char *              p;
    int                 bit;
    int                 len;

    /* Strip off any trailing newline */
    len = strlen (lbuf) - 1;
    if (lbuf[len] == '\n')
        lbuf[len] = '\0';

    d->next = NULL;
    /* WARNING:  flagfield might be the same as mask! See ispell.h. */
    d->flagfield = 0;
    (void) bzero ((char *) d->mask, sizeof (d->mask));
    d->flagfield |= USED;
    d->flagfield &= ~KEEP;

    p = index (lbuf, hashheader.flagmarker);
    if (p != NULL)
        *p = 0;

    /*
    ** Convert the word to an ichar_t and back;  this makes sure that
    ** it is in canonical form and thus that the length is correct.
    */
    if (strtoichar (ibuf, lbuf, INPUTWORDLEN * sizeof (ichar_t), 1)
      ||  ichartostr (lbuf, ibuf, lbuflen, 1))
        {
        (void) fprintf (stderr, WORD_TOO_LONG (lbuf));
        return (-1);
        }
    /*
    ** Make sure the word is well-formed (contains only legal characters).
    */
    for (ip = ibuf;  *ip != 0;  ip++)
        {
        if (!iswordch (*ip))
            {
            /* Boundary characters are legal as long as they're not at edges */
            if (!isboundarych (*ip)
              ||  ip == ibuf  ||  ip[1] == 0)
                {
                (void) fprintf (stderr, MAKEDENT_C_BAD_WORD_CHAR, lbuf);
                return -1;
                }
            }
        }
    len = strlen (lbuf);
#ifndef NO_CAPITALIZATION_SUPPORT
    /*
    ** Figure out the capitalization rules from the capitalization of
    ** the sample entry.
    */
    d->flagfield |= whatcap (ibuf);
#endif

    if (len > INPUTWORDLEN - 1)
        {
        (void) fprintf (stderr, WORD_TOO_LONG (lbuf));
        return (-1);
        }

    d->word = mymalloc ((unsigned) len + 1);
    if (d->word == NULL)
        {
        (void) fprintf (stderr, MAKEDENT_C_NO_WORD_SPACE, lbuf);
        return -1;
        }

    (void) strcpy (d->word, lbuf);
#ifdef NO_CAPITALIZATION_SUPPORT
    chupcase (d->word);
#else /* NO_CAPITALIZATION_SUPPORT */
    if (captype (d->flagfield) != FOLLOWCASE)
        chupcase (d->word);
#endif /* NO_CAPITALIZATION_SUPPORT */
    if (p == NULL)
        return (0);

    p++;
    while (*p != '\0'  &&  *p != '\n')
        {
        bit = CHARTOBIT ((unsigned char) *p);
        if (bit >= 0  &&  bit <= LARGESTFLAG)
            SETMASKBIT (d->mask, bit);
        else
            (void) fprintf (stderr, BAD_FLAG, (unsigned char) *p);
        p++;
        if (*p == hashheader.flagmarker)
            p++;                /* Handle old-format dictionaries too */
        }
    return (0);
    }

#ifndef NO_CAPITALIZATION_SUPPORT
/*
** Classify the capitalization of a sample entry.  Returns one of the
** four capitalization codes ANYCASE, ALLCAPS, CAPITALIZED, or FOLLOWCASE.
*/

long whatcap (word)
    register ichar_t *  word;
    {
    register ichar_t *  p;

    for (p = word;  *p;  p++)
        {
        if (mylower (*p))
            break;
        }
    if (*p == '\0')
        return ALLCAPS;
    else
        {
        for (  ;  *p;  p++)
            {
            if (myupper (*p))
                break;
            }
        if (*p == '\0')
            {
            /*
            ** No uppercase letters follow the lowercase ones.
            ** If there is more than one uppercase letter, it's
            ** "followcase". If only the first one is capitalized,
            ** it's "capitalize".  If there are no capitals
            ** at all, it's ANYCASE.
            */
            if (myupper (word[0]))
                {
                for (p = word + 1;  *p != '\0';  p++)
                    {
                    if (myupper (*p))
                        return FOLLOWCASE;
                    }
                return CAPITALIZED;
                }
            else
                return ANYCASE;
            }
        else
            return FOLLOWCASE;  /* .../lower/upper */
        }
    }

