levenshtein.c

Go to the documentation of this file.

/*-------------------------------------------------------------------------
 *
 * levenshtein.c
 *    Levenshtein distance implementation.
 *
 * Original author:  Joe Conway <mail@joeconway.com>
 *
 * This file is included by varlena.c twice, to provide matching code for (1)
 * Levenshtein distance with custom costings, and (2) Levenshtein distance with
 * custom costings and a "max" value above which exact distances are not
 * interesting.  Before the inclusion, we rely on the presence of the inline
 * function rest_of_char_same().
 *
 * Written based on a description of the algorithm by Michael Gilleland found
 * at http://www.merriampark.com/ld.htm.  Also looked at levenshtein.c in the
 * PHP 4.0.6 distribution for inspiration.  Configurable penalty costs
 * extension is introduced by Volkan YAZICI <volkan.yazici@gmail.com.
 *
 * Copyright (c) 2001-2026, PostgreSQL Global Development Group
 *
 * IDENTIFICATION
 *  src/backend/utils/adt/levenshtein.c
 *
 *-------------------------------------------------------------------------
 */
#define MAX_LEVENSHTEIN_STRLEN      255
 
/*
 * Calculates Levenshtein distance metric between supplied strings, which are
 * not necessarily null-terminated.
 *
 * source: source string, of length slen bytes.
 * target: target string, of length tlen bytes.
 * ins_c, del_c, sub_c: costs to charge for character insertion, deletion,
 *      and substitution respectively; (1, 1, 1) costs suffice for common
 *      cases, but your mileage may vary.
 * max_d: if provided and >= 0, maximum distance we care about; see below.
 * trusted: caller is trusted and need not obey MAX_LEVENSHTEIN_STRLEN.
 *
 * One way to compute Levenshtein distance is to incrementally construct
 * an (m+1)x(n+1) matrix where cell (i, j) represents the minimum number
 * of operations required to transform the first i characters of s into
 * the first j characters of t.  The last column of the final row is the
 * answer.
 *
 * We use that algorithm here with some modification.  In lieu of holding
 * the entire array in memory at once, we'll just use two arrays of size
 * m+1 for storing accumulated values. At each step one array represents
 * the "previous" row and one is the "current" row of the notional large
 * array.
 *
 * If max_d >= 0, we only need to provide an accurate answer when that answer
 * is less than or equal to max_d.  From any cell in the matrix, there is
 * theoretical "minimum residual distance" from that cell to the last column
 * of the final row.  This minimum residual distance is zero when the
 * untransformed portions of the strings are of equal length (because we might
 * get lucky and find all the remaining characters matching) and is otherwise
 * based on the minimum number of insertions or deletions needed to make them
 * equal length.  The residual distance grows as we move toward the upper
 * right or lower left corners of the matrix.  When the max_d bound is
 * usefully tight, we can use this property to avoid computing the entirety
 * of each row; instead, we maintain a start_column and stop_column that
 * identify the portion of the matrix close to the diagonal which can still
 * affect the final answer.
 */
int
#ifdef LEVENSHTEIN_LESS_EQUAL
varstr_levenshtein_less_equal(const char *source, int slen,
                              const char *target, int tlen,
                              int ins_c, int del_c, int sub_c,
                              int max_d, bool trusted)
#else
varstr_levenshtein(const char *source, int slen,
                   const char *target, int tlen,
                   int ins_c, int del_c, int sub_c,
                   bool trusted)
#endif
{
    int         m,
                n;
    int        *prev;
    int        *curr;
    int        *s_char_len = NULL;
    int         j;
    const char *y;
    const char *send = source + slen;
    const char *tend = target + tlen;
 
    /*
     * For varstr_levenshtein_less_equal, we have real variables called
     * start_column and stop_column; otherwise it's just short-hand for 0 and
     * m.
     */
#ifdef LEVENSHTEIN_LESS_EQUAL
    int         start_column,
                stop_column;
 
#undef START_COLUMN
#undef STOP_COLUMN
#define START_COLUMN start_column
#define STOP_COLUMN stop_column
#else
#undef START_COLUMN
#undef STOP_COLUMN
#define START_COLUMN 0
#define STOP_COLUMN m
#endif
 
    /* Convert string lengths (in bytes) to lengths in characters */
    m = pg_mbstrlen_with_len(source, slen);
    n = pg_mbstrlen_with_len(target, tlen);
 
    /*
     * We can transform an empty s into t with n insertions, or a non-empty t
     * into an empty s with m deletions.
     */
    if (!m)
        return n * ins_c;
    if (!n)
        return m * del_c;
 
