ustring.cpp   [plain text]

// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
/*
******************************************************************************
*
*   Copyright (C) 1998-2016, International Business Machines
*   Corporation and others.  All Rights Reserved.
*
******************************************************************************
*
* File ustring.cpp
*
* Modification History:
*
*   Date        Name        Description
*   12/07/98    bertrand    Creation.
******************************************************************************
*/

#include "unicode/utypes.h"
#include "unicode/putil.h"
#include "unicode/uchar.h"
#include "unicode/ustring.h"
#include "unicode/utf16.h"
#include "cstring.h"
#include "cwchar.h"
#include "cmemory.h"
#include "ustr_imp.h"

/* ANSI string.h - style functions ------------------------------------------ */

/* U+ffff is the highest BMP code point, the highest one that fits into a 16-bit UChar */
#define U_BMP_MAX 0xffff

/* Forward binary string search functions ----------------------------------- */

/*
 * Test if a substring match inside a string is at code point boundaries.
 * All pointers refer to the same buffer.
 * The limit pointer may be NULL, all others must be real pointers.
 */
static inline UBool
isMatchAtCPBoundary(const UChar *start, const UChar *match, const UChar *matchLimit, const UChar *limit) {
    if(U16_IS_TRAIL(*match) && start!=match && U16_IS_LEAD(*(match-1))) {
        /* the leading edge of the match is in the middle of a surrogate pair */
        return FALSE;
    }
    if(U16_IS_LEAD(*(matchLimit-1)) && match!=limit && U16_IS_TRAIL(*matchLimit)) {
        /* the trailing edge of the match is in the middle of a surrogate pair */
        return FALSE;
    }
    return TRUE;
}

U_CAPI UChar * U_EXPORT2
u_strFindFirst(const UChar *s, int32_t length,
               const UChar *sub, int32_t subLength) {
    const UChar *start, *p, *q, *subLimit;
    UChar c, cs, cq;

    if(sub==NULL || subLength<-1) {
        return (UChar *)s;
    }
    if(s==NULL || length<-1) {
        return NULL;
    }

    start=s;

    if(length<0 && subLength<0) {
        /* both strings are NUL-terminated */
        if((cs=*sub++)==0) {
            return (UChar *)s;
        }
        if(*sub==0 && !U16_IS_SURROGATE(cs)) {
            /* the substring consists of a single, non-surrogate BMP code point */
            return u_strchr(s, cs);
        }

        while((c=*s++)!=0) {
            if(c==cs) {
                /* found first substring UChar, compare rest */
                p=s;
                q=sub;
                for(;;) {
                    if((cq=*q)==0) {
                        if(isMatchAtCPBoundary(start, s-1, p, NULL)) {
                            return (UChar *)(s-1); /* well-formed match */
                        } else {
                            break; /* no match because surrogate pair is split */
                        }
                    }
                    if((c=*p)==0) {
                        return NULL; /* no match, and none possible after s */
                    }
                    if(c!=cq) {
                        break; /* no match */
                    }
                    ++p;
                    ++q;
                }
            }
        }

        /* not found */
        return NULL;
    }

    if(subLength<0) {
        subLength=u_strlen(sub);
    }
    if(subLength==0) {
        return (UChar *)s;
    }

    /* get sub[0] to search for it fast */
    cs=*sub++;
    --subLength;
    subLimit=sub+subLength;

    if(subLength==0 && !U16_IS_SURROGATE(cs)) {
        /* the substring consists of a single, non-surrogate BMP code point */
        return length<0 ? u_strchr(s, cs) : u_memchr(s, cs, length);
    }

    if(length<0) {
        /* s is NUL-terminated */
        while((c=*s++)!=0) {
            if(c==cs) {
                /* found first substring UChar, compare rest */
                p=s;
                q=sub;
                for(;;) {
                    if(q==subLimit) {
                        if(isMatchAtCPBoundary(start, s-1, p, NULL)) {
                            return (UChar *)(s-1); /* well-formed match */
                        } else {
                            break; /* no match because surrogate pair is split */
                        }
                    }
                    if((c=*p)==0) {
                        return NULL; /* no match, and none possible after s */
                    }
                    if(c!=*q) {
                        break; /* no match */
                    }
                    ++p;
                    ++q;
                }
            }
        }
    } else {
        const UChar *limit, *preLimit;

        /* subLength was decremented above */
        if(length<=subLength) {
            return NULL; /* s is shorter than sub */
        }

        limit=s+length;

        /* the substring must start before preLimit */
        preLimit=limit-subLength;

        while(s!=preLimit) {
            c=*s++;
            if(c==cs) {
                /* found first substring UChar, compare rest */
                p=s;
                q=sub;
                for(;;) {
                    if(q==subLimit) {
                        if(isMatchAtCPBoundary(start, s-1, p, limit)) {
                            return (UChar *)(s-1); /* well-formed match */
                        } else {
                            break; /* no match because surrogate pair is split */
                        }
                    }
                    if(*p!=*q) {
                        break; /* no match */
                    }
                    ++p;
                    ++q;
                }
            }
        }
    }

    /* not found */
    return NULL;
}

U_CAPI UChar * U_EXPORT2
u_strstr(const UChar *s, const UChar *substring) {
    return u_strFindFirst(s, -1, substring, -1);
}

U_CAPI UChar * U_EXPORT2
u_strchr(const UChar *s, UChar c) {
    if(U16_IS_SURROGATE(c)) {
        /* make sure to not find half of a surrogate pair */
        return u_strFindFirst(s, -1, &c, 1);
    } else {
        UChar cs;

        /* trivial search for a BMP code point */
        for(;;) {
            if((cs=*s)==c) {
                return (UChar *)s;
            }
            if(cs==0) {
                return NULL;
            }
            ++s;
        }
    }
}

U_CAPI UChar * U_EXPORT2
u_strchr32(const UChar *s, UChar32 c) {
    if((uint32_t)c<=U_BMP_MAX) {
        /* find BMP code point */
        return u_strchr(s, (UChar)c);
    } else if((uint32_t)c<=UCHAR_MAX_VALUE) {
        /* find supplementary code point as surrogate pair */
        UChar cs, lead=U16_LEAD(c), trail=U16_TRAIL(c);

        while((cs=*s++)!=0) {
            if(cs==lead && *s==trail) {
                return (UChar *)(s-1);
            }
        }
        return NULL;
    } else {
        /* not a Unicode code point, not findable */
        return NULL;
    }
}