/*
** Add a variant-capitalization header to a word.  This routine may be
** called even for a followcase word that doesn't yet have a header.
**
** Returns 0 if all was ok, -1 if allocation error.
*/
int addvheader (dp)
    register struct dent *      dp;     /* Entry to update */
    {
    register struct dent *      tdent; /* Copy of entry */

    /*
    ** Add a second entry with the correct capitalization, and then make
    ** dp into a special dummy entry.
    */
    tdent = (struct dent *) mymalloc (sizeof (struct dent));
    if (tdent == NULL)
        {
        (void) fprintf (stderr, MAKEDENT_C_NO_WORD_SPACE, dp->word);
        return -1;
        }
    *tdent = *dp;
    if (captype (tdent->flagfield) != FOLLOWCASE)
        tdent->word = NULL;
    else
        {
        /* Followcase words need a copy of the capitalization */
        tdent->word = mymalloc ((unsigned int) strlen (tdent->word) + 1);
        if (tdent->word == NULL)
            {
            (void) fprintf (stderr, MAKEDENT_C_NO_WORD_SPACE, dp->word);
            myfree ((char *) tdent);
            return -1;
            }
        (void) strcpy (tdent->word, dp->word);
        }
    chupcase (dp->word);
    dp->next = tdent;
    dp->flagfield &= ~CAPTYPEMASK;
    dp->flagfield |= (ALLCAPS | MOREVARIANTS);
    return 0;
    }
#endif /* NO_CAPITALIZATION_SUPPORT */

/*
** Combine and resolve the entries describing two capitalizations of the same
** word.  This may require allocating yet more entries.
**
** Hdrp is a pointer into a hash table.  If the word covered by hdrp has
** variations, hdrp must point to the header.  Newp is a pointer to temporary
** storage, and space is malloc'ed if newp is to be kept.  The newp->word
** field must have been allocated with mymalloc, so that this routine may free
** the space if it keeps newp but not the word.
**
** Return value:  0 if the word was added, 1 if the word was combined
** with an existing entry, and -1 if trouble occurred (e.g., malloc).
** If 1 is returned, newp->word may have been be freed using myfree.
**
** Life is made much more difficult by the KEEP flag's possibilities.  We
** must ensure that a !KEEP word doesn't find its way into the personal
** dictionary as a result of this routine's actions.  However, a !KEEP
** word that has affixes must have come from the main dictionary, so it
** is acceptable to combine entries in that case (got that?).
**
** The net result of all this is a set of rules that is a bloody pain
** to figure out.  Basically, we want to choose one of the following actions:
**
**      (1) Add newp's affixes and KEEP flag to oldp, and discard newp.
**      (2) Add oldp's affixes and KEEP flag to newp, replace oldp with
**          newp, and discard newp.
#ifndef NO_CAPITALIZATION_SUPPORT
**      (3) Insert newp as a new entry in the variants list.  If there is
**          currently no variant header, this requires adding one.  Adding a
**          header splits into two sub-cases:
**
**          (3a) If oldp is ALLCAPS and the KEEP flags match, just turn it
**              into the header.
**          (3b) Otherwise, add a new entry to serve as the header.
**              To ease list linking, this is done by copying oldp into
**              the new entry, and then performing (3a).
**
**          After newp has been added as a variant, its affixes and KEEP
**          flag are OR-ed into the variant header.
#endif
**
** So how to choose which?  The default is always case (3), which adds newp
** as a new entry in the variants list.  Cases (1) and (2) are symmetrical
** except for which entry is discarded.  We can use case (1) or (2) whenever
** one entry "covers" the other.  "Covering" is defined as follows:
**
**      (4) For entries with matching capitalization types, A covers B
**          if:
**
**          (4a) B's affix flags are a subset of A's, or the KEEP flags
**               match, and
**          (4b) either the KEEP flags match, or A's KEEP flag is set.
**              (Since A has more suffixes, combining B with it won't
**              cause any extra suffixes to be added to the dictionary.)
**          (4c) If the words are FOLLOWCASE, the capitalizations match
**              exactly.
**
#ifndef NO_CAPITALIZATION_SUPPORT
**      (5) For entries with mismatched capitalization types, A covers B
**          if (4a) and (4b) are true, and:
**
**          (5a) B is ALLCAPS, or
**          (5b) A is ANYCASE, and B is CAPITALIZED.
#endif
**
** For any "hdrp" without variants, oldp is the same as hdrp.  Otherwise,
** the above tests are applied using each variant in turn for oldp.
*/
int combinecaps (hdrp, newp)
    struct dent *       hdrp;   /* Header of entry currently in dictionary */
    register struct dent *
                        newp;   /* Entry to add */
    {
    register struct dent *
                        oldp;   /* Current "oldp" entry */
#ifndef NO_CAPITALIZATION_SUPPORT
    register struct dent *
                        tdent; /* Entry we'll add to the dictionary */
#endif /* NO_CAPITALIZATION_SUPPORT */
    register int        retval = 0; /* Return value from combine_two_entries */

