[p5sagit/p5-mst-13.2.git] / utf8.c

/*    utf8.c
 *
 *    Copyright (c) 1998-1999, Larry Wall
 *
 *    You may distribute under the terms of either the GNU General Public
 *    License or the Artistic License, as specified in the README file.
 *
 */

/*
 * 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
 * heard of that we don't want to see any closer; and that's the one place
 * we're trying to get to!  And that's just where we can't get, nohow.'
 *
 * 'Well do I understand your speech,' he answered in the same language;
 * 'yet few strangers do so.  Why then do you not speak in the Common Tongue,
 * as is the custom in the West, if you wish to be answered?'
 *
 * ...the travellers perceived that the floor was paved with stones of many
 * hues; branching runes and strange devices intertwined beneath their feet.
 */

#include "EXTERN.h"
#define PERL_IN_UTF8_C
#include "perl.h"

/* Unicode support */

U8 *
Perl_uv_to_utf8(pTHX_ U8 *d, UV uv)
{
    if (uv < 0x80) {
	*d++ = uv;
	return d;
    }
    if (uv < 0x800) {
	*d++ = (( uv >>  6)         | 0xc0);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
    if (uv < 0x10000) {
	*d++ = (( uv >> 12)         | 0xe0);
	*d++ = (((uv >>  6) & 0x3f) | 0x80);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
    if (uv < 0x200000) {
	*d++ = (( uv >> 18)         | 0xf0);
	*d++ = (((uv >> 12) & 0x3f) | 0x80);
	*d++ = (((uv >>  6) & 0x3f) | 0x80);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
    if (uv < 0x4000000) {
	*d++ = (( uv >> 24)         | 0xf8);
	*d++ = (((uv >> 18) & 0x3f) | 0x80);
	*d++ = (((uv >> 12) & 0x3f) | 0x80);
	*d++ = (((uv >>  6) & 0x3f) | 0x80);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
    if (uv < 0x80000000) {
	*d++ = (( uv >> 30)         | 0xfc);
	*d++ = (((uv >> 24) & 0x3f) | 0x80);
	*d++ = (((uv >> 18) & 0x3f) | 0x80);
	*d++ = (((uv >> 12) & 0x3f) | 0x80);
	*d++ = (((uv >>  6) & 0x3f) | 0x80);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
#ifdef Quad_t
    if (uv < 0x2000000000)
#endif
    {
	*d++ =                        0xfe;	/* Can't match U+FEFF! */
	*d++ = (((uv >> 30) & 0x3f) | 0x80);
	*d++ = (((uv >> 24) & 0x3f) | 0x80);
	*d++ = (((uv >> 18) & 0x3f) | 0x80);
	*d++ = (((uv >> 12) & 0x3f) | 0x80);
	*d++ = (((uv >>  6) & 0x3f) | 0x80);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
#ifdef Quad_t
    {
	*d++ =                        0xff;	/* Can't match U+FFFE! */
	*d++ = (((uv >> 36) & 0x3f) | 0x80);
	*d++ = (((uv >> 30) & 0x3f) | 0x80);
	*d++ = (((uv >> 24) & 0x3f) | 0x80);
	*d++ = (((uv >> 18) & 0x3f) | 0x80);
	*d++ = (((uv >> 12) & 0x3f) | 0x80);
	*d++ = (((uv >>  6) & 0x3f) | 0x80);
	*d++ = (( uv        & 0x3f) | 0x80);
	return d;
    }
#endif
}

UV
Perl_utf8_to_uv(pTHX_ U8* s, I32* retlen)
{
    UV uv = *s;
    int len;
    if (!(uv & 0x80)) {
	if (retlen)
	    *retlen = 1;
	return *s;
    }
    if (!(uv & 0x40)) {
        dTHR;
	if (ckWARN_d(WARN_UTF8))     
	    Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character");
	if (retlen)
	    *retlen = 1;
	return *s;
    }

    if      (!(uv & 0x20))	{ len = 2; uv &= 0x1f; }
    else if (!(uv & 0x10))	{ len = 3; uv &= 0x0f; }
    else if (!(uv & 0x08))	{ len = 4; uv &= 0x07; }
    else if (!(uv & 0x04))	{ len = 5; uv &= 0x03; }
    else if (!(uv & 0x02))	{ len = 6; uv &= 0x01; }
    else if (!(uv & 0x01))	{ len = 7; uv &= 0x00; }
    else 			  len = 8;	/* whoa! */

    if (retlen)
	*retlen = len;
    --len;
    s++;
    while (len--) {
	if ((*s & 0xc0) != 0x80) {
            dTHR;
	    if (ckWARN_d(WARN_UTF8))     
	        Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character");
	    if (retlen)
		*retlen -= len + 1;
	    return 0xfffd;
	}
	else
	    uv = (uv << 6) | (*s++ & 0x3f);
    }
    return uv;
}

/* utf8_distance(a,b) is intended to be a - b in pointer arithmetic */

I32
Perl_utf8_distance(pTHX_ U8 *a, U8 *b)
{
    I32 off = 0;
    if (a < b) {
	while (a < b) {
	    a += UTF8SKIP(a);
	    off--;
	}
    }
    else {
	while (b < a) {
	    b += UTF8SKIP(b);
	    off++;
	}
    }
    return off;
}

/* WARNING: do not use the following unless you *know* off is within bounds */

U8 *
Perl_utf8_hop(pTHX_ U8 *s, I32 off)
{
    if (off >= 0) {
	while (off--)
	    s += UTF8SKIP(s);
    }
    else {
	while (off++) {
	    s--;
	    if (*s & 0x80) {
		while ((*s & 0xc0) == 0x80)
		    s--;
	    }
	}
    }
    return s;
}

/* XXX NOTHING CALLS THE FOLLOWING TWO ROUTINES YET!!! */
/*
 * Convert native or reversed UTF-16 to UTF-8.
 *
 * Destination must be pre-extended to 3/2 source.  Do not use in-place.
 * We optimize for native, for obvious reasons. */