U_CAPI UChar * U_EXPORT2
u_memchr(const UChar *s, UChar c, int32_t count) {
    if(count<=0) {
        return NULL; /* no string */
    } else if(U16_IS_SURROGATE(c)) {
        /* make sure to not find half of a surrogate pair */
        return u_strFindFirst(s, count, &c, 1);
    } else {
        /* trivial search for a BMP code point */
        const UChar *limit=s+count;
        do {
            if(*s==c) {
                return (UChar *)s;
            }
        } while(++s!=limit);
        return NULL;
    }
}

U_CAPI UChar * U_EXPORT2
u_memchr32(const UChar *s, UChar32 c, int32_t count) {
    if((uint32_t)c<=U_BMP_MAX) {
        /* find BMP code point */
        return u_memchr(s, (UChar)c, count);
    } else if(count<2) {
        /* too short for a surrogate pair */
        return NULL;
    } else if((uint32_t)c<=UCHAR_MAX_VALUE) {
        /* find supplementary code point as surrogate pair */
        const UChar *limit=s+count-1; /* -1 so that we do not need a separate check for the trail unit */
        UChar lead=U16_LEAD(c), trail=U16_TRAIL(c);

        do {
            if(*s==lead && *(s+1)==trail) {
                return (UChar *)s;
            }
        } while(++s!=limit);
        return NULL;
    } else {
        /* not a Unicode code point, not findable */
        return NULL;
    }
}

/* Backward binary string search functions ---------------------------------- */

U_CAPI UChar * U_EXPORT2
u_strFindLast(const UChar *s, int32_t length,
              const UChar *sub, int32_t subLength) {
    const UChar *start, *limit, *p, *q, *subLimit;
    UChar c, cs;

    if(sub==NULL || subLength<-1) {
        return (UChar *)s;
    }
    if(s==NULL || length<-1) {
        return NULL;
    }

    /*
     * This implementation is more lazy than the one for u_strFindFirst():
     * There is no special search code for NUL-terminated strings.
     * It does not seem to be worth it for searching substrings to
     * search forward and find all matches like in u_strrchr() and similar.
     * Therefore, we simply get both string lengths and search backward.
     *
     * markus 2002oct23
     */

    if(subLength<0) {
        subLength=u_strlen(sub);
    }
    if(subLength==0) {
        return (UChar *)s;
    }

    /* get sub[subLength-1] to search for it fast */
    subLimit=sub+subLength;
    cs=*(--subLimit);
    --subLength;

    if(subLength==0 && !U16_IS_SURROGATE(cs)) {
        /* the substring consists of a single, non-surrogate BMP code point */
        return length<0 ? u_strrchr(s, cs) : u_memrchr(s, cs, length);
    }

    if(length<0) {
        length=u_strlen(s);
    }

    /* subLength was decremented above */
    if(length<=subLength) {
        return NULL; /* s is shorter than sub */
    }

    start=s;
    limit=s+length;

    /* the substring must start no later than s+subLength */
    s+=subLength;

    while(s!=limit) {
        c=*(--limit);
        if(c==cs) {
            /* found last substring UChar, compare rest */
            p=limit;
            q=subLimit;
            for(;;) {
                if(q==sub) {
                    if(isMatchAtCPBoundary(start, p, limit+1, start+length)) {
                        return (UChar *)p; /* well-formed match */
                    } else {
                        break; /* no match because surrogate pair is split */
                    }
                }
                if(*(--p)!=*(--q)) {
                    break; /* no match */
                }
            }
        }
    }

    /* not found */
    return NULL;
}

U_CAPI UChar * U_EXPORT2
u_strrstr(const UChar *s, const UChar *substring) {
    return u_strFindLast(s, -1, substring, -1);
}

U_CAPI UChar * U_EXPORT2
u_strrchr(const UChar *s, UChar c) {
    if(U16_IS_SURROGATE(c)) {
        /* make sure to not find half of a surrogate pair */
        return u_strFindLast(s, -1, &c, 1);
    } else {
        const UChar *result=NULL;
        UChar cs;

        /* trivial search for a BMP code point */
        for(;;) {
            if((cs=*s)==c) {
                result=s;
            }
            if(cs==0) {
                return (UChar *)result;
            }
            ++s;
        }
    }
}

U_CAPI UChar * U_EXPORT2
u_strrchr32(const UChar *s, UChar32 c) {
    if((uint32_t)c<=U_BMP_MAX) {
        /* find BMP code point */
        return u_strrchr(s, (UChar)c);
    } else if((uint32_t)c<=UCHAR_MAX_VALUE) {
        /* find supplementary code point as surrogate pair */
        const UChar *result=NULL;
        UChar cs, lead=U16_LEAD(c), trail=U16_TRAIL(c);

        while((cs=*s++)!=0) {
            if(cs==lead && *s==trail) {
                result=s-1;
            }
        }
        return (UChar *)result;
    } else {
        /* not a Unicode code point, not findable */
        return NULL;
    }
}

U_CAPI UChar * U_EXPORT2
u_memrchr(const UChar *s, UChar c, int32_t count) {
    if(count<=0) {
        return NULL; /* no string */
    } else if(U16_IS_SURROGATE(c)) {
        /* make sure to not find half of a surrogate pair */
        return u_strFindLast(s, count, &c, 1);
    } else {
        /* trivial search for a BMP code point */
        const UChar *limit=s+count;
        do {
            if(*(--limit)==c) {
                return (UChar *)limit;
            }
        } while(s!=limit);
        return NULL;
    }
}

U_CAPI UChar * U_EXPORT2
u_memrchr32(const UChar *s, UChar32 c, int32_t count) {
    if((uint32_t)c<=U_BMP_MAX) {
        /* find BMP code point */
        return u_memrchr(s, (UChar)c, count);
    } else if(count<2) {
        /* too short for a surrogate pair */
        return NULL;
    } else if((uint32_t)c<=UCHAR_MAX_VALUE) {
        /* find supplementary code point as surrogate pair */
        const UChar *limit=s+count-1;
        UChar lead=U16_LEAD(c), trail=U16_TRAIL(c);

        do {
            if(*limit==trail && *(limit-1)==lead) {
                return (UChar *)(limit-1);
            }
        } while(s!=--limit);
        return NULL;
    } else {
        /* not a Unicode code point, not findable */
        return NULL;
    }
}