    /*
     * For security concerns, restrict excessive CPU+RAM usage. (This
     * implementation uses O(m) memory and has O(mn) complexity.)  If
     * "trusted" is true, caller is responsible for not making excessive
     * requests, typically by using a small max_d along with strings that are
     * bounded, though not necessarily to MAX_LEVENSHTEIN_STRLEN exactly.
     */
    if (!trusted &&
        (m > MAX_LEVENSHTEIN_STRLEN ||
         n > MAX_LEVENSHTEIN_STRLEN))
        ereport(ERROR,
                (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
                 errmsg("levenshtein argument exceeds maximum length of %d characters",
                        MAX_LEVENSHTEIN_STRLEN)));
 
#ifdef LEVENSHTEIN_LESS_EQUAL
    /* Initialize start and stop columns. */
    start_column = 0;
    stop_column = m + 1;
 
    /*
     * If max_d >= 0, determine whether the bound is impossibly tight.  If so,
     * return max_d + 1 immediately.  Otherwise, determine whether it's tight
     * enough to limit the computation we must perform.  If so, figure out
     * initial stop column.
     */
    if (max_d >= 0)
    {
        int         min_theo_d; /* Theoretical minimum distance. */
        int         max_theo_d; /* Theoretical maximum distance. */
        int         net_inserts = n - m;
 
        min_theo_d = net_inserts < 0 ?
            -net_inserts * del_c : net_inserts * ins_c;
        if (min_theo_d > max_d)
            return max_d + 1;
        if (ins_c + del_c < sub_c)
            sub_c = ins_c + del_c;
        max_theo_d = min_theo_d + sub_c * Min(m, n);
        if (max_d >= max_theo_d)
            max_d = -1;
        else if (ins_c + del_c > 0)
        {
            /*
             * Figure out how much of the first row of the notional matrix we
             * need to fill in.  If the string is growing, the theoretical
             * minimum distance already incorporates the cost of deleting the
             * number of characters necessary to make the two strings equal in
             * length.  Each additional deletion forces another insertion, so
             * the best-case total cost increases by ins_c + del_c. If the
             * string is shrinking, the minimum theoretical cost assumes no
             * excess deletions; that is, we're starting no further right than
             * column n - m.  If we do start further right, the best-case
             * total cost increases by ins_c + del_c for each move right.
             */
            int         slack_d = max_d - min_theo_d;
            int         best_column = net_inserts < 0 ? -net_inserts : 0;
 
            stop_column = best_column + (slack_d / (ins_c + del_c)) + 1;
            if (stop_column > m)
                stop_column = m + 1;
        }
    }
#endif
 
    /*
     * In order to avoid calling pg_mblen_range() repeatedly on each character
     * in s, we cache all the lengths before starting the main loop -- but if
     * all the characters in both strings are single byte, then we skip this
     * and use a fast-path in the main loop.  If only one string contains
     * multi-byte characters, we still build the array, so that the fast-path
     * needn't deal with the case where the array hasn't been initialized.
     */
    if (m != slen || n != tlen)
    {
        int         i;
        const char *cp = source;
 
        s_char_len = (int *) palloc((m + 1) * sizeof(int));
        for (i = 0; i < m; ++i)
        {
            s_char_len[i] = pg_mblen_range(cp, send);
            cp += s_char_len[i];
        }
        s_char_len[i] = 0;
    }
 
    /* One more cell for initialization column and row. */
    ++m;
    ++n;
 
    /* Previous and current rows of notional array. */
    prev = (int *) palloc(2 * m * sizeof(int));
    curr = prev + m;
 
    /*
     * To transform the first i characters of s into the first 0 characters of
     * t, we must perform i deletions.
     */
    for (int i = START_COLUMN; i < STOP_COLUMN; i++)
        prev[i] = i * del_c;
 
    /* Loop through rows of the notional array */
    for (y = target, j = 1; j < n; j++)
    {
        int        *temp;
        const char *x = source;
        int         y_char_len = n != tlen + 1 ? pg_mblen_range(y, tend) : 1;
        int         i;
 
#ifdef LEVENSHTEIN_LESS_EQUAL
 
        /*
         * In the best case, values percolate down the diagonal unchanged, so
         * we must increment stop_column unless it's already on the right end
         * of the array.  The inner loop will read prev[stop_column], so we
         * have to initialize it even though it shouldn't affect the result.
         */
        if (stop_column < m)
        {
            prev[stop_column] = max_d + 1;
            ++stop_column;
        }
 