U8*
Perl_utf16_to_utf8(pTHX_ U16* p, U8* d, I32 bytelen)
{
    U16* pend = p + bytelen / 2;
    while (p < pend) {
	UV uv = *p++;
	if (uv < 0x80) {
	    *d++ = uv;
	    continue;
	}
	if (uv < 0x800) {
	    *d++ = (( uv >>  6)         | 0xc0);
	    *d++ = (( uv        & 0x3f) | 0x80);
	    continue;
	}
	if (uv >= 0xd800 && uv < 0xdbff) {	/* surrogates */
            dTHR;
	    int low = *p++;
	    if (low < 0xdc00 || low >= 0xdfff) {
		if (ckWARN_d(WARN_UTF8))     
	    	    Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-16 surrogate");
		p--;
		uv = 0xfffd;
	    }
	    uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
	}
	if (uv < 0x10000) {
	    *d++ = (( uv >> 12)         | 0xe0);
	    *d++ = (((uv >>  6) & 0x3f) | 0x80);
	    *d++ = (( uv        & 0x3f) | 0x80);
	    continue;
	}
	else {
	    *d++ = (( uv >> 18)         | 0xf0);
	    *d++ = (((uv >> 12) & 0x3f) | 0x80);
	    *d++ = (((uv >>  6) & 0x3f) | 0x80);
	    *d++ = (( uv        & 0x3f) | 0x80);
	    continue;
	}
    }
    return d;
}

/* Note: this one is slightly destructive of the source. */

U8*
Perl_utf16_to_utf8_reversed(pTHX_ U16* p, U8* d, I32 bytelen)
{
    U8* s = (U8*)p;
    U8* send = s + bytelen;
    while (s < send) {
	U8 tmp = s[0];
	s[0] = s[1];
	s[1] = tmp;
	s += 2;
    }
    return utf16_to_utf8(p, d, bytelen);
}

/* for now these are all defined (inefficiently) in terms of the utf8 versions */

bool
Perl_is_uni_alnum(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_alnum(tmpbuf);
}

bool
Perl_is_uni_alnumc(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_alnumc(tmpbuf);
}

bool
Perl_is_uni_idfirst(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_idfirst(tmpbuf);
}

bool
Perl_is_uni_alpha(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_alpha(tmpbuf);
}

bool
Perl_is_uni_space(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_space(tmpbuf);
}

bool
Perl_is_uni_digit(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_digit(tmpbuf);
}

bool
Perl_is_uni_upper(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_upper(tmpbuf);
}

bool
Perl_is_uni_lower(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_lower(tmpbuf);
}

bool
Perl_is_uni_cntrl(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_cntrl(tmpbuf);
}

bool
Perl_is_uni_graph(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_graph(tmpbuf);
}

bool
Perl_is_uni_print(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_print(tmpbuf);
}

bool
Perl_is_uni_punct(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return is_utf8_punct(tmpbuf);
}

U32
Perl_to_uni_upper(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return to_utf8_upper(tmpbuf);
}

U32
Perl_to_uni_title(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return to_utf8_title(tmpbuf);
}

U32
Perl_to_uni_lower(pTHX_ U32 c)
{
    U8 tmpbuf[10];
    uv_to_utf8(tmpbuf, (UV)c);
    return to_utf8_lower(tmpbuf);
}

/* for now these all assume no locale info available for Unicode > 255 */

bool
Perl_is_uni_alnum_lc(pTHX_ U32 c)
{
    return is_uni_alnum(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_alnumc_lc(pTHX_ U32 c)
{
    return is_uni_alnumc(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_idfirst_lc(pTHX_ U32 c)
{
    return is_uni_idfirst(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_alpha_lc(pTHX_ U32 c)
{
    return is_uni_alpha(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_space_lc(pTHX_ U32 c)
{
    return is_uni_space(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_digit_lc(pTHX_ U32 c)
{
    return is_uni_digit(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_upper_lc(pTHX_ U32 c)
{
    return is_uni_upper(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_lower_lc(pTHX_ U32 c)
{
    return is_uni_lower(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_cntrl_lc(pTHX_ U32 c)
{
    return is_uni_cntrl(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_graph_lc(pTHX_ U32 c)
{
    return is_uni_graph(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_print_lc(pTHX_ U32 c)
{
    return is_uni_print(c);	/* XXX no locale support yet */
}

bool
Perl_is_uni_punct_lc(pTHX_ U32 c)
{
    return is_uni_punct(c);	/* XXX no locale support yet */
}

U32
Perl_to_uni_upper_lc(pTHX_ U32 c)
{
    return to_uni_upper(c);	/* XXX no locale support yet */
}

U32
Perl_to_uni_title_lc(pTHX_ U32 c)
{
    return to_uni_title(c);	/* XXX no locale support yet */
}

U32
Perl_to_uni_lower_lc(pTHX_ U32 c)
{
    return to_uni_lower(c);	/* XXX no locale support yet */
}

bool
Perl_is_utf8_alnum(pTHX_ U8 *p)
{
    if (!PL_utf8_alnum)
	PL_utf8_alnum = swash_init("utf8", "IsAlnum", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_alnum, p);
/*    return *p == '_' || is_utf8_alpha(p) || is_utf8_digit(p); */
#ifdef SURPRISINGLY_SLOWER  /* probably because alpha is usually true */
    if (!PL_utf8_alnum)
	PL_utf8_alnum = swash_init("utf8", "",
	    sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0);
    return swash_fetch(PL_utf8_alnum, p);
#endif
}

bool
Perl_is_utf8_alnumc(pTHX_ U8 *p)
{
    if (!PL_utf8_alnum)
	PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_alnum, p);
/*    return is_utf8_alpha(p) || is_utf8_digit(p); */
#ifdef SURPRISINGLY_SLOWER  /* probably because alpha is usually true */
    if (!PL_utf8_alnum)
	PL_utf8_alnum = swash_init("utf8", "",
	    sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0);
    return swash_fetch(PL_utf8_alnum, p);
#endif
}

bool
Perl_is_utf8_idfirst(pTHX_ U8 *p)
{
    return *p == '_' || is_utf8_alpha(p);
}

bool
Perl_is_utf8_alpha(pTHX_ U8 *p)
{
    if (!PL_utf8_alpha)
	PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_alpha, p);
}

bool
Perl_is_utf8_ascii(pTHX_ U8 *p)
{
    if (!PL_utf8_ascii)
	PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_ascii, p);
}

bool
Perl_is_utf8_space(pTHX_ U8 *p)
{
    if (!PL_utf8_space)
	PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_space, p);
}

bool
Perl_is_utf8_digit(pTHX_ U8 *p)
{
    if (!PL_utf8_digit)
	PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_digit, p);
}

bool
Perl_is_utf8_upper(pTHX_ U8 *p)
{
    if (!PL_utf8_upper)
	PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_upper, p);
}