/* Tokenization functions --------------------------------------------------- */

/*
 * Match each code point in a string against each code point in the matchSet.
 * Return the index of the first string code point that
 * is (polarity==TRUE) or is not (FALSE) contained in the matchSet.
 * Return -(string length)-1 if there is no such code point.
 */
static int32_t
_matchFromSet(const UChar *string, const UChar *matchSet, UBool polarity) {
    int32_t matchLen, matchBMPLen, strItr, matchItr;
    UChar32 stringCh, matchCh;
    UChar c, c2;

    /* first part of matchSet contains only BMP code points */
    matchBMPLen = 0;
    while((c = matchSet[matchBMPLen]) != 0 && U16_IS_SINGLE(c)) {
        ++matchBMPLen;
    }

    /* second part of matchSet contains BMP and supplementary code points */
    matchLen = matchBMPLen;
    while(matchSet[matchLen] != 0) {
        ++matchLen;
    }

    for(strItr = 0; (c = string[strItr]) != 0;) {
        ++strItr;
        if(U16_IS_SINGLE(c)) {
            if(polarity) {
                for(matchItr = 0; matchItr < matchLen; ++matchItr) {
                    if(c == matchSet[matchItr]) {
                        return strItr - 1; /* one matches */
                    }
                }
            } else {
                for(matchItr = 0; matchItr < matchLen; ++matchItr) {
                    if(c == matchSet[matchItr]) {
                        goto endloop;
                    }
                }
                return strItr - 1; /* none matches */
            }
        } else {
            /*
             * No need to check for string length before U16_IS_TRAIL
             * because c2 could at worst be the terminating NUL.
             */
            if(U16_IS_SURROGATE_LEAD(c) && U16_IS_TRAIL(c2 = string[strItr])) {
                ++strItr;
                stringCh = U16_GET_SUPPLEMENTARY(c, c2);
            } else {
                stringCh = c; /* unpaired trail surrogate */
            }

            if(polarity) {
                for(matchItr = matchBMPLen; matchItr < matchLen;) {
                    U16_NEXT(matchSet, matchItr, matchLen, matchCh);
                    if(stringCh == matchCh) {
                        return strItr - U16_LENGTH(stringCh); /* one matches */
                    }
                }
            } else {
                for(matchItr = matchBMPLen; matchItr < matchLen;) {
                    U16_NEXT(matchSet, matchItr, matchLen, matchCh);
                    if(stringCh == matchCh) {
                        goto endloop;
                    }
                }
                return strItr - U16_LENGTH(stringCh); /* none matches */
            }
        }
endloop:
        /* wish C had continue with labels like Java... */;
    }

    /* Didn't find it. */
    return -strItr-1;
}

/* Search for a codepoint in a string that matches one of the matchSet codepoints. */
U_CAPI UChar * U_EXPORT2
u_strpbrk(const UChar *string, const UChar *matchSet)
{
    int32_t idx = _matchFromSet(string, matchSet, TRUE);
    if(idx >= 0) {
        return (UChar *)string + idx;
    } else {
        return NULL;
    }
}

/* Search for a codepoint in a string that matches one of the matchSet codepoints. */
U_CAPI int32_t U_EXPORT2
u_strcspn(const UChar *string, const UChar *matchSet)
{
    int32_t idx = _matchFromSet(string, matchSet, TRUE);
    if(idx >= 0) {
        return idx;
    } else {
        return -idx - 1; /* == u_strlen(string) */
    }
}

/* Search for a codepoint in a string that does not match one of the matchSet codepoints. */
U_CAPI int32_t U_EXPORT2
u_strspn(const UChar *string, const UChar *matchSet)
{
    int32_t idx = _matchFromSet(string, matchSet, FALSE);
    if(idx >= 0) {
        return idx;
    } else {
        return -idx - 1; /* == u_strlen(string) */
    }
}

/* ----- Text manipulation functions --- */

U_CAPI UChar* U_EXPORT2
u_strtok_r(UChar    *src, 
     const UChar    *delim,
           UChar   **saveState)
{
    UChar *tokSource;
    UChar *nextToken;
    uint32_t nonDelimIdx;

    /* If saveState is NULL, the user messed up. */
    if (src != NULL) {
        tokSource = src;
        *saveState = src; /* Set to "src" in case there are no delimiters */
    }
    else if (*saveState) {
        tokSource = *saveState;
    }
    else {
        /* src == NULL && *saveState == NULL */
        /* This shouldn't happen. We already finished tokenizing. */
        return NULL;
    }

    /* Skip initial delimiters */
    nonDelimIdx = u_strspn(tokSource, delim);
    tokSource = &tokSource[nonDelimIdx];

    if (*tokSource) {
        nextToken = u_strpbrk(tokSource, delim);
        if (nextToken != NULL) {
            /* Create a token */
            *(nextToken++) = 0;
            *saveState = nextToken;
            return tokSource;
        }
        else if (*saveState) {
            /* Return the last token */
            *saveState = NULL;
            return tokSource;
        }
    }
    else {
        /* No tokens were found. Only delimiters were left. */
        *saveState = NULL;
    }
    return NULL;
}

/* Miscellaneous functions -------------------------------------------------- */

U_CAPI UChar* U_EXPORT2
u_strcat(UChar     *dst, 
    const UChar     *src)
{
    UChar *anchor = dst;            /* save a pointer to start of dst */

    while(*dst != 0) {              /* To end of first string          */
        ++dst;
    }
    while((*(dst++) = *(src++)) != 0) {     /* copy string 2 over              */
    }

    return anchor;
}

U_CAPI UChar*  U_EXPORT2
u_strncat(UChar     *dst, 
     const UChar     *src, 
     int32_t     n ) 
{
    if(n > 0) {
        UChar *anchor = dst;            /* save a pointer to start of dst */

        while(*dst != 0) {              /* To end of first string          */
            ++dst;
        }
        while((*dst = *src) != 0) {     /* copy string 2 over              */
            ++dst;
            if(--n == 0) {
                *dst = 0;
                break;
            }
            ++src;
        }

        return anchor;
    } else {
        return dst;
    }
}

/* ----- Text property functions --- */

U_CAPI int32_t   U_EXPORT2
u_strcmp(const UChar *s1, 
    const UChar *s2) 
{
    UChar  c1, c2;

    for(;;) {
        c1=*s1++;
        c2=*s2++;
        if (c1 != c2 || c1 == 0) {
            break;
        }
    }
    return (int32_t)c1 - (int32_t)c2;
}

U_CFUNC int32_t U_EXPORT2
uprv_strCompare(const UChar *s1, int32_t length1,
                const UChar *s2, int32_t length2,
                UBool strncmpStyle, UBool codePointOrder) {
    const UChar *start1, *start2, *limit1, *limit2;
    UChar c1, c2;