    /*
    ** First, see if we can combine the two entries (cases 1 and 2).  If
    ** combine_two_entries does so, it will return 1.  If it has trouble,
    ** it will return zero.
    */
    oldp = hdrp;
#ifdef NO_CAPITALIZATION_SUPPORT
    retval = combine_two_entries (hdrp, oldp, newp);
#else /* NO_CAPITALIZATION_SUPPORT */
    if ((oldp->flagfield & (CAPTYPEMASK | MOREVARIANTS))
      == (ALLCAPS | MOREVARIANTS))
        {
        while (oldp->flagfield & MOREVARIANTS)
            {
            oldp = oldp->next;
            retval = combine_two_entries (hdrp, oldp, newp);
            if (retval != 0)            /* Did we combine them? */
                break;
            }
        }
    else
        retval = combine_two_entries (hdrp, oldp, newp);
    if (retval == 0)
        {
        /*
        ** Couldn't combine the two entries.  Add a new variant.  For
        ** ease, we'll stick it right behind the header, rather than
        ** at the end of the list.
        */
        forcevheader (hdrp, oldp, newp);
        tdent = (struct dent *) mymalloc (sizeof (struct dent));
        if (tdent == NULL)
            {
            (void) fprintf (stderr, MAKEDENT_C_NO_WORD_SPACE, newp->word);
            return -1;
            }
        *tdent = *newp;
        tdent->next = hdrp->next;
        hdrp->next = tdent;
        tdent->flagfield |= (hdrp->flagfield & MOREVARIANTS);
        hdrp->flagfield |= MOREVARIANTS;
        combineaffixes (hdrp, newp);
        hdrp->flagfield |= (newp->flagfield & KEEP);
        if (captype (newp->flagfield) == FOLLOWCASE)
            tdent->word = newp->word;
        else
            {
            tdent->word = NULL;
            myfree (newp->word);                /* newp->word isn't needed */
            }
        }
#endif /* NO_CAPITALIZATION_SUPPORT */
    return retval;
    }

#ifndef NO_CAPITALIZATION_SUPPORT
/*
** The following routine implements steps 3a and 3b in the commentary
** for "combinecaps".
*/
static void forcevheader (hdrp, oldp, newp)
    register struct dent *      hdrp;
    struct dent *               oldp;
    struct dent *               newp;
    {

    if ((hdrp->flagfield & (CAPTYPEMASK | MOREVARIANTS)) == ALLCAPS
      &&  ((oldp->flagfield ^ newp->flagfield) & KEEP) == 0)
        return;                 /* Caller will set MOREVARIANTS */
    else if ((hdrp->flagfield & (CAPTYPEMASK | MOREVARIANTS))
      != (ALLCAPS | MOREVARIANTS))
        (void) addvheader (hdrp);
    }
#endif /* NO_CAPITALIZATION_SUPPORT */

