hash.c

/**********************************************************************

  hash.c -

  $Author$
  created at: Mon Nov 22 18:51:18 JST 1993

  Copyright (C) 1993-2007 Yukihiro Matsumoto
  Copyright (C) 2000  Network Applied Communication Laboratory, Inc.
  Copyright (C) 2000  Information-technology Promotion Agency, Japan

**********************************************************************/

#include "ruby/encoding.h"
#include "ruby/st.h"
#include "ruby/util.h"
#include "internal.h"
#include <errno.h>
#include "probes.h"
#include "id.h"
#include "symbol.h"
#include "debug_counter.h"
#include "transient_heap.h"
#include "ruby_assert.h"
#ifdef __APPLE__
# ifdef HAVE_CRT_EXTERNS_H
#  include <crt_externs.h>
# else
#  include "missing/crt_externs.h"
# endif
#endif

#ifndef HASH_DEBUG
#define HASH_DEBUG 0
#endif

#if HASH_DEBUG
#include "gc.h"
#endif

#define HAS_EXTRA_STATES(hash, klass) ( \
    ((klass = has_extra_methods(rb_obj_class(hash))) != 0) || \
    FL_TEST((hash), FL_EXIVAR|RHASH_PROC_DEFAULT) || \
    !NIL_P(RHASH_IFNONE(hash)))

#define SET_DEFAULT(hash, ifnone) ( \
    FL_UNSET_RAW(hash, RHASH_PROC_DEFAULT), \
    RHASH_SET_IFNONE(hash, ifnone))

#define SET_PROC_DEFAULT(hash, proc) set_proc_default(hash, proc)

#define COPY_DEFAULT(hash, hash2) copy_default(RHASH(hash), RHASH(hash2))

static inline void
copy_default(struct RHash *hash, const struct RHash *hash2)
{
    hash->basic.flags &= ~RHASH_PROC_DEFAULT;
    hash->basic.flags |= hash2->basic.flags & RHASH_PROC_DEFAULT;
    RHASH_SET_IFNONE(hash, RHASH_IFNONE(hash2));
}

static VALUE
has_extra_methods(VALUE klass)
{
    const VALUE base = rb_cHash;
    VALUE c = klass;
    while (c != base) {
	if (rb_class_has_methods(c)) return klass;
	c = RCLASS_SUPER(c);
    }
    return 0;
}

static VALUE rb_hash_s_try_convert(VALUE, VALUE);

/*
 * Hash WB strategy:
 *  1. Check mutate st_* functions
 *     * st_insert()
 *     * st_insert2()
 *     * st_update()
 *     * st_add_direct()
 *  2. Insert WBs
 */

VALUE
rb_hash_freeze(VALUE hash)
{
    return rb_obj_freeze(hash);
}

VALUE rb_cHash;

static VALUE envtbl;
static ID id_hash, id_yield, id_default, id_flatten_bang;
static ID id_hash_iter_lev;

VALUE
rb_hash_set_ifnone(VALUE hash, VALUE ifnone)
{
    RB_OBJ_WRITE(hash, (&RHASH(hash)->ifnone), ifnone);
    return hash;
}

static int
rb_any_cmp(VALUE a, VALUE b)
{
    if (a == b) return 0;
    if (RB_TYPE_P(a, T_STRING) && RBASIC(a)->klass == rb_cString &&
	RB_TYPE_P(b, T_STRING) && RBASIC(b)->klass == rb_cString) {
	return rb_str_hash_cmp(a, b);
    }
    if (a == Qundef || b == Qundef) return -1;
    if (SYMBOL_P(a) && SYMBOL_P(b)) {
	return a != b;
    }

    return !rb_eql(a, b);
}

static VALUE
hash_recursive(VALUE obj, VALUE arg, int recurse)
{
    if (recurse) return INT2FIX(0);
    return rb_funcallv(obj, id_hash, 0, 0);
}

VALUE
rb_hash(VALUE obj)
{
    VALUE hval = rb_exec_recursive_outer(hash_recursive, obj, 0);

    while (!FIXNUM_P(hval)) {
        if (RB_TYPE_P(hval, T_BIGNUM)) {
            int sign;
            unsigned long ul;
            sign = rb_integer_pack(hval, &ul, 1, sizeof(ul), 0,
                    INTEGER_PACK_NATIVE_BYTE_ORDER);
            ul &= (1UL << (sizeof(long)*CHAR_BIT-1)) - 1;
            if (sign < 0)
                return LONG2FIX(-(long)ul);
            return LONG2FIX((long)ul);
        }
	hval = rb_to_int(hval);
    }
    return hval;
}

static long rb_objid_hash(st_index_t index);

static st_index_t
dbl_to_index(double d)
{
    union {double d; st_index_t i;} u;
    u.d = d;
    return u.i;
}