    /* setup for fix-up */
    start1=s1;
    start2=s2;

    /* compare identical prefixes - they do not need to be fixed up */
    if(length1<0 && length2<0) {
        /* strcmp style, both NUL-terminated */
        if(s1==s2) {
            return 0;
        }

        for(;;) {
            c1=*s1;
            c2=*s2;
            if(c1!=c2) {
                break;
            }
            if(c1==0) {
                return 0;
            }
            ++s1;
            ++s2;
        }

        /* setup for fix-up */
        limit1=limit2=NULL;
    } else if(strncmpStyle) {
        /* special handling for strncmp, assume length1==length2>=0 but also check for NUL */
        if(s1==s2) {
            return 0;
        }

        limit1=start1+length1;

        for(;;) {
            /* both lengths are same, check only one limit */
            if(s1==limit1) {
                return 0;
            }

            c1=*s1;
            c2=*s2;
            if(c1!=c2) {
                break;
            }
            if(c1==0) {
                return 0;
            }
            ++s1;
            ++s2;
        }

        /* setup for fix-up */
        limit2=start2+length1; /* use length1 here, too, to enforce assumption */
    } else {
        /* memcmp/UnicodeString style, both length-specified */
        int32_t lengthResult;

        if(length1<0) {
            length1=u_strlen(s1);
        }
        if(length2<0) {
            length2=u_strlen(s2);
        }

        /* limit1=start1+min(lenght1, length2) */
        if(length1<length2) {
            lengthResult=-1;
            limit1=start1+length1;
        } else if(length1==length2) {
            lengthResult=0;
            limit1=start1+length1;
        } else /* length1>length2 */ {
            lengthResult=1;
            limit1=start1+length2;
        }

        if(s1==s2) {
            return lengthResult;
        }

        for(;;) {
            /* check pseudo-limit */
            if(s1==limit1) {
                return lengthResult;
            }

            c1=*s1;
            c2=*s2;
            if(c1!=c2) {
                break;
            }
            ++s1;
            ++s2;
        }

        /* setup for fix-up */
        limit1=start1+length1;
        limit2=start2+length2;
    }

    /* if both values are in or above the surrogate range, fix them up */
    if(c1>=0xd800 && c2>=0xd800 && codePointOrder) {
        /* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
        if(
            (c1<=0xdbff && (s1+1)!=limit1 && U16_IS_TRAIL(*(s1+1))) ||
            (U16_IS_TRAIL(c1) && start1!=s1 && U16_IS_LEAD(*(s1-1)))
        ) {
            /* part of a surrogate pair, leave >=d800 */
        } else {
            /* BMP code point - may be surrogate code point - make <d800 */
            c1-=0x2800;
        }

        if(
            (c2<=0xdbff && (s2+1)!=limit2 && U16_IS_TRAIL(*(s2+1))) ||
            (U16_IS_TRAIL(c2) && start2!=s2 && U16_IS_LEAD(*(s2-1)))
        ) {
            /* part of a surrogate pair, leave >=d800 */
        } else {
            /* BMP code point - may be surrogate code point - make <d800 */
            c2-=0x2800;
        }
    }

    /* now c1 and c2 are in the requested (code unit or code point) order */
    return (int32_t)c1-(int32_t)c2;
}

/*
 * Compare two strings as presented by UCharIterators.
 * Use code unit or code point order.
 * When the function returns, it is undefined where the iterators
 * have stopped.
 */
U_CAPI int32_t U_EXPORT2
u_strCompareIter(UCharIterator *iter1, UCharIterator *iter2, UBool codePointOrder) {
    UChar32 c1, c2;

    /* argument checking */
    if(iter1==NULL || iter2==NULL) {
        return 0; /* bad arguments */
    }
    if(iter1==iter2) {
        return 0; /* identical iterators */
    }

    /* reset iterators to start? */
    iter1->move(iter1, 0, UITER_START);
    iter2->move(iter2, 0, UITER_START);

    /* compare identical prefixes - they do not need to be fixed up */
    for(;;) {
        c1=iter1->next(iter1);
        c2=iter2->next(iter2);
        if(c1!=c2) {
            break;
        }
        if(c1==-1) {
            return 0;
        }
    }

    /* if both values are in or above the surrogate range, fix them up */
    if(c1>=0xd800 && c2>=0xd800 && codePointOrder) {
        /* subtract 0x2800 from BMP code points to make them smaller than supplementary ones */
        if(
            (c1<=0xdbff && U16_IS_TRAIL(iter1->current(iter1))) ||
            (U16_IS_TRAIL(c1) && (iter1->previous(iter1), U16_IS_LEAD(iter1->previous(iter1))))
        ) {
            /* part of a surrogate pair, leave >=d800 */
        } else {
            /* BMP code point - may be surrogate code point - make <d800 */
            c1-=0x2800;
        }

        if(
            (c2<=0xdbff && U16_IS_TRAIL(iter2->current(iter2))) ||
            (U16_IS_TRAIL(c2) && (iter2->previous(iter2), U16_IS_LEAD(iter2->previous(iter2))))
        ) {
            /* part of a surrogate pair, leave >=d800 */
        } else {
            /* BMP code point - may be surrogate code point - make <d800 */
            c2-=0x2800;
        }
    }

    /* now c1 and c2 are in the requested (code unit or code point) order */
    return (int32_t)c1-(int32_t)c2;
}

#if 0
/*
 * u_strCompareIter() does not leave the iterators _on_ the different units.
 * This is possible but would cost a few extra indirect function calls to back
 * up if the last unit (c1 or c2 respectively) was >=0.
 *
 * Consistently leaving them _behind_ the different units is not an option
 * because the current "unit" is the end of the string if that is reached,
 * and in such a case the iterator does not move.
 * For example, when comparing "ab" with "abc", both iterators rest _on_ the end
 * of their strings. Calling previous() on each does not move them to where
 * the comparison fails.
 *
 * So the simplest semantics is to not define where the iterators end up.
 *
 * The following fragment is part of what would need to be done for backing up.
 */
void fragment {
        /* iff a surrogate is part of a surrogate pair, leave >=d800 */
        if(c1<=0xdbff) {
            if(!U16_IS_TRAIL(iter1->current(iter1))) {
                /* lead surrogate code point - make <d800 */
                c1-=0x2800;
            }
        } else if(c1<=0xdfff) {
            int32_t idx=iter1->getIndex(iter1, UITER_CURRENT);
            iter1->previous(iter1); /* ==c1 */
            if(!U16_IS_LEAD(iter1->previous(iter1))) {
                /* trail surrogate code point - make <d800 */
                c1-=0x2800;
            }
            /* go back to behind where the difference is */
            iter1->move(iter1, idx, UITER_ZERO);
        } else /* 0xe000<=c1<=0xffff */ {
            /* BMP code point - make <d800 */
            c1-=0x2800;
        }
}
#endif