/*
** This routine implements steps 4 and 5 of the commentary for "combinecaps".
**
** Returns 1 if newp can be discarded, 0 if nothing done.
*/
static int combine_two_entries (hdrp, oldp, newp)
    struct dent *       hdrp;   /* (Possible) header of variant chain */
    register struct dent *
                        oldp;   /* Pre-existing dictionary entry */
    register struct dent *
                        newp;   /* Entry to possibly combine */
    {

    if (acoversb (oldp, newp))
        {
        /* newp is superfluous.  Drop it, preserving affixes and keep flag */
        combineaffixes (oldp, newp);
        oldp->flagfield |= (newp->flagfield & KEEP);
        hdrp->flagfield |= (newp->flagfield & KEEP);
        myfree (newp->word);
        return 1;
        }
    else if (acoversb (newp, oldp))
        {
        /*
        ** oldp is superfluous.  Replace it with newp, preserving affixes and
        ** the keep flag.
        */
        combineaffixes (newp, oldp);
#ifdef NO_CAPITALIZATION_SUPPORT
        newp->flagfield |= (oldp->flagfield & KEEP);
#else /* NO_CAPITALIZATION_SUPPORT */
        newp->flagfield |= (oldp->flagfield & (KEEP | MOREVARIANTS));
#endif /* NO_CAPITALIZATION_SUPPORT */
        hdrp->flagfield |= (newp->flagfield & KEEP);
        newp->next = oldp->next;
        /*
        ** We really want to free oldp->word, but that might be part of
        ** "hashstrings".  So we'll futz around to arrange things so we can
        ** free newp->word instead.  This depends very much on the fact
        ** that both words are the same length.
        */
        if (oldp->word != NULL)
            (void) strcpy (oldp->word, newp->word);
        myfree (newp->word);    /* No longer needed */
        newp->word = oldp->word;
        *oldp = *newp;
#ifndef NO_CAPITALIZATION_SUPPORT
        /* We may need to add a header if newp is followcase */
        if (captype (newp->flagfield) == FOLLOWCASE
          &&  (hdrp->flagfield & (CAPTYPEMASK | MOREVARIANTS))
            != (ALLCAPS | MOREVARIANTS))
            (void) addvheader (hdrp);
#endif /* NO_CAPITALIZATION_SUPPORT */
        return 1;
        }
    else
        return 0;
    }

/*
** Determine if enta covers entb, according to the rules in steps 4 and 5
** of the commentary for "combinecaps".
*/
static int acoversb (enta, entb)
    register struct dent *      enta;   /* "A" in the rules */
    register struct dent *      entb;   /* "B" in the rules */
    {
    int                         subset; /* NZ if entb is a subset of enta */

    if ((subset = issubset (entb, enta)) != 0)
        {
        /* entb is a subset of enta;  thus enta might cover entb */
        if (((enta->flagfield ^ entb->flagfield) & KEEP) != 0
          &&  (enta->flagfield & KEEP) == 0)    /* Inverse of condition (4b) */
            return 0;
        }
    else
        {
        /* not a subset;  KEEP flags must match exactly (both (4a) and (4b)) */
        if (((enta->flagfield ^ entb->flagfield) & KEEP) != 0)
            return 0;
        }

    /* Rules (4a) and (4b) are satisfied;  check for capitalization match */
#ifdef NO_CAPITALIZATION_SUPPORT
#ifdef lint
    return subset;                              /* Just so it gets used */
#else /* lint */
    return 1;                                   /* All words match */
#endif /* lint */
#else /* NO_CAPITALIZATION_SUPPORT */
    if (((enta->flagfield ^ entb->flagfield) & CAPTYPEMASK) == 0)
        {
        if (captype (enta->flagfield) != FOLLOWCASE     /* Condition (4c) */
          ||  strcmp (enta->word, entb->word) == 0)
            return 1;                           /* Perfect match */
        else
            return 0;
        }
    else if (subset == 0)                       /* No flag subset, refuse */
        return 0;                               /* ..near matches */
    else if (captype (entb->flagfield) == ALLCAPS)
        return 1;
    else if (captype (enta->flagfield) == ANYCASE
      &&  captype (entb->flagfield) == CAPITALIZED)
        return 1;
    else
        return 0;
#endif /* NO_CAPITALIZATION_SUPPORT */
    }