bool
Perl_is_utf8_lower(pTHX_ U8 *p)
{
    if (!PL_utf8_lower)
	PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_lower, p);
}

bool
Perl_is_utf8_cntrl(pTHX_ U8 *p)
{
    if (!PL_utf8_cntrl)
	PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_cntrl, p);
}

bool
Perl_is_utf8_graph(pTHX_ U8 *p)
{
    if (!PL_utf8_graph)
	PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_graph, p);
}

bool
Perl_is_utf8_print(pTHX_ U8 *p)
{
    if (!PL_utf8_print)
	PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_print, p);
}

bool
Perl_is_utf8_punct(pTHX_ U8 *p)
{
    if (!PL_utf8_punct)
	PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_punct, p);
}

bool
Perl_is_utf8_xdigit(pTHX_ U8 *p)
{
    if (!PL_utf8_xdigit)
	PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_xdigit, p);
}

bool
Perl_is_utf8_mark(pTHX_ U8 *p)
{
    if (!PL_utf8_mark)
	PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0);
    return swash_fetch(PL_utf8_mark, p);
}

UV
Perl_to_utf8_upper(pTHX_ U8 *p)
{
    UV uv;

    if (!PL_utf8_toupper)
	PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0);
    uv = swash_fetch(PL_utf8_toupper, p);
    return uv ? uv : utf8_to_uv(p,0);
}

UV
Perl_to_utf8_title(pTHX_ U8 *p)
{
    UV uv;

    if (!PL_utf8_totitle)
	PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0);
    uv = swash_fetch(PL_utf8_totitle, p);
    return uv ? uv : utf8_to_uv(p,0);
}

UV
Perl_to_utf8_lower(pTHX_ U8 *p)
{
    UV uv;

    if (!PL_utf8_tolower)
	PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0);
    uv = swash_fetch(PL_utf8_tolower, p);
    return uv ? uv : utf8_to_uv(p,0);
}

/* a "swash" is a swatch hash */

SV*
Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none)
{
    SV* retval;
    char tmpbuf[256];
    dSP;    
    PUSHSTACKi(PERLSI_MAGIC);
    PUSHMARK(SP);
    EXTEND(SP,5);
    PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg))));
    PUSHs(sv_2mortal(newSVpvn(name, strlen(name))));
    PUSHs(listsv);
    PUSHs(sv_2mortal(newSViv(minbits)));
    PUSHs(sv_2mortal(newSViv(none)));
    PUTBACK;
    ENTER;
    SAVEI32(PL_hints);
    PL_hints = 0;
    save_re_context();
    if (PL_curcop == &PL_compiling)	/* XXX ought to be handled by lex_start */
	strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf);
    if (call_method("SWASHNEW", G_SCALAR))
	retval = newSVsv(*PL_stack_sp--);    
    else
	retval = &PL_sv_undef;
    LEAVE;
    POPSTACK;
    if (PL_curcop == &PL_compiling) {
	strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf);
	PL_curcop->op_private = PL_hints;
    }
    if (!SvROK(retval) || SvTYPE(SvRV(retval)) != SVt_PVHV)
	Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
    return retval;
}

UV
Perl_swash_fetch(pTHX_ SV *sv, U8 *ptr)
{
    HV* hv = (HV*)SvRV(sv);
    U32 klen = UTF8SKIP(ptr) - 1;
    U32 off = ptr[klen] & 127;  /* NB: 64 bit always 0 when len > 1 */
    STRLEN slen;
    STRLEN needents = (klen ? 64 : 128);
    U8 *tmps;
    U32 bit;
    SV *retval;

    /*
     * This single-entry cache saves about 1/3 of the utf8 overhead in test
     * suite.  (That is, only 7-8% overall over just a hash cache.  Still,
     * it's nothing to sniff at.)  Pity we usually come through at least
     * two function calls to get here...
     *
     * NB: this code assumes that swatches are never modified, once generated!
     */

    if (hv == PL_last_swash_hv &&
	klen == PL_last_swash_klen &&
	(!klen || memEQ(ptr,PL_last_swash_key,klen)) )
    {
	tmps = PL_last_swash_tmps;
	slen = PL_last_swash_slen;
    }
    else {
	/* Try our second-level swatch cache, kept in a hash. */
	SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE);

	/* If not cached, generate it via utf8::SWASHGET */
	if (!svp || !SvPOK(*svp) || !(tmps = (U8*)SvPV(*svp, slen))) {
	    dSP;
	    ENTER;
	    SAVETMPS;
	    save_re_context();
	    PUSHSTACKi(PERLSI_MAGIC);
	    PUSHMARK(SP);
	    EXTEND(SP,3);
	    PUSHs((SV*)sv);
	    PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, 0) & ~(needents - 1))));
	    PUSHs(sv_2mortal(newSViv(needents)));
	    PUTBACK;
	    if (call_method("SWASHGET", G_SCALAR))
		retval = newSVsv(*PL_stack_sp--);    
	    else
		retval = &PL_sv_undef;
	    POPSTACK;
	    FREETMPS;
	    LEAVE;
	    if (PL_curcop == &PL_compiling)
		PL_curcop->op_private = PL_hints;

	    svp = hv_store(hv, (char*)ptr, klen, retval, 0);

	    if (!svp || !(tmps = (U8*)SvPV(*svp, slen)) || slen < 8)
		Perl_croak(aTHX_ "SWASHGET didn't return result of proper length");
	}

	PL_last_swash_hv = hv;
	PL_last_swash_klen = klen;
	PL_last_swash_tmps = tmps;
	PL_last_swash_slen = slen;
	if (klen)
	    Copy(ptr, PL_last_swash_key, klen, U8);
    }