U_CAPI int32_t U_EXPORT2
u_strCompare(const UChar *s1, int32_t length1,
             const UChar *s2, int32_t length2,
             UBool codePointOrder) {
    /* argument checking */
    if(s1==NULL || length1<-1 || s2==NULL || length2<-1) {
        return 0;
    }
    return uprv_strCompare(s1, length1, s2, length2, FALSE, codePointOrder);
}

/* String compare in code point order - u_strcmp() compares in code unit order. */
U_CAPI int32_t U_EXPORT2
u_strcmpCodePointOrder(const UChar *s1, const UChar *s2) {
    return uprv_strCompare(s1, -1, s2, -1, FALSE, TRUE);
}

U_CAPI int32_t   U_EXPORT2
u_strncmp(const UChar     *s1, 
     const UChar     *s2, 
     int32_t     n) 
{
    if(n > 0) {
        int32_t rc;
        for(;;) {
            rc = (int32_t)*s1 - (int32_t)*s2;
            if(rc != 0 || *s1 == 0 || --n == 0) {
                return rc;
            }
            ++s1;
            ++s2;
        }
    } else {
        return 0;
    }
}

U_CAPI int32_t U_EXPORT2
u_strncmpCodePointOrder(const UChar *s1, const UChar *s2, int32_t n) {
    return uprv_strCompare(s1, n, s2, n, TRUE, TRUE);
}

U_CAPI UChar* U_EXPORT2
u_strcpy(UChar     *dst, 
    const UChar     *src) 
{
    UChar *anchor = dst;            /* save a pointer to start of dst */

    while((*(dst++) = *(src++)) != 0) {     /* copy string 2 over              */
    }

    return anchor;
}

U_CAPI UChar*  U_EXPORT2
u_strncpy(UChar     *dst, 
     const UChar     *src, 
     int32_t     n) 
{
    UChar *anchor = dst;            /* save a pointer to start of dst */

    /* copy string 2 over */
    while(n > 0 && (*(dst++) = *(src++)) != 0) {
        --n;
    }

    return anchor;
}

U_CAPI int32_t   U_EXPORT2
u_strlen(const UChar *s) 
{
#if U_SIZEOF_WCHAR_T == U_SIZEOF_UCHAR
    return (int32_t)uprv_wcslen((const wchar_t *)s);
#else
    const UChar *t = s;
    while(*t != 0) {
      ++t;
    }
    return t - s;
#endif
}

U_CAPI int32_t U_EXPORT2
u_countChar32(const UChar *s, int32_t length) {
    int32_t count;

    if(s==NULL || length<-1) {
        return 0;
    }

    count=0;
    if(length>=0) {
        while(length>0) {
            ++count;
            if(U16_IS_LEAD(*s) && length>=2 && U16_IS_TRAIL(*(s+1))) {
                s+=2;
                length-=2;
            } else {
                ++s;
                --length;
            }
        }
    } else /* length==-1 */ {
        UChar c;

        for(;;) {
            if((c=*s++)==0) {
                break;
            }
            ++count;

            /*
             * sufficient to look ahead one because of UTF-16;
             * safe to look ahead one because at worst that would be the terminating NUL
             */
            if(U16_IS_LEAD(c) && U16_IS_TRAIL(*s)) {
                ++s;
            }
        }
    }
    return count;
}

U_CAPI UBool U_EXPORT2
u_strHasMoreChar32Than(const UChar *s, int32_t length, int32_t number) {

    if(number<0) {
        return TRUE;
    }
    if(s==NULL || length<-1) {
        return FALSE;
    }

    if(length==-1) {
        /* s is NUL-terminated */
        UChar c;

        /* count code points until they exceed */
        for(;;) {
            if((c=*s++)==0) {
                return FALSE;
            }
            if(number==0) {
                return TRUE;
            }
            if(U16_IS_LEAD(c) && U16_IS_TRAIL(*s)) {
                ++s;
            }
            --number;
        }
    } else {
        /* length>=0 known */
        const UChar *limit;
        int32_t maxSupplementary;

        /* s contains at least (length+1)/2 code points: <=2 UChars per cp */
        if(((length+1)/2)>number) {
            return TRUE;
        }

        /* check if s does not even contain enough UChars */
        maxSupplementary=length-number;
        if(maxSupplementary<=0) {
            return FALSE;
        }
        /* there are maxSupplementary=length-number more UChars than asked-for code points */

        /*
         * count code points until they exceed and also check that there are
         * no more than maxSupplementary supplementary code points (UChar pairs)
         */
        limit=s+length;
        for(;;) {
            if(s==limit) {
                return FALSE;
            }
            if(number==0) {
                return TRUE;
            }
            if(U16_IS_LEAD(*s++) && s!=limit && U16_IS_TRAIL(*s)) {
                ++s;
                if(--maxSupplementary<=0) {
                    /* too many pairs - too few code points */
                    return FALSE;
                }
            }
            --number;
        }
    }
}

/* ----- String validation functions --- */

/*
 * Check whether the string is well-formed according to various criteria:
 * - No code points that are defined as non-characters (e.g. 0xFFFF) or are undefined in
 *   the version of Unicode currently supported.
 * - No isolated surrogate code points.
 * - No overly-long sequences of non-starter combining marks, i.e. more than 30 characters
 *   in a row with non-zero combining class (which may have category Mn or Mc); this
 *   violates Stream-Safe Text Format per UAX #15. This test does not ensure that the
 *   string satisfies Stream-Safe Text Format (because it does not convert to NFKC first),
 *   but any string that fails this test is certainly not Stream-Safe.
 * - No emoji variation selectors applied to non-emoji code points. This function may
 *   also check for other non-standard variation sequences.
 * - No tag sequences that are ill-formed per definition ED-14a in UTS #51 (e.g. tag
 *   sequences must have an emoji base and a terminator).
 *
 * @internal Apple only 
 */
enum { kBidiMaxDepth = 125 };