long
rb_dbl_long_hash(double d)
{
    /* normalize -0.0 to 0.0 */
    if (d == 0.0) d = 0.0;
#if SIZEOF_INT == SIZEOF_VOIDP
    return rb_memhash(&d, sizeof(d));
#else
    return rb_objid_hash(dbl_to_index(d));
#endif
}

static inline long
any_hash(VALUE a, st_index_t (*other_func)(VALUE))
{
    VALUE hval;
    st_index_t hnum;

    if (SPECIAL_CONST_P(a)) {
	if (STATIC_SYM_P(a)) {
            hnum = a >> (RUBY_SPECIAL_SHIFT + ID_SCOPE_SHIFT);
            hnum = rb_hash_start(hnum);
            goto out;
        }
	else if (FLONUM_P(a)) {
	    /* prevent pathological behavior: [Bug #10761] */
	    goto flt;
	}
	hnum = rb_objid_hash((st_index_t)a);
    }
    else if (BUILTIN_TYPE(a) == T_STRING) {
	hnum = rb_str_hash(a);
    }
    else if (BUILTIN_TYPE(a) == T_SYMBOL) {
	hnum = RSYMBOL(a)->hashval;
    }
    else if (BUILTIN_TYPE(a) == T_BIGNUM) {
	hval = rb_big_hash(a);
	hnum = FIX2LONG(hval);
    }
    else if (BUILTIN_TYPE(a) == T_FLOAT) {
      flt:
	hnum = rb_dbl_long_hash(rb_float_value(a));
    }
    else {
	hnum = other_func(a);
    }
  out:
#if SIZEOF_LONG < SIZEOF_ST_INDEX_T
    if (hnum > 0)
	hnum &= (unsigned long)-1 >> 2;
    else
	hnum |= ~((unsigned long)-1 >> 2);
#else
    hnum <<= 1;
    hnum = RSHIFT(hnum, 1);
#endif
    return (long)hnum;
}

static st_index_t
obj_any_hash(VALUE obj)
{
    obj = rb_hash(obj);
    return FIX2LONG(obj);
}

static st_index_t
rb_any_hash(VALUE a)
{
    return any_hash(a, obj_any_hash);
}

/* Here is a hash function for 64-bit key.  It is about 5 times faster
   (2 times faster when uint128 type is absent) on Haswell than
   tailored Spooky or City hash function can be.  */

/* Here we two primes with random bit generation.  */
static const uint64_t prime1 = ((uint64_t)0x2e0bb864 << 32) | 0xe9ea7df5;
static const uint32_t prime2 = 0x830fcab9;


static inline uint64_t
mult_and_mix(uint64_t m1, uint64_t m2)
{
#if defined HAVE_UINT128_T
    uint128_t r = (uint128_t) m1 * (uint128_t) m2;
    return (uint64_t) (r >> 64) ^ (uint64_t) r;
#else
    uint64_t hm1 = m1 >> 32, hm2 = m2 >> 32;
    uint64_t lm1 = m1, lm2 = m2;
    uint64_t v64_128 = hm1 * hm2;
    uint64_t v32_96 = hm1 * lm2 + lm1 * hm2;
    uint64_t v1_32 = lm1 * lm2;

    return (v64_128 + (v32_96 >> 32)) ^ ((v32_96 << 32) + v1_32);
#endif
}

static inline uint64_t
key64_hash(uint64_t key, uint32_t seed)
{
    return mult_and_mix(key + seed, prime1);
}

/* Should cast down the result for each purpose */
#define st_index_hash(index) key64_hash(rb_hash_start(index), prime2)

static long
rb_objid_hash(st_index_t index)
{
    return (long)st_index_hash(index);
}

static st_index_t
objid_hash(VALUE obj)
{
    VALUE object_id = rb_obj_id(obj);
    if (!FIXNUM_P(object_id))
        object_id = rb_big_hash(object_id);

#if SIZEOF_LONG == SIZEOF_VOIDP
    return (st_index_t)st_index_hash((st_index_t)NUM2LONG(object_id));
#elif SIZEOF_LONG_LONG == SIZEOF_VOIDP
    return (st_index_t)st_index_hash((st_index_t)NUM2LL(object_id));
#endif
}

/**
 * call-seq:
 *    obj.hash    -> integer
 *
 * Generates an Integer hash value for this object.  This function must have the
 * property that <code>a.eql?(b)</code> implies <code>a.hash == b.hash</code>.
 *
 * The hash value is used along with #eql? by the Hash class to determine if
 * two objects reference the same hash key.  Any hash value that exceeds the
 * capacity of an Integer will be truncated before being used.
 *
 * The hash value for an object may not be identical across invocations or
 * implementations of Ruby.  If you need a stable identifier across Ruby
 * invocations and implementations you will need to generate one with a custom
 * method.
 *--
 * \private
 *++
 */
VALUE
rb_obj_hash(VALUE obj)
{
    long hnum = any_hash(obj, objid_hash);
    return ST2FIX(hnum);
}

static const struct st_hash_type objhash = {
    rb_any_cmp,
    rb_any_hash,
};