    switch ((slen << 3) / needents) {
    case 1:
	bit = 1 << (off & 7);
	off >>= 3;
	return (tmps[off] & bit) != 0;
    case 8:
	return tmps[off];
    case 16:
	off <<= 1;
	return (tmps[off] << 8) + tmps[off + 1] ;
    case 32:
	off <<= 2;
	return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
    }
    Perl_croak(aTHX_ "panic: swash_fetch");
    return 0;
}
Commit	Line	Data
a0ed51b3	1	/* utf8.c
a0ed51b3	2	*
4eb8286e	3	* Copyright (c) 1998-1999, Larry Wall
a0ed51b3	4	*
	5	* You may distribute under the terms of either the GNU General Public
	6	* License or the Artistic License, as specified in the README file.
	7	*
	8	*/
	9
	10	/*
	11	* 'What a fix!' said Sam. 'That's the one place in all the lands we've ever
	12	* heard of that we don't want to see any closer; and that's the one place
	13	* we're trying to get to! And that's just where we can't get, nohow.'
	14	*
	15	* 'Well do I understand your speech,' he answered in the same language;
	16	* 'yet few strangers do so. Why then do you not speak in the Common Tongue,
	17	* as is the custom in the West, if you wish to be answered?'
	18	*
	19	* ...the travellers perceived that the floor was paved with stones of many
	20	* hues; branching runes and strange devices intertwined beneath their feet.
	21	*/
	22
	23	#include "EXTERN.h"
864dbfa3	24	#define PERL_IN_UTF8_C
a0ed51b3	25	#include "perl.h"
	26
	27	/* Unicode support */
	28
dfe13c55	29	U8 *
864dbfa3	30	Perl_uv_to_utf8(pTHX_ U8 *d, UV uv)
a0ed51b3	31	{
	32	if (uv < 0x80) {
	33	*d++ = uv;
	34	return d;
	35	}
	36	if (uv < 0x800) {
	37	*d++ = (( uv >> 6) \| 0xc0);
	38	*d++ = (( uv & 0x3f) \| 0x80);
	39	return d;
	40	}
	41	if (uv < 0x10000) {
	42	*d++ = (( uv >> 12) \| 0xe0);
	43	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	44	*d++ = (( uv & 0x3f) \| 0x80);
	45	return d;
	46	}
	47	if (uv < 0x200000) {
	48	*d++ = (( uv >> 18) \| 0xf0);
	49	*d++ = (((uv >> 12) & 0x3f) \| 0x80);
	50	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	51	*d++ = (( uv & 0x3f) \| 0x80);
	52	return d;
	53	}
	54	if (uv < 0x4000000) {
	55	*d++ = (( uv >> 24) \| 0xf8);
	56	*d++ = (((uv >> 18) & 0x3f) \| 0x80);
	57	*d++ = (((uv >> 12) & 0x3f) \| 0x80);
	58	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	59	*d++ = (( uv & 0x3f) \| 0x80);
	60	return d;
	61	}
	62	if (uv < 0x80000000) {
	63	*d++ = (( uv >> 30) \| 0xfc);
	64	*d++ = (((uv >> 24) & 0x3f) \| 0x80);
	65	*d++ = (((uv >> 18) & 0x3f) \| 0x80);
	66	*d++ = (((uv >> 12) & 0x3f) \| 0x80);
	67	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	68	*d++ = (( uv & 0x3f) \| 0x80);
	69	return d;
	70	}
	71	#ifdef Quad_t
	72	if (uv < 0x2000000000)
	73	#endif
	74	{
	75	d++ = 0xfe; / Can't match U+FEFF! */
	76	*d++ = (((uv >> 30) & 0x3f) \| 0x80);
	77	*d++ = (((uv >> 24) & 0x3f) \| 0x80);
	78	*d++ = (((uv >> 18) & 0x3f) \| 0x80);
	79	*d++ = (((uv >> 12) & 0x3f) \| 0x80);
	80	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	81	*d++ = (( uv & 0x3f) \| 0x80);
	82	return d;
	83	}
	84	#ifdef Quad_t
	85	{
	86	d++ = 0xff; / Can't match U+FFFE! */
	87	*d++ = (((uv >> 36) & 0x3f) \| 0x80);
	88	*d++ = (((uv >> 30) & 0x3f) \| 0x80);
	89	*d++ = (((uv >> 24) & 0x3f) \| 0x80);
	90	*d++ = (((uv >> 18) & 0x3f) \| 0x80);
	91	*d++ = (((uv >> 12) & 0x3f) \| 0x80);
	92	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	93	*d++ = (( uv & 0x3f) \| 0x80);
	94	return d;
95	}
96	#endif
97	}
98
99	UV
864dbfa3	100	Perl_utf8_to_uv(pTHX_ U8* s, I32* retlen)
a0ed51b3	101	{
	102	UV uv = *s;
	103	int len;
	104	if (!(uv & 0x80)) {
	105	if (retlen)
	106	*retlen = 1;
	107	return *s;
	108	}
	109	if (!(uv & 0x40)) {
0453d815	110	dTHR;
	111	if (ckWARN_d(WARN_UTF8))
	112	Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character");
a0ed51b3	113	if (retlen)
	114	*retlen = 1;
	115	return *s;
	116	}
	117
	118	if (!(uv & 0x20)) { len = 2; uv &= 0x1f; }
	119	else if (!(uv & 0x10)) { len = 3; uv &= 0x0f; }
	120	else if (!(uv & 0x08)) { len = 4; uv &= 0x07; }
	121	else if (!(uv & 0x04)) { len = 5; uv &= 0x03; }
	122	else if (!(uv & 0x02)) { len = 6; uv &= 0x01; }
	123	else if (!(uv & 0x01)) { len = 7; uv &= 0x00; }
	124	else len = 8; /* whoa! */
	125
	126	if (retlen)
	127	*retlen = len;
	128	--len;
	129	s++;