static UBool isWellFormed(UChar32 c, UChar32 cLast, int32_t *nonStarterCountP, UBool *inTagSeqP,
                        uint8_t* dirStatus, int32_t* dirStatusIndexP, int32_t* validIsolateCountP) {
    if (*inTagSeqP) {
        // can only have tag_spec or tag_term
        if (c == 0xE007F) { // tag_term
            *inTagSeqP = FALSE;
        } else if (c < 0xE0020 || c > 0xE007E) {
            return FALSE;
        }
    } else if (c < 0x0300) {
        // Everything in this range (includes ASCII) is a valid character with combining class 0
        *nonStarterCountP = 0;
        if (c == 0x000A || c == 0x000D || c == 0x0085 || (c >= 0x001C && c <= 0x001E)) {
            // paragraph sep, reset bidi
            *dirStatusIndexP = 0;
            *validIsolateCountP = 0;
        }
    } else if ((c >= 0x2029 && c <= 0x202E) || (c >= 0x2066 && c <= 0x2069)) {
        // para sep & bidi controls, all have combining class 0. The bidi control actions here
        // are from [https://www.unicode.org/reports/tr9/#Explicit_Levels_and_Directions]
        *nonStarterCountP = 0;
        if (c == 0x2029) { // paragraph sep, reset bidi
            *dirStatusIndexP = 0;
            *validIsolateCountP = 0;
        } else if (c == 0x2069) { // PDI
            if (*validIsolateCountP > 0) {
                while (*dirStatusIndexP > 0 && (dirStatus[(*dirStatusIndexP)--] & 0x80) == 0);
                (*validIsolateCountP)--;
            }
        } else if (c == 0x202C) { // PDF
            if (*dirStatusIndexP > 0 && (dirStatus[*dirStatusIndexP] & 0x80) == 0) {
                (*dirStatusIndexP)--;
            }
        } else {
            // embedding/override initiator. Need to increment the level by at least 1, and possibly 2 if the
            // embedding/override direction matches the current direction (i.e. R and current odd, or L and current even).
            // Since we increment first, the test for odd/even is flipped. For FSI, we do not actually determine
            // whether it should be treated as RLI or LRI, so we just do the minimum increment.
            uint8_t newEntry = (dirStatus[*dirStatusIndexP] & 0x7F) + 1; // min increment, flips odd/even status compared to current
            if ( ((c == 0x202B || c == 0x202E || c == 0x2067) && (newEntry & 0x01) == 0) ||  // RLE/RLO/RLI and current was odd
                 ((c == 0x202A || c == 0x202D || c == 0x2066) && (newEntry & 0x01) != 0) ) { // LRE/LRO/LRI and current was even
                newEntry++;
            }
            if (newEntry > kBidiMaxDepth || *dirStatusIndexP > kBidiMaxDepth) {
                return FALSE; // Checking for this is the whole point.
            }
            if (c >= 0x2066 &&  c <= 0x2068) { // LRI/RLI/FSI
                newEntry |= 0x80; // set directional isolate status
                (*validIsolateCountP)++;
            }
            dirStatus[++(*dirStatusIndexP)] = newEntry;
        }
    } else if (c == 0xFE0F) { // emoji variation selector
        if (!u_isEmoji(cLast)) { // previous char must be emoji
            return FALSE;
        }
        // previous character would have set *nonStarterCountP = 0;
    } else if (c >= 0xE0020 && c <= 0xE007E) { // tag_spec
        if (!u_isEmoji(cLast) && cLast != 0xFE0F) { // previous char must be emoji or FE0F
              return FALSE;
        } 
        *inTagSeqP = TRUE;
        // previous character would have set  *nonStarterCountP = 0;
    } else if (c == 0xE007F) { // tag_term
         return FALSE;
    } else {
        // we have checked specific ranges/chars, now check general info for others
        int8_t genCat = u_charType(c);
        if (genCat == U_UNASSIGNED || genCat == U_SURROGATE) {
            return FALSE;
        }
        if ((genCat == U_NON_SPACING_MARK || genCat == U_COMBINING_SPACING_MARK) && u_getCombiningClass(c) != 0) {
            // non-starter
            if (++(*nonStarterCountP) > 30) {
                return FALSE;
            }
        } else {
             *nonStarterCountP = 0;
        }
    }
    return TRUE;
}

U_CAPI UBool U_EXPORT2
u_strIsWellFormed(const UChar *s, int32_t length) {
    if (s==NULL || length<-1) {
        return FALSE;
    }
    UChar32 c, c2, cLast = 0;
    int32_t nonStarterCount = 0;
    UBool inTagSeq = FALSE;
    uint8_t dirStatus[kBidiMaxDepth + 3]; // low 7 bits is embed level, high bit is direction override status
    int32_t dirStatusIndex = 0;
    int32_t validIsolateCount = 0;
    dirStatus[0] = 0; // assume initial paragraph direction L (most conservative)
    if (length < 0) {
        // NUL terminated
        while ((c = *s++) != 0) {
            // get next UChar32 c
            if (U16_IS_LEAD(c)) {
                if (U16_IS_TRAIL(c2 = *s)) {
                    s++;
                    c = U16_GET_SUPPLEMENTARY(c,c2);
                }
            }
            // check current c
            if (!isWellFormed(c, cLast, &nonStarterCount, &inTagSeq, dirStatus, &dirStatusIndex, &validIsolateCount)) {
                return FALSE;
            }
            // setup next iteration
            cLast = c;
        }
    } else {
        // use length
        const UChar *sLimit = s + length;
        while (s < sLimit) {
            // get next UChar32 c
            c = *s++;
            if (U16_IS_LEAD(c)) {
                if (s < sLimit && U16_IS_TRAIL(c2 = *s)) {
                    s++;
                    c = U16_GET_SUPPLEMENTARY(c,c2);
                }
            }
            // check current c
            if (!isWellFormed(c, cLast, &nonStarterCount, &inTagSeq, dirStatus, &dirStatusIndex, &validIsolateCount)) {
                return FALSE;
            }
            // setup next iteration
            cLast = c;
        }
    }
    return TRUE;
}

/* ----- U_mem functions --- */

U_CAPI UChar * U_EXPORT2
u_memcpy(UChar *dest, const UChar *src, int32_t count) {
    if(count > 0) {
        uprv_memcpy(dest, src, (size_t)count*U_SIZEOF_UCHAR);
    }
    return dest;
}

U_CAPI UChar * U_EXPORT2
u_memmove(UChar *dest, const UChar *src, int32_t count) {
    if(count > 0) {
        uprv_memmove(dest, src, (size_t)count*U_SIZEOF_UCHAR);
    }
    return dest;
}