void upcase (s)
    register ichar_t *  s;
    {

    while (*s)
        {
        *s = mytoupper (*s);
        s++;
        }
    }

void lowcase (s)
    register ichar_t *  s;
    {

    while (*s)
        {
        *s = mytolower (*s);
        s++;
        }
    }

/*
 * Upcase variant that works on normal strings.  Note that it is a lot
 * slower than the normal upcase.  The input must be in canonical form.
 */
void chupcase (s)
    char *      s;
    {
    ichar_t *   is;

    is = strtosichar (s, 1);
    upcase (is);
    (void) ichartostr (s, is, strlen (s) + 1, 1);
    }

/*
** See if one affix field is a subset of another.  Returns NZ if ent1
** is a subset of ent2.  The KEEP flag is not taken into consideration.
*/
static int issubset (ent1, ent2)
    register struct dent *      ent1;
    register struct dent *      ent2;
    {
/* The following is really testing for MASKSIZE > 1, but cpp can't do that */
#if MASKBITS > 32
    register int                flagword;

#ifdef FULLMASKSET
#define MASKMAX MASKSIZE
#else
#define MASKMAX MASKSIZE - 1
#endif /* FULLMASKSET */
    for (flagword = MASKMAX;  --flagword >= 0;  )
        {
        if ((ent1->mask[flagword] & ent2->mask[flagword])
          != ent1->mask[flagword])
            return 0;
        }
#endif /* MASKBITS > 32 */
#ifdef FULLMASKSET
    return ((ent1->mask[MASKSIZE - 1] & ent2->mask[MASKSIZE - 1])
      == ent1->mask[MASKSIZE - 1]);
#else
    if (((ent1->mask[MASKSIZE - 1] & ent2->mask[MASKSIZE - 1])
      ^ ent1->mask[MASKSIZE - 1]) & ~ALLFLAGS)
        return 0;
    else
        return 1;
#endif /* FULLMASKSET */
    }

/*
** Add ent2's affix flags to ent1.
*/
static void combineaffixes (ent1, ent2)
    register struct dent *      ent1;
    register struct dent *      ent2;
    {
/* The following is really testing for MASKSIZE > 1, but cpp can't do that */
#if MASKBITS > 32
    register int                flagword;

    if (ent1 == ent2)
        return;
    /* MASKMAX is defined in issubset, just above */
    for (flagword = MASKMAX;  --flagword >= 0;  )
        ent1->mask[flagword] |= ent2->mask[flagword];
#endif /* MASKBITS > 32 */
#ifndef FULLMASKSET
    ent1->mask[MASKSIZE - 1] |= ent2->mask[MASKSIZE - 1] & ~ALLFLAGS;
#endif
    }

/*
** Write out a dictionary entry, including capitalization variants.
** If onlykeep is true, only those variants with KEEP set will be
** written.
*/
void toutent (toutfile, hent, onlykeep)
    register FILE *     toutfile;
    struct dent *       hent;
    register int        onlykeep;
    {
#ifdef NO_CAPITALIZATION_SUPPORT
    if (!onlykeep  ||  (hent->flagfield & KEEP))
        toutword (toutfile, hent->word, hent);
#else
    register struct dent * cent;
    ichar_t             wbuf[INPUTWORDLEN + MAXAFFIXLEN];

    cent = hent;
    if (strtoichar (wbuf, cent->word, INPUTWORDLEN, 1))
        (void) fprintf (stderr, WORD_TOO_LONG (cent->word));
    for (  ;  ;  )
        {
        if (!onlykeep  ||  (cent->flagfield & KEEP))
            {
            switch (captype (cent->flagfield))
                {
                case ANYCASE:
                    lowcase (wbuf);
                    toutword (toutfile, ichartosstr (wbuf, 1), cent);
                    break;
                case ALLCAPS:
                    if ((cent->flagfield & MOREVARIANTS) == 0
                      ||  cent != hent)
                        {
                        upcase (wbuf);
                        toutword (toutfile, ichartosstr (wbuf, 1), cent);
                        }
                    break;
                case CAPITALIZED:
                    lowcase (wbuf);
                    wbuf[0] = mytoupper (wbuf[0]);
                    toutword (toutfile, ichartosstr (wbuf, 1), cent);
                    break;
                case FOLLOWCASE:
                    toutword (toutfile, cent->word, cent);
                    break;
                }
            }
        if (cent->flagfield & MOREVARIANTS)
            cent = cent->next;
        else
            break;
        }
#endif
    }
                