	130	while (len--) {
	131	if ((*s & 0xc0) != 0x80) {
0453d815	132	dTHR;
	133	if (ckWARN_d(WARN_UTF8))
	134	Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-8 character");
a0ed51b3	135	if (retlen)
	136	*retlen -= len + 1;
	137	return 0xfffd;
	138	}
	139	else
	140	uv = (uv << 6) \| (*s++ & 0x3f);
	141	}
	142	return uv;
	143	}
	144
	145	/* utf8_distance(a,b) is intended to be a - b in pointer arithmetic */
	146
	147	I32
864dbfa3	148	Perl_utf8_distance(pTHX_ U8 a, U8 b)
a0ed51b3	149	{
	150	I32 off = 0;
	151	if (a < b) {
	152	while (a < b) {
	153	a += UTF8SKIP(a);
	154	off--;
	155	}
	156	}
	157	else {
	158	while (b < a) {
	159	b += UTF8SKIP(b);
	160	off++;
	161	}
	162	}
	163	return off;
	164	}
	165
	166	/* WARNING: do not use the following unless you know off is within bounds */
	167
	168	U8 *
864dbfa3	169	Perl_utf8_hop(pTHX_ U8 *s, I32 off)
a0ed51b3	170	{
	171	if (off >= 0) {
	172	while (off--)
	173	s += UTF8SKIP(s);
	174	}
	175	else {
	176	while (off++) {
	177	s--;
	178	if (*s & 0x80) {
	179	while ((*s & 0xc0) == 0x80)
	180	s--;
	181	}
	182	}
	183	}
	184	return s;
	185	}
	186
	187	/* XXX NOTHING CALLS THE FOLLOWING TWO ROUTINES YET!!! */
	188	/*
	189	* Convert native or reversed UTF-16 to UTF-8.
	190	*
	191	* Destination must be pre-extended to 3/2 source. Do not use in-place.
	192	* We optimize for native, for obvious reasons. */
	193
	194	U8*
864dbfa3	195	Perl_utf16_to_utf8(pTHX_ U16* p, U8* d, I32 bytelen)
a0ed51b3	196	{
	197	U16* pend = p + bytelen / 2;
	198	while (p < pend) {
	199	UV uv = *p++;
	200	if (uv < 0x80) {
	201	*d++ = uv;
	202	continue;
	203	}
	204	if (uv < 0x800) {
	205	*d++ = (( uv >> 6) \| 0xc0);
	206	*d++ = (( uv & 0x3f) \| 0x80);
	207	continue;
	208	}
	209	if (uv >= 0xd800 && uv < 0xdbff) { /* surrogates */
0453d815	210	dTHR;
a0ed51b3	211	int low = *p++;
a0ed51b3	212	if (low < 0xdc00 \|\| low >= 0xdfff) {
0453d815	213	if (ckWARN_d(WARN_UTF8))
0453d815	214	Perl_warner(aTHX_ WARN_UTF8, "Malformed UTF-16 surrogate");
a0ed51b3	215	p--;
	216	uv = 0xfffd;
	217	}
	218	uv = ((uv - 0xd800) << 10) + (low - 0xdc00) + 0x10000;
	219	}
	220	if (uv < 0x10000) {
	221	*d++ = (( uv >> 12) \| 0xe0);
	222	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	223	*d++ = (( uv & 0x3f) \| 0x80);
	224	continue;
	225	}
	226	else {
	227	*d++ = (( uv >> 18) \| 0xf0);
	228	*d++ = (((uv >> 12) & 0x3f) \| 0x80);
	229	*d++ = (((uv >> 6) & 0x3f) \| 0x80);
	230	*d++ = (( uv & 0x3f) \| 0x80);
	231	continue;
	232	}
	233	}
	234	return d;
	235	}
	236
	237	/* Note: this one is slightly destructive of the source. */
	238
	239	U8*
864dbfa3	240	Perl_utf16_to_utf8_reversed(pTHX_ U16* p, U8* d, I32 bytelen)
a0ed51b3	241	{
	242	U8* s = (U8*)p;
	243	U8* send = s + bytelen;
	244	while (s < send) {
	245	U8 tmp = s[0];
	246	s[0] = s[1];
	247	s[1] = tmp;
	248	s += 2;
	249	}
	250	return utf16_to_utf8(p, d, bytelen);
	251	}
	252
	253	/* for now these are all defined (inefficiently) in terms of the utf8 versions */
	254
	255	bool
864dbfa3	256	Perl_is_uni_alnum(pTHX_ U32 c)
a0ed51b3	257	{
dfe13c55	258	U8 tmpbuf[10];
a0ed51b3	259	uv_to_utf8(tmpbuf, (UV)c);
	260	return is_utf8_alnum(tmpbuf);
	261	}
	262
	263	bool
b8c5462f	264	Perl_is_uni_alnumc(pTHX_ U32 c)
	265	{
	266	U8 tmpbuf[10];
	267	uv_to_utf8(tmpbuf, (UV)c);
	268	return is_utf8_alnumc(tmpbuf);
	269	}
	270
	271	bool
864dbfa3	272	Perl_is_uni_idfirst(pTHX_ U32 c)
a0ed51b3	273	{
dfe13c55	274	U8 tmpbuf[10];
a0ed51b3	275	uv_to_utf8(tmpbuf, (UV)c);
	276	return is_utf8_idfirst(tmpbuf);
	277	}
	278
	279	bool
864dbfa3	280	Perl_is_uni_alpha(pTHX_ U32 c)
a0ed51b3	281	{
dfe13c55	282	U8 tmpbuf[10];
a0ed51b3	283	uv_to_utf8(tmpbuf, (UV)c);
	284	return is_utf8_alpha(tmpbuf);
	285	}
	286
	287	bool
864dbfa3	288	Perl_is_uni_space(pTHX_ U32 c)
a0ed51b3	289	{
dfe13c55	290	U8 tmpbuf[10];
a0ed51b3	291	uv_to_utf8(tmpbuf, (UV)c);
	292	return is_utf8_space(tmpbuf);
	293	}
	294
	295	bool
864dbfa3	296	Perl_is_uni_digit(pTHX_ U32 c)
a0ed51b3	297	{
dfe13c55	298	U8 tmpbuf[10];
a0ed51b3	299	uv_to_utf8(tmpbuf, (UV)c);
	300	return is_utf8_digit(tmpbuf);
	301	}
	302
	303	bool
864dbfa3	304	Perl_is_uni_upper(pTHX_ U32 c)
a0ed51b3	305	{
dfe13c55	306	U8 tmpbuf[10];
a0ed51b3	307	uv_to_utf8(tmpbuf, (UV)c);
	308	return is_utf8_upper(tmpbuf);