        /*
         * The main loop fills in curr, but curr[0] needs a special case: to
         * transform the first 0 characters of s into the first j characters
         * of t, we must perform j insertions.  However, if start_column > 0,
         * this special case does not apply.
         */
        if (start_column == 0)
        {
            curr[0] = j * ins_c;
            i = 1;
        }
        else
            i = start_column;
#else
        curr[0] = j * ins_c;
        i = 1;
#endif
 
        /*
         * This inner loop is critical to performance, so we include a
         * fast-path to handle the (fairly common) case where no multibyte
         * characters are in the mix.  The fast-path is entitled to assume
         * that if s_char_len is not initialized then BOTH strings contain
         * only single-byte characters.
         */
        if (s_char_len != NULL)
        {
            for (; i < STOP_COLUMN; i++)
            {
                int         ins;
                int         del;
                int         sub;
                int         x_char_len = s_char_len[i - 1];
 
                /*
                 * Calculate costs for insertion, deletion, and substitution.
                 *
                 * When calculating cost for substitution, we compare the last
                 * character of each possibly-multibyte character first,
                 * because that's enough to rule out most mis-matches.  If we
                 * get past that test, then we compare the lengths and the
                 * remaining bytes.
                 */
                ins = prev[i] + ins_c;
                del = curr[i - 1] + del_c;
                if (x[x_char_len - 1] == y[y_char_len - 1]
                    && x_char_len == y_char_len &&
                    (x_char_len == 1 || rest_of_char_same(x, y, x_char_len)))
                    sub = prev[i - 1];
                else
                    sub = prev[i - 1] + sub_c;
 
                /* Take the one with minimum cost. */
                curr[i] = Min(ins, del);
                curr[i] = Min(curr[i], sub);
 
                /* Point to next character. */
                x += x_char_len;
            }
        }
        else
        {
            for (; i < STOP_COLUMN; i++)
            {
                int         ins;
                int         del;
                int         sub;
 
                /* Calculate costs for insertion, deletion, and substitution. */
                ins = prev[i] + ins_c;
                del = curr[i - 1] + del_c;
                sub = prev[i - 1] + ((*x == *y) ? 0 : sub_c);
 
                /* Take the one with minimum cost. */
                curr[i] = Min(ins, del);
                curr[i] = Min(curr[i], sub);
 
                /* Point to next character. */
                x++;
            }
        }
 
        /* Swap current row with previous row. */
        temp = curr;
        curr = prev;
        prev = temp;
 
        /* Point to next character. */
        y += y_char_len;
 
#ifdef LEVENSHTEIN_LESS_EQUAL
 
        /*
         * This chunk of code represents a significant performance hit if used
         * in the case where there is no max_d bound.  This is probably not
         * because the max_d >= 0 test itself is expensive, but rather because
         * the possibility of needing to execute this code prevents tight
         * optimization of the loop as a whole.
         */
        if (max_d >= 0)
        {
            /*
             * The "zero point" is the column of the current row where the
             * remaining portions of the strings are of equal length.  There
             * are (n - 1) characters in the target string, of which j have
             * been transformed.  There are (m - 1) characters in the source
             * string, so we want to find the value for zp where (n - 1) - j =
             * (m - 1) - zp.
             */
            int         zp = j - (n - m);
 
            /* Check whether the stop column can slide left. */
            while (stop_column > 0)
            {
                int         ii = stop_column - 1;
                int         net_inserts = ii - zp;
 
                if (prev[ii] + (net_inserts > 0 ? net_inserts * ins_c :
                                -net_inserts * del_c) <= max_d)
                    break;
                stop_column--;
            }
 
            /* Check whether the start column can slide right. */
            while (start_column < stop_column)
            {
                int         net_inserts = start_column - zp;
 
                if (prev[start_column] +
                    (net_inserts > 0 ? net_inserts * ins_c :
                     -net_inserts * del_c) <= max_d)
                    break;
 
                /*
                 * We'll never again update these values, so we must make sure
                 * there's nothing here that could confuse any future
                 * iteration of the outer loop.
                 */
                prev[start_column] = max_d + 1;
                curr[start_column] = max_d + 1;
                if (start_column != 0)
                    source += (s_char_len != NULL) ? s_char_len[start_column - 1] : 1;
                start_column++;
            }
 
            /* If they cross, we're going to exceed the bound. */
            if (start_column >= stop_column)
                return max_d + 1;
        }
#endif
    }
 
    /*
     * Because the final value was swapped from the previous row to the
     * current row, that's where we'll find it.
     */
    return prev[m - 1];
}