static void toutword (toutfile, word, cent)
    register FILE *     toutfile;
    char *              word;
    register struct dent * cent;
    {
    register int        bit;

    has_marker = 0;
    (void) fprintf (toutfile, "%s", word);
    for (bit = 0;  bit < LARGESTFLAG;  bit++)
        {
        if (TSTMASKBIT (cent->mask, bit))
          flagout (toutfile, BITTOCHAR (bit));
        }
    (void) fprintf (toutfile, "\n");
    }

static void flagout (toutfile, flag)
    register FILE *     toutfile;
    int                 flag;
    {
    if (!has_marker)
        (void) putc (hashheader.flagmarker, toutfile);
    has_marker = 1;
    (void) putc (flag, toutfile);
    }

/*
 * If the string under the given pointer begins with a string character,
 * return the length of that "character".  If not, return 0.
 * May be called any time, but it's best if "isstrstart" is first
 * used to filter out unnecessary calls.
 *
 * As a side effect, "laststringch" is set to the number of the string
 * found, or to -1 if none was found.  This can be useful for such things
 * as case conversion.
 */
int stringcharlen (bufp, canonical)
    char *              bufp;
    int                 canonical;      /* NZ if input is in canonical form */
    {
#ifdef SLOWMULTIPLY
    static char *       sp[MAXSTRINGCHARS];
    static int          inited = 0;
#endif /* SLOWMULTIPLY */
    register char *     bufcur;
    register char *     stringcur;
    register int        stringno;
    register int        lowstringno;
    register int        highstringno;
    int                 dupwanted;

#ifdef SLOWMULTIPLY
    if (!inited)
        {
        inited = 1;
        for (stringno = 0;  stringno < MAXSTRINGCHARS;  stringno++)
            sp[stringno] = &hashheader.stringchars[stringno][0];
        }
#endif /* SLOWMULTIPLY */
    lowstringno = 0;
    highstringno = hashheader.nstrchars - 1;
    dupwanted = canonical ? 0 : defdupchar;
    while (lowstringno <= highstringno)
        {
        stringno = (lowstringno + highstringno) >> 1;
#ifdef SLOWMULTIPLY
        stringcur = sp[stringno];
#else /* SLOWMULTIPLY */
        stringcur = &hashheader.stringchars[stringno][0];
#endif /* SLOWMULTIPLY */
        bufcur = bufp;
        while (*stringcur)
            {
#ifdef NO8BIT
            if (((*bufcur++ ^ *stringcur) & 0x7F) != 0)
#else /* NO8BIT */
            if (*bufcur++ != *stringcur)
#endif /* NO8BIT */
                break;
            /*
            ** We can't use autoincrement above because of the
            ** test below.
            */
            stringcur++;
            }
        if (*stringcur == '\0')
            {
            if (hashheader.dupnos[stringno] == dupwanted)
                {
                /* We have a match */
                laststringch = hashheader.stringdups[stringno];
#ifdef SLOWMULTIPLY
                return stringcur - sp[stringno];
#else /* SLOWMULTIPLY */
                return stringcur - &hashheader.stringchars[stringno][0];
#endif /* SLOWMULTIPLY */
                }
            else
                --stringcur;
            }
        /* No match - choose which side to search on */
#ifdef NO8BIT
        if ((*--bufcur & 0x7F) < (*stringcur & 0x7F))
            highstringno = stringno - 1;
        else if ((*bufcur & 0x7F) > (*stringcur & 0x7F))
            lowstringno = stringno + 1;
#else /* NO8BIT */
        if (*--bufcur < *stringcur)
            highstringno = stringno - 1;
        else if (*bufcur > *stringcur)
            lowstringno = stringno + 1;
#endif /* NO8BIT */
        else if (dupwanted < hashheader.dupnos[stringno])
            highstringno = stringno - 1;
        else
            lowstringno = stringno + 1;
        }
    laststringch = -1;
    return 0;                   /* Not a string character */
    }