U_CAPI UChar * U_EXPORT2
u_memset(UChar *dest, UChar c, int32_t count) {
    if(count > 0) {
        UChar *ptr = dest;
        UChar *limit = dest + count;

        while (ptr < limit) {
            *(ptr++) = c;
        }
    }
    return dest;
}

U_CAPI int32_t U_EXPORT2
u_memcmp(const UChar *buf1, const UChar *buf2, int32_t count) {
    if(count > 0) {
        const UChar *limit = buf1 + count;
        int32_t result;

        while (buf1 < limit) {
            result = (int32_t)(uint16_t)*buf1 - (int32_t)(uint16_t)*buf2;
            if (result != 0) {
                return result;
            }
            buf1++;
            buf2++;
        }
    }
    return 0;
}

U_CAPI int32_t U_EXPORT2
u_memcmpCodePointOrder(const UChar *s1, const UChar *s2, int32_t count) {
    return uprv_strCompare(s1, count, s2, count, FALSE, TRUE);
}

/* u_unescape & support fns ------------------------------------------------- */

/* This map must be in ASCENDING ORDER OF THE ESCAPE CODE */
static const UChar UNESCAPE_MAP[] = {
    /*"   0x22, 0x22 */
    /*'   0x27, 0x27 */
    /*?   0x3F, 0x3F */
    /*\   0x5C, 0x5C */
    /*a*/ 0x61, 0x07,
    /*b*/ 0x62, 0x08,
    /*e*/ 0x65, 0x1b,
    /*f*/ 0x66, 0x0c,
    /*n*/ 0x6E, 0x0a,
    /*r*/ 0x72, 0x0d,
    /*t*/ 0x74, 0x09,
    /*v*/ 0x76, 0x0b
};
enum { UNESCAPE_MAP_LENGTH = UPRV_LENGTHOF(UNESCAPE_MAP) };

/* Convert one octal digit to a numeric value 0..7, or -1 on failure */
static int8_t _digit8(UChar c) {
    if (c >= 0x0030 && c <= 0x0037) {
        return (int8_t)(c - 0x0030);
    }
    return -1;
}

/* Convert one hex digit to a numeric value 0..F, or -1 on failure */
static int8_t _digit16(UChar c) {
    if (c >= 0x0030 && c <= 0x0039) {
        return (int8_t)(c - 0x0030);
    }
    if (c >= 0x0041 && c <= 0x0046) {
        return (int8_t)(c - (0x0041 - 10));
    }
    if (c >= 0x0061 && c <= 0x0066) {
        return (int8_t)(c - (0x0061 - 10));
    }
    return -1;
}

/* Parse a single escape sequence.  Although this method deals in
 * UChars, it does not use C++ or UnicodeString.  This allows it to
 * be used from C contexts. */
U_CAPI UChar32 U_EXPORT2
u_unescapeAt(UNESCAPE_CHAR_AT charAt,
             int32_t *offset,
             int32_t length,
             void *context) {

    int32_t start = *offset;
    UChar c;
    UChar32 result = 0;
    int8_t n = 0;
    int8_t minDig = 0;
    int8_t maxDig = 0;
    int8_t bitsPerDigit = 4; 
    int8_t dig;
    int32_t i;
    UBool braces = FALSE;

    /* Check that offset is in range */
    if (*offset < 0 || *offset >= length) {
        goto err;
    }

    /* Fetch first UChar after '\\' */
    c = charAt((*offset)++, context);

    /* Convert hexadecimal and octal escapes */
    switch (c) {
    case 0x0075 /*'u'*/:
        minDig = maxDig = 4;
        break;
    case 0x0055 /*'U'*/:
        minDig = maxDig = 8;
        break;
    case 0x0078 /*'x'*/:
        minDig = 1;
        if (*offset < length && charAt(*offset, context) == 0x7B /*{*/) {
            ++(*offset);
            braces = TRUE;
            maxDig = 8;
        } else {
            maxDig = 2;
        }
        break;
    default:
        dig = _digit8(c);
        if (dig >= 0) {
            minDig = 1;
            maxDig = 3;
            n = 1; /* Already have first octal digit */
            bitsPerDigit = 3;
            result = dig;
        }
        break;
    }
    if (minDig != 0) {
        while (*offset < length && n < maxDig) {
            c = charAt(*offset, context);
            dig = (int8_t)((bitsPerDigit == 3) ? _digit8(c) : _digit16(c));
            if (dig < 0) {
                break;
            }
            result = (result << bitsPerDigit) | dig;
            ++(*offset);
            ++n;
        }
        if (n < minDig) {
            goto err;
        }
        if (braces) {
            if (c != 0x7D /*}*/) {
                goto err;
            }
            ++(*offset);
        }
        if (result < 0 || result >= 0x110000) {
            goto err;
        }
        /* If an escape sequence specifies a lead surrogate, see if
         * there is a trail surrogate after it, either as an escape or
         * as a literal.  If so, join them up into a supplementary.
         */
        if (*offset < length && U16_IS_LEAD(result)) {
            int32_t ahead = *offset + 1;
            c = charAt(*offset, context);
            if (c == 0x5C /*'\\'*/ && ahead < length) {
                c = (UChar) u_unescapeAt(charAt, &ahead, length, context);
            }
            if (U16_IS_TRAIL(c)) {
                *offset = ahead;
                result = U16_GET_SUPPLEMENTARY(result, c);
            }
        }
        return result;
    }

    /* Convert C-style escapes in table */
    for (i=0; i<UNESCAPE_MAP_LENGTH; i+=2) {
        if (c == UNESCAPE_MAP[i]) {
            return UNESCAPE_MAP[i+1];
        } else if (c < UNESCAPE_MAP[i]) {
            break;
        }
    }

    /* Map \cX to control-X: X & 0x1F */
    if (c == 0x0063 /*'c'*/ && *offset < length) {
        c = charAt((*offset)++, context);
        if (U16_IS_LEAD(c) && *offset < length) {
            UChar c2 = charAt(*offset, context);
            if (U16_IS_TRAIL(c2)) {
                ++(*offset);
                c = (UChar) U16_GET_SUPPLEMENTARY(c, c2); /* [sic] */
            }
        }
        return 0x1F & c;
    }

    /* If no special forms are recognized, then consider
     * the backslash to generically escape the next character.
     * Deal with surrogate pairs. */
    if (U16_IS_LEAD(c) && *offset < length) {
        UChar c2 = charAt(*offset, context);
        if (U16_IS_TRAIL(c2)) {
            ++(*offset);
            return U16_GET_SUPPLEMENTARY(c, c2);
        }
    }
    return c;

 err:
    /* Invalid escape sequence */
    *offset = start; /* Reset to initial value */
    return (UChar32)0xFFFFFFFF;
}