	309	}
	310
	311	bool
864dbfa3	312	Perl_is_uni_lower(pTHX_ U32 c)
a0ed51b3	313	{
dfe13c55	314	U8 tmpbuf[10];
a0ed51b3	315	uv_to_utf8(tmpbuf, (UV)c);
	316	return is_utf8_lower(tmpbuf);
	317	}
	318
	319	bool
b8c5462f	320	Perl_is_uni_cntrl(pTHX_ U32 c)
	321	{
	322	U8 tmpbuf[10];
	323	uv_to_utf8(tmpbuf, (UV)c);
	324	return is_utf8_cntrl(tmpbuf);
	325	}
	326
	327	bool
	328	Perl_is_uni_graph(pTHX_ U32 c)
	329	{
	330	U8 tmpbuf[10];
	331	uv_to_utf8(tmpbuf, (UV)c);
	332	return is_utf8_graph(tmpbuf);
	333	}
	334
	335	bool
864dbfa3	336	Perl_is_uni_print(pTHX_ U32 c)
a0ed51b3	337	{
dfe13c55	338	U8 tmpbuf[10];
a0ed51b3	339	uv_to_utf8(tmpbuf, (UV)c);
	340	return is_utf8_print(tmpbuf);
	341	}
	342
b8c5462f	343	bool
f248d071	344	Perl_is_uni_punct(pTHX_ U32 c)
b8c5462f	345	{
	346	U8 tmpbuf[10];
	347	uv_to_utf8(tmpbuf, (UV)c);
	348	return is_utf8_punct(tmpbuf);
	349	}
	350
a0ed51b3	351	U32
864dbfa3	352	Perl_to_uni_upper(pTHX_ U32 c)
a0ed51b3	353	{
dfe13c55	354	U8 tmpbuf[10];
a0ed51b3	355	uv_to_utf8(tmpbuf, (UV)c);
	356	return to_utf8_upper(tmpbuf);
	357	}
	358
	359	U32
864dbfa3	360	Perl_to_uni_title(pTHX_ U32 c)
a0ed51b3	361	{
dfe13c55	362	U8 tmpbuf[10];
a0ed51b3	363	uv_to_utf8(tmpbuf, (UV)c);
	364	return to_utf8_title(tmpbuf);
	365	}
	366
	367	U32
864dbfa3	368	Perl_to_uni_lower(pTHX_ U32 c)
a0ed51b3	369	{
dfe13c55	370	U8 tmpbuf[10];
a0ed51b3	371	uv_to_utf8(tmpbuf, (UV)c);
	372	return to_utf8_lower(tmpbuf);
	373	}
	374
	375	/* for now these all assume no locale info available for Unicode > 255 */
	376
	377	bool
864dbfa3	378	Perl_is_uni_alnum_lc(pTHX_ U32 c)
a0ed51b3	379	{
	380	return is_uni_alnum(c); /* XXX no locale support yet */
	381	}
	382
	383	bool
b8c5462f	384	Perl_is_uni_alnumc_lc(pTHX_ U32 c)
	385	{
	386	return is_uni_alnumc(c); /* XXX no locale support yet */
	387	}
	388
	389	bool
864dbfa3	390	Perl_is_uni_idfirst_lc(pTHX_ U32 c)
a0ed51b3	391	{
	392	return is_uni_idfirst(c); /* XXX no locale support yet */
	393	}
	394
	395	bool
864dbfa3	396	Perl_is_uni_alpha_lc(pTHX_ U32 c)
a0ed51b3	397	{
	398	return is_uni_alpha(c); /* XXX no locale support yet */
	399	}
	400
	401	bool
864dbfa3	402	Perl_is_uni_space_lc(pTHX_ U32 c)
a0ed51b3	403	{
	404	return is_uni_space(c); /* XXX no locale support yet */
	405	}
	406
	407	bool
864dbfa3	408	Perl_is_uni_digit_lc(pTHX_ U32 c)
a0ed51b3	409	{
	410	return is_uni_digit(c); /* XXX no locale support yet */
	411	}
	412
	413	bool
864dbfa3	414	Perl_is_uni_upper_lc(pTHX_ U32 c)
a0ed51b3	415	{
	416	return is_uni_upper(c); /* XXX no locale support yet */
	417	}
	418
	419	bool
864dbfa3	420	Perl_is_uni_lower_lc(pTHX_ U32 c)
a0ed51b3	421	{
	422	return is_uni_lower(c); /* XXX no locale support yet */
	423	}
	424
	425	bool
b8c5462f	426	Perl_is_uni_cntrl_lc(pTHX_ U32 c)
	427	{
	428	return is_uni_cntrl(c); /* XXX no locale support yet */
	429	}
	430
	431	bool
	432	Perl_is_uni_graph_lc(pTHX_ U32 c)
	433	{
	434	return is_uni_graph(c); /* XXX no locale support yet */
	435	}
	436
	437	bool
864dbfa3	438	Perl_is_uni_print_lc(pTHX_ U32 c)
a0ed51b3	439	{
	440	return is_uni_print(c); /* XXX no locale support yet */
	441	}
	442
b8c5462f	443	bool
	444	Perl_is_uni_punct_lc(pTHX_ U32 c)
	445	{
	446	return is_uni_punct(c); /* XXX no locale support yet */
	447	}
	448
a0ed51b3	449	U32
864dbfa3	450	Perl_to_uni_upper_lc(pTHX_ U32 c)
a0ed51b3	451	{
	452	return to_uni_upper(c); /* XXX no locale support yet */
	453	}
	454
	455	U32
864dbfa3	456	Perl_to_uni_title_lc(pTHX_ U32 c)
a0ed51b3	457	{
	458	return to_uni_title(c); /* XXX no locale support yet */
	459	}
	460
	461	U32
864dbfa3	462	Perl_to_uni_lower_lc(pTHX_ U32 c)
a0ed51b3	463	{
	464	return to_uni_lower(c); /* XXX no locale support yet */
	465	}
	466
a0ed51b3	467	bool
864dbfa3	468	Perl_is_utf8_alnum(pTHX_ U8 *p)
a0ed51b3	469	{
a0ed51b3	470	if (!PL_utf8_alnum)
e24b16f9	471	PL_utf8_alnum = swash_init("utf8", "IsAlnum", &PL_sv_undef, 0, 0);
a0ed51b3	472	return swash_fetch(PL_utf8_alnum, p);
	473	/* return p == '_' \|\| is_utf8_alpha(p) \|\| is_utf8_digit(p); /
	474	#ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */
	475	if (!PL_utf8_alnum)
	476	PL_utf8_alnum = swash_init("utf8", "",
	477	sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0);