/*
 * Convert an external string to an ichar_t string.  If necessary, the parity
 * bit is stripped off as part of the process.
 *
 * Returns NZ if the output string overflowed.
 */
int strtoichar (out, in, outlen, canonical)
    register ichar_t *  out;            /* Where to put result */
    register char *     in;             /* String to convert */
    int                 outlen;         /* Size of output buffer, *BYTES* */
    int                 canonical;      /* NZ if input is in canonical form */
    {
    register int        len;            /* Length of next character */

    outlen /= sizeof (ichar_t);         /* Convert to an ichar_t count */
    for (  ;  --outlen > 0  &&  *in != '\0';  in += len)
        {
        if (l1_isstringch (in, len, canonical))
            *out++ = SET_SIZE + laststringch;
        else
            *out++ = *in & NOPARITY;
        }
    *out = 0;
    return outlen <= 0;
    }

/*
 * Convert an ichar_t string to an external string.
 *
 * WARNING: the resulting string may wind up being longer than the
 * original.  In fact, even the sequence strtoichar->ichartostr may
 * produce a result longer than the original, because the output form
 * may use a different string type set than the original input form.
 *
 * Returns NZ if the output string overflowed.
 */
int ichartostr (out, in, outlen, canonical)
    register char *     out;            /* Where to put result */
    register ichar_t *  in;             /* String to convert */
    int                 outlen;         /* Size of output buffer, bytes */
    int                 canonical;      /* NZ for canonical form */
    {
    register int        ch;             /* Next character to store */
    register int        i;              /* Index into duplicates list */
    register char *     scharp;         /* Pointer into a string char */

    while (--outlen > 0  &&  (ch = *in++) != 0)
        {
        if (ch < SET_SIZE)
            *out++ = (char) ch;
        else
            {
            ch -= SET_SIZE;
            if (!canonical)
                {
                for (i = hashheader.nstrchars;  --i >= 0;  )
                    {
                    if (hashheader.dupnos[i] == defdupchar
                      &&  hashheader.stringdups[i] == ch)
                        {
                        ch = i;
                        break;
                        }
                    }
                }
            scharp = hashheader.stringchars[(unsigned) ch];
            while ((*out++ = *scharp++) != '\0')
                ;
            out--;
            }
        }
    *out = '\0';
    return outlen <= 0;
    }

/*
 * Convert a string to an ichar_t, storing the result in a static area.
 */
ichar_t * strtosichar (in, canonical)
    char *              in;             /* String to convert */
    int                 canonical;      /* NZ if input is in canonical form */
    {
    static ichar_t      out[STRTOSICHAR_SIZE / sizeof (ichar_t)];

    if (strtoichar (out, in, sizeof out, canonical))
        (void) fprintf (stderr, WORD_TOO_LONG (in));
    return out;
    }

/*
 * Convert an ichar_t to a string, storing the result in a static area.
 */
char * ichartosstr (in, canonical)
    ichar_t *           in;             /* Internal string to convert */
    int                 canonical;      /* NZ for canonical conversion */
    {
    static char         out[ICHARTOSSTR_SIZE];

    if (ichartostr (out, in, sizeof out, canonical))
        (void) fprintf (stderr, WORD_TOO_LONG (out));
    return out;
    }

/*
 * Convert a single ichar to a printable string, storing the result in
 * a static area.
 */
char * printichar (in)
    int                 in;
    {
    static char         out[MAXSTRINGCHARLEN + 1];

    if (in < SET_SIZE)
        {
        out[0] = (char) in;
        out[1] = '\0';
        }
    else
        (void) strcpy (out, hashheader.stringchars[(unsigned) in - SET_SIZE]);
    return out;
    }