/* u_unescapeAt() callback to return a UChar from a char* */
static UChar U_CALLCONV
_charPtr_charAt(int32_t offset, void *context) {
    UChar c16;
    /* It would be more efficient to access the invariant tables
     * directly but there is no API for that. */
    u_charsToUChars(((char*) context) + offset, &c16, 1);
    return c16;
}

/* Append an escape-free segment of the text; used by u_unescape() */
static void _appendUChars(UChar *dest, int32_t destCapacity,
                          const char *src, int32_t srcLen) {
    if (destCapacity < 0) {
        destCapacity = 0;
    }
    if (srcLen > destCapacity) {
        srcLen = destCapacity;
    }
    u_charsToUChars(src, dest, srcLen);
}

/* Do an invariant conversion of char* -> UChar*, with escape parsing */
U_CAPI int32_t U_EXPORT2
u_unescape(const char *src, UChar *dest, int32_t destCapacity) {
    const char *segment = src;
    int32_t i = 0;
    char c;

    while ((c=*src) != 0) {
        /* '\\' intentionally written as compiler-specific
         * character constant to correspond to compiler-specific
         * char* constants. */
        if (c == '\\') {
            int32_t lenParsed = 0;
            UChar32 c32;
            if (src != segment) {
                if (dest != NULL) {
                    _appendUChars(dest + i, destCapacity - i,
                                  segment, (int32_t)(src - segment));
                }
                i += (int32_t)(src - segment);
            }
            ++src; /* advance past '\\' */
            c32 = (UChar32)u_unescapeAt(_charPtr_charAt, &lenParsed, (int32_t)uprv_strlen(src), (void*)src);
            if (lenParsed == 0) {
                goto err;
            }
            src += lenParsed; /* advance past escape seq. */
            if (dest != NULL && U16_LENGTH(c32) <= (destCapacity - i)) {
                U16_APPEND_UNSAFE(dest, i, c32);
            } else {
                i += U16_LENGTH(c32);
            }
            segment = src;
        } else {
            ++src;
        }
    }
    if (src != segment) {
        if (dest != NULL) {
            _appendUChars(dest + i, destCapacity - i,
                          segment, (int32_t)(src - segment));
        }
        i += (int32_t)(src - segment);
    }
    if (dest != NULL && i < destCapacity) {
        dest[i] = 0;
    }
    return i;

 err:
    if (dest != NULL && destCapacity > 0) {
        *dest = 0;
    }
    return 0;
}

/* NUL-termination of strings ----------------------------------------------- */

/**
 * NUL-terminate a string no matter what its type.
 * Set warning and error codes accordingly.
 */
#define __TERMINATE_STRING(dest, destCapacity, length, pErrorCode)      \
    if(pErrorCode!=NULL && U_SUCCESS(*pErrorCode)) {                    \
        /* not a public function, so no complete argument checking */   \
                                                                        \
        if(length<0) {                                                  \
            /* assume that the caller handles this */                   \
        } else if(length<destCapacity) {                                \
            /* NUL-terminate the string, the NUL fits */                \
            dest[length]=0;                                             \
            /* unset the not-terminated warning but leave all others */ \
            if(*pErrorCode==U_STRING_NOT_TERMINATED_WARNING) {          \
                *pErrorCode=U_ZERO_ERROR;                               \
            }                                                           \
        } else if(length==destCapacity) {                               \
            /* unable to NUL-terminate, but the string itself fit - set a warning code */ \
            *pErrorCode=U_STRING_NOT_TERMINATED_WARNING;                \
        } else /* length>destCapacity */ {                              \
            /* even the string itself did not fit - set an error code */ \
            *pErrorCode=U_BUFFER_OVERFLOW_ERROR;                        \
        }                                                               \
    }

U_CAPI int32_t U_EXPORT2
u_terminateUChars(UChar *dest, int32_t destCapacity, int32_t length, UErrorCode *pErrorCode) {
    __TERMINATE_STRING(dest, destCapacity, length, pErrorCode);
    return length;
}

U_CAPI int32_t U_EXPORT2
u_terminateChars(char *dest, int32_t destCapacity, int32_t length, UErrorCode *pErrorCode) {
    __TERMINATE_STRING(dest, destCapacity, length, pErrorCode);
    return length;
}

U_CAPI int32_t U_EXPORT2
u_terminateUChar32s(UChar32 *dest, int32_t destCapacity, int32_t length, UErrorCode *pErrorCode) {
    __TERMINATE_STRING(dest, destCapacity, length, pErrorCode);
    return length;
}

U_CAPI int32_t U_EXPORT2
u_terminateWChars(wchar_t *dest, int32_t destCapacity, int32_t length, UErrorCode *pErrorCode) {
    __TERMINATE_STRING(dest, destCapacity, length, pErrorCode);
    return length;
}

// Compute the hash code for a string -------------------------------------- ***

// Moved here from uhash.c so that UnicodeString::hashCode() does not depend
// on UHashtable code.

/*
  Compute the hash by iterating sparsely over about 32 (up to 63)
  characters spaced evenly through the string.  For each character,
  multiply the previous hash value by a prime number and add the new
  character in, like a linear congruential random number generator,
  producing a pseudorandom deterministic value well distributed over
  the output range. [LIU]
*/

#define STRING_HASH(TYPE, STR, STRLEN, DEREF) \
    uint32_t hash = 0;                        \
    const TYPE *p = (const TYPE*) STR;        \
    if (p != NULL) {                          \
        int32_t len = (int32_t)(STRLEN);      \
        int32_t inc = ((len - 32) / 32) + 1;  \
        const TYPE *limit = p + len;          \
        while (p<limit) {                     \
            hash = (hash * 37) + DEREF;       \
            p += inc;                         \
        }                                     \
    }                                         \
    return static_cast<int32_t>(hash)

/* Used by UnicodeString to compute its hashcode - Not public API. */
U_CAPI int32_t U_EXPORT2
ustr_hashUCharsN(const UChar *str, int32_t length) {
    STRING_HASH(UChar, str, length, *p);
}

U_CAPI int32_t U_EXPORT2
ustr_hashCharsN(const char *str, int32_t length) {
    STRING_HASH(uint8_t, str, length, *p);
}

U_CAPI int32_t U_EXPORT2
ustr_hashICharsN(const char *str, int32_t length) {
    STRING_HASH(char, str, length, (uint8_t)uprv_tolower(*p));
}