	478	return swash_fetch(PL_utf8_alnum, p);
	479	#endif
	480	}
	481
	482	bool
b8c5462f	483	Perl_is_utf8_alnumc(pTHX_ U8 *p)
	484	{
	485	if (!PL_utf8_alnum)
	486	PL_utf8_alnum = swash_init("utf8", "IsAlnumC", &PL_sv_undef, 0, 0);
	487	return swash_fetch(PL_utf8_alnum, p);
	488	/* return is_utf8_alpha(p) \|\| is_utf8_digit(p); */
	489	#ifdef SURPRISINGLY_SLOWER /* probably because alpha is usually true */
	490	if (!PL_utf8_alnum)
	491	PL_utf8_alnum = swash_init("utf8", "",
	492	sv_2mortal(newSVpv("+utf8::IsAlpha\n+utf8::IsDigit\n005F\n",0)), 0, 0);
	493	return swash_fetch(PL_utf8_alnum, p);
	494	#endif
	495	}
	496
	497	bool
864dbfa3	498	Perl_is_utf8_idfirst(pTHX_ U8 *p)
a0ed51b3	499	{
	500	return *p == '_' \|\| is_utf8_alpha(p);
	501	}
	502
	503	bool
864dbfa3	504	Perl_is_utf8_alpha(pTHX_ U8 *p)
a0ed51b3	505	{
a0ed51b3	506	if (!PL_utf8_alpha)
e24b16f9	507	PL_utf8_alpha = swash_init("utf8", "IsAlpha", &PL_sv_undef, 0, 0);
a0ed51b3	508	return swash_fetch(PL_utf8_alpha, p);
	509	}
	510
	511	bool
b8c5462f	512	Perl_is_utf8_ascii(pTHX_ U8 *p)
	513	{
	514	if (!PL_utf8_ascii)
	515	PL_utf8_ascii = swash_init("utf8", "IsAscii", &PL_sv_undef, 0, 0);
	516	return swash_fetch(PL_utf8_ascii, p);
	517	}
	518
	519	bool
864dbfa3	520	Perl_is_utf8_space(pTHX_ U8 *p)
a0ed51b3	521	{
a0ed51b3	522	if (!PL_utf8_space)
e24b16f9	523	PL_utf8_space = swash_init("utf8", "IsSpace", &PL_sv_undef, 0, 0);
a0ed51b3	524	return swash_fetch(PL_utf8_space, p);
	525	}
	526
	527	bool
864dbfa3	528	Perl_is_utf8_digit(pTHX_ U8 *p)
a0ed51b3	529	{
a0ed51b3	530	if (!PL_utf8_digit)
e24b16f9	531	PL_utf8_digit = swash_init("utf8", "IsDigit", &PL_sv_undef, 0, 0);
a0ed51b3	532	return swash_fetch(PL_utf8_digit, p);
	533	}
	534
	535	bool
864dbfa3	536	Perl_is_utf8_upper(pTHX_ U8 *p)
a0ed51b3	537	{
a0ed51b3	538	if (!PL_utf8_upper)
e24b16f9	539	PL_utf8_upper = swash_init("utf8", "IsUpper", &PL_sv_undef, 0, 0);
a0ed51b3	540	return swash_fetch(PL_utf8_upper, p);
	541	}
	542
	543	bool
864dbfa3	544	Perl_is_utf8_lower(pTHX_ U8 *p)
a0ed51b3	545	{
a0ed51b3	546	if (!PL_utf8_lower)
e24b16f9	547	PL_utf8_lower = swash_init("utf8", "IsLower", &PL_sv_undef, 0, 0);
a0ed51b3	548	return swash_fetch(PL_utf8_lower, p);
	549	}
	550
	551	bool
b8c5462f	552	Perl_is_utf8_cntrl(pTHX_ U8 *p)
	553	{
	554	if (!PL_utf8_cntrl)
	555	PL_utf8_cntrl = swash_init("utf8", "IsCntrl", &PL_sv_undef, 0, 0);
	556	return swash_fetch(PL_utf8_cntrl, p);
	557	}
	558
	559	bool
	560	Perl_is_utf8_graph(pTHX_ U8 *p)
	561	{
	562	if (!PL_utf8_graph)
	563	PL_utf8_graph = swash_init("utf8", "IsGraph", &PL_sv_undef, 0, 0);
	564	return swash_fetch(PL_utf8_graph, p);
	565	}
	566
	567	bool
864dbfa3	568	Perl_is_utf8_print(pTHX_ U8 *p)
a0ed51b3	569	{
a0ed51b3	570	if (!PL_utf8_print)
e24b16f9	571	PL_utf8_print = swash_init("utf8", "IsPrint", &PL_sv_undef, 0, 0);
a0ed51b3	572	return swash_fetch(PL_utf8_print, p);
	573	}
	574
	575	bool
b8c5462f	576	Perl_is_utf8_punct(pTHX_ U8 *p)
	577	{
	578	if (!PL_utf8_punct)
	579	PL_utf8_punct = swash_init("utf8", "IsPunct", &PL_sv_undef, 0, 0);
	580	return swash_fetch(PL_utf8_punct, p);
	581	}
	582
	583	bool
	584	Perl_is_utf8_xdigit(pTHX_ U8 *p)
	585	{
	586	if (!PL_utf8_xdigit)
	587	PL_utf8_xdigit = swash_init("utf8", "IsXDigit", &PL_sv_undef, 0, 0);
	588	return swash_fetch(PL_utf8_xdigit, p);
	589	}
	590
	591	bool
864dbfa3	592	Perl_is_utf8_mark(pTHX_ U8 *p)
a0ed51b3	593	{
a0ed51b3	594	if (!PL_utf8_mark)
e24b16f9	595	PL_utf8_mark = swash_init("utf8", "IsM", &PL_sv_undef, 0, 0);
a0ed51b3	596	return swash_fetch(PL_utf8_mark, p);
	597	}
	598
2104c8d9	599	UV
864dbfa3	600	Perl_to_utf8_upper(pTHX_ U8 *p)
a0ed51b3	601	{
	602	UV uv;
	603
	604	if (!PL_utf8_toupper)
e24b16f9	605	PL_utf8_toupper = swash_init("utf8", "ToUpper", &PL_sv_undef, 4, 0);
a0ed51b3	606	uv = swash_fetch(PL_utf8_toupper, p);
	607	return uv ? uv : utf8_to_uv(p,0);
	608	}
	609
2104c8d9	610	UV
864dbfa3	611	Perl_to_utf8_title(pTHX_ U8 *p)
a0ed51b3	612	{
	613	UV uv;
	614
	615	if (!PL_utf8_totitle)
e24b16f9	616	PL_utf8_totitle = swash_init("utf8", "ToTitle", &PL_sv_undef, 4, 0);
a0ed51b3	617	uv = swash_fetch(PL_utf8_totitle, p);
	618	return uv ? uv : utf8_to_uv(p,0);
	619	}
	620
2104c8d9	621	UV
864dbfa3	622	Perl_to_utf8_lower(pTHX_ U8 *p)
a0ed51b3	623	{