#ifndef ICHAR_IS_CHAR
/*
 * Copy an ichar_t.
 */
ichar_t * icharcpy (out, in)
    register ichar_t *  out;            /* Destination */
    register ichar_t *  in;             /* Source */
    {
    ichar_t *           origout;        /* Copy of destination for return */

    origout = out;
    while ((*out++ = *in++) != 0)
        ;
    return origout;
    }

/*
 * Return the length of an ichar_t.
 */
int icharlen (in)
    register ichar_t *  in;             /* String to count */
    {
    register int        len;            /* Length so far */

    for (len = 0;  *in++ != 0;  len++)
        ;
    return len;
    }

/*
 * Compare two ichar_t's.
 */
int icharcmp (s1, s2)
    register ichar_t *  s1;
    register ichar_t *  s2;
    {

    while (*s1 != 0)
        {
        if (*s1++ != *s2++)
            return *--s1 - *--s2;
        }
    return *s1 - *s2;
    }

/*
 * Strncmp for two ichar_t's.
 */
int icharncmp (s1, s2, n)
    register ichar_t *  s1;
    register ichar_t *  s2;
    register int        n;
    {

    while (--n >= 0  &&  *s1 != 0)
        {
        if (*s1++ != *s2++)
            return *--s1 - *--s2;
        }
    if (n < 0)
        return 0;
    else
        return *s1 - *s2;
    }

#endif /* ICHAR_IS_CHAR */

int findfiletype (name, searchnames, deformatter)
    char *              name;           /* Name to look up in suffix table */
    int                 searchnames;    /* NZ to search name field of table */
    int *               deformatter;    /* Where to set deformatter type */
    {
    char *              cp;             /* Pointer into suffix list */
    int                 cplen;          /* Length of current suffix */
    register int        i;              /* Index into type table */
    int                 len;            /* Length of the name */

    /*
     * Note:  for now, the deformatter is set to 1 for tex, 0 for nroff.
     * Further, we assume that it's one or the other, so that a test
     * for tex is sufficient.  This needs to be generalized.
     */
    len = strlen (name);
    if (searchnames)
        {
        for (i = 0;  i < hashheader.nstrchartype;  i++)
            {
            if (strcmp (name, chartypes[i].name) == 0)
                {
                if (deformatter != NULL)
                    *deformatter =
                      (strcmp (chartypes[i].deformatter, "tex") == 0);
                return i;
                }
            }
        }
    for (i = 0;  i < hashheader.nstrchartype;  i++)
        {
        for (cp = chartypes[i].suffixes;  *cp != '\0';  cp += cplen + 1)
            {
            cplen = strlen (cp);
            if (len >= cplen  &&  strcmp (&name[len - cplen], cp) == 0)
                {
                if (deformatter != NULL)
                    *deformatter =
                      (strcmp (chartypes[i].deformatter, "tex") == 0);
                return i;
                }
            }
        }
    return -1;
    }

/*
 * The following routines are all dummies for the benefit of lint.
 */
#ifdef lint
int TSTMASKBIT (mask, bit) MASKTYPE * mask; int bit;
    { return bit + (int) *mask; }
void CLRMASKBIT (mask, bit) MASKTYPE * mask; int bit; { bit += (int) *mask; }
void SETMASKBIT (mask, bit) MASKTYPE * mask; int bit; { bit += (int) *mask; }
int BITTOCHAR (bit) int bit; { return bit; }
int CHARTOBIT (ch) int ch; { return ch; }
int myupper (ch) unsigned int ch; { return (int) ch; }
int mylower (ch) unsigned int ch; { return (int) ch; }
int myspace (ch) unsigned int ch; { return (int) ch; }
int iswordch (ch) unsigned int ch; { return (int) ch; }
int isboundarych (ch) unsigned int ch; { return (int) ch; }
int isstringstart (ch) unsigned int ch; { return ch; }
ichar_t mytolower (ch) unsigned int ch; { return (ichar_t) ch; }
ichar_t mytoupper (ch) unsigned int ch; { return (ichar_t) ch; }
#endif /* lint */