	624	UV uv;
	625
	626	if (!PL_utf8_tolower)
e24b16f9	627	PL_utf8_tolower = swash_init("utf8", "ToLower", &PL_sv_undef, 4, 0);
a0ed51b3	628	uv = swash_fetch(PL_utf8_tolower, p);
	629	return uv ? uv : utf8_to_uv(p,0);
	630	}
	631
	632	/* a "swash" is a swatch hash */
	633
	634	SV*
864dbfa3	635	Perl_swash_init(pTHX_ char* pkg, char* name, SV *listsv, I32 minbits, I32 none)
a0ed51b3	636	{
	637	SV* retval;
	638	char tmpbuf[256];
	639	dSP;
	640	PUSHSTACKi(PERLSI_MAGIC);
	641	PUSHMARK(SP);
	642	EXTEND(SP,5);
	643	PUSHs(sv_2mortal(newSVpvn(pkg, strlen(pkg))));
	644	PUSHs(sv_2mortal(newSVpvn(name, strlen(name))));
	645	PUSHs(listsv);
	646	PUSHs(sv_2mortal(newSViv(minbits)));
	647	PUSHs(sv_2mortal(newSViv(none)));
	648	PUTBACK;
	649	ENTER;
	650	SAVEI32(PL_hints);
	651	PL_hints = 0;
	652	save_re_context();
e24b16f9	653	if (PL_curcop == &PL_compiling) /* XXX ought to be handled by lex_start */
a0ed51b3	654	strncpy(tmpbuf, PL_tokenbuf, sizeof tmpbuf);
864dbfa3	655	if (call_method("SWASHNEW", G_SCALAR))
e24b16f9	656	retval = newSVsv(*PL_stack_sp--);
a0ed51b3	657	else
e24b16f9	658	retval = &PL_sv_undef;
a0ed51b3	659	LEAVE;
a0ed51b3	660	POPSTACK;
e24b16f9	661	if (PL_curcop == &PL_compiling) {
a0ed51b3	662	strncpy(PL_tokenbuf, tmpbuf, sizeof tmpbuf);
e24b16f9	663	PL_curcop->op_private = PL_hints;
a0ed51b3	664	}
a0ed51b3	665	if (!SvROK(retval) \|\| SvTYPE(SvRV(retval)) != SVt_PVHV)
cea2e8a9	666	Perl_croak(aTHX_ "SWASHNEW didn't return an HV ref");
a0ed51b3	667	return retval;
	668	}
	669
	670	UV
864dbfa3	671	Perl_swash_fetch(pTHX_ SV sv, U8 ptr)
a0ed51b3	672	{
	673	HV* hv = (HV*)SvRV(sv);
	674	U32 klen = UTF8SKIP(ptr) - 1;
	675	U32 off = ptr[klen] & 127; /* NB: 64 bit always 0 when len > 1 */
	676	STRLEN slen;
	677	STRLEN needents = (klen ? 64 : 128);
dfe13c55	678	U8 *tmps;
a0ed51b3	679	U32 bit;
	680	SV *retval;
	681
	682	/*
	683	* This single-entry cache saves about 1/3 of the utf8 overhead in test
	684	* suite. (That is, only 7-8% overall over just a hash cache. Still,
	685	* it's nothing to sniff at.) Pity we usually come through at least
	686	* two function calls to get here...
	687	*
	688	* NB: this code assumes that swatches are never modified, once generated!
	689	*/
	690
	691	if (hv == PL_last_swash_hv &&
	692	klen == PL_last_swash_klen &&
	693	(!klen \|\| memEQ(ptr,PL_last_swash_key,klen)) )
	694	{
	695	tmps = PL_last_swash_tmps;
	696	slen = PL_last_swash_slen;
	697	}
	698	else {
	699	/* Try our second-level swatch cache, kept in a hash. */
dfe13c55	700	SV** svp = hv_fetch(hv, (char*)ptr, klen, FALSE);
a0ed51b3	701
a0ed51b3	702	/* If not cached, generate it via utf8::SWASHGET */
dfe13c55	703	if (!svp \|\| !SvPOK(svp) \|\| !(tmps = (U8)SvPV(*svp, slen))) {
a0ed51b3	704	dSP;
	705	ENTER;
	706	SAVETMPS;
	707	save_re_context();
	708	PUSHSTACKi(PERLSI_MAGIC);
	709	PUSHMARK(SP);
	710	EXTEND(SP,3);
	711	PUSHs((SV*)sv);
	712	PUSHs(sv_2mortal(newSViv(utf8_to_uv(ptr, 0) & ~(needents - 1))));
	713	PUSHs(sv_2mortal(newSViv(needents)));
	714	PUTBACK;
864dbfa3	715	if (call_method("SWASHGET", G_SCALAR))
e24b16f9	716	retval = newSVsv(*PL_stack_sp--);
a0ed51b3	717	else
e24b16f9	718	retval = &PL_sv_undef;
a0ed51b3	719	POPSTACK;
	720	FREETMPS;
	721	LEAVE;
e24b16f9	722	if (PL_curcop == &PL_compiling)
e24b16f9	723	PL_curcop->op_private = PL_hints;
a0ed51b3	724
dfe13c55	725	svp = hv_store(hv, (char*)ptr, klen, retval, 0);
a0ed51b3	726
dfe13c55	727	if (!svp \|\| !(tmps = (U8)SvPV(svp, slen)) \|\| slen < 8)
cea2e8a9	728	Perl_croak(aTHX_ "SWASHGET didn't return result of proper length");
a0ed51b3	729	}
	730
	731	PL_last_swash_hv = hv;
	732	PL_last_swash_klen = klen;
	733	PL_last_swash_tmps = tmps;
	734	PL_last_swash_slen = slen;
	735	if (klen)
	736	Copy(ptr, PL_last_swash_key, klen, U8);
	737	}
	738
	739	switch ((slen << 3) / needents) {
	740	case 1:
	741	bit = 1 << (off & 7);
	742	off >>= 3;
	743	return (tmps[off] & bit) != 0;
	744	case 8:
	745	return tmps[off];
	746	case 16:
	747	off <<= 1;
	748	return (tmps[off] << 8) + tmps[off + 1] ;
	749	case 32:
	750	off <<= 2;
	751	return (tmps[off] << 24) + (tmps[off+1] << 16) + (tmps[off+2] << 8) + tmps[off + 3] ;
	752	}
cea2e8a9	753	Perl_croak(aTHX_ "panic: swash_fetch");
a0ed51b3	754	return 0;
a0ed51b3	755	}