#define rb_ident_cmp st_numcmp

static st_index_t
rb_ident_hash(st_data_t n)
{
#ifdef USE_FLONUM /* RUBY */
    /*
     * - flonum (on 64-bit) is pathologically bad, mix the actual
     *   float value in, but do not use the float value as-is since
     *   many integers get interpreted as 2.0 or -2.0 [Bug #10761]
     */
    if (FLONUM_P(n)) {
        n ^= dbl_to_index(rb_float_value(n));
    }
#endif

    return (st_index_t)st_index_hash((st_index_t)n);
}

#define identhash rb_hashtype_ident
const struct st_hash_type rb_hashtype_ident = {
    rb_ident_cmp,
    rb_ident_hash,
};

typedef st_index_t st_hash_t;

/*
 * RHASH_AR_TABLE_P(h):
 * * as.ar == NULL or
 *   as.ar points ar_table.
 * * as.ar is allocated by transient heap or xmalloc.
 *
 * !RHASH_AR_TABLE_P(h):
 * * as.st points st_table.
 */

#define RHASH_AR_TABLE_MAX_BOUND     RHASH_AR_TABLE_MAX_SIZE

#define RHASH_AR_TABLE_REF(hash, n) (&RHASH_AR_TABLE(hash)->pairs[n])
#define RHASH_AR_CLEARED_HINT 0xff

typedef struct ar_table_pair_struct {
    VALUE key;
    VALUE val;
} ar_table_pair;

typedef struct ar_table_struct {
    /* 64bit CPU: 8B * 2 * 8 = 128B */
    ar_table_pair pairs[RHASH_AR_TABLE_MAX_SIZE];
} ar_table;

size_t
rb_hash_ar_table_size(void)
{
    return sizeof(ar_table);
}

static inline st_hash_t
ar_do_hash(st_data_t key)
{
    return (st_hash_t)rb_any_hash(key);
}

static inline ar_hint_t
ar_do_hash_hint(st_hash_t hash_value)
{
    return (ar_hint_t)hash_value;
}

static inline ar_hint_t
ar_hint(VALUE hash, unsigned int index)
{
    return RHASH(hash)->ar_hint.ary[index];
}

static inline void
ar_hint_set_hint(VALUE hash, unsigned int index, ar_hint_t hint)
{
    RHASH(hash)->ar_hint.ary[index] = hint;
}

static inline void
ar_hint_set(VALUE hash, unsigned int index, st_hash_t hash_value)
{
    ar_hint_set_hint(hash, index, ar_do_hash_hint(hash_value));
}

static inline void
ar_clear_entry(VALUE hash, unsigned int index)
{
    ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, index);
    pair->key = Qundef;
    ar_hint_set_hint(hash, index, RHASH_AR_CLEARED_HINT);
}

static inline int
ar_cleared_entry(VALUE hash, unsigned int index)
{
    if (ar_hint(hash, index) == RHASH_AR_CLEARED_HINT) {
        /* RHASH_AR_CLEARED_HINT is only a hint, not mean cleared entry,
         * so you need to check key == Qundef
         */
        ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, index);
        return pair->key == Qundef;
    }
    else {
        return FALSE;
    }
}

static inline void
ar_set_entry(VALUE hash, unsigned int index, st_data_t key, st_data_t val, st_hash_t hash_value)
{
    ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, index);
    pair->key = key;
    pair->val = val;
    ar_hint_set(hash, index, hash_value);
}

#define RHASH_AR_TABLE_SIZE(h) (HASH_ASSERT(RHASH_AR_TABLE_P(h)), \
                                RHASH_AR_TABLE_SIZE_RAW(h))

#define RHASH_AR_TABLE_BOUND_RAW(h) \
  ((unsigned int)((RBASIC(h)->flags >> RHASH_AR_TABLE_BOUND_SHIFT) & \
                  (RHASH_AR_TABLE_BOUND_MASK >> RHASH_AR_TABLE_BOUND_SHIFT)))

#define RHASH_AR_TABLE_BOUND(h) (HASH_ASSERT(RHASH_AR_TABLE_P(h)), \
                                 RHASH_AR_TABLE_BOUND_RAW(h))

#define RHASH_ST_TABLE_SET(h, s)  rb_hash_st_table_set(h, s)
#define RHASH_TYPE(hash) (RHASH_AR_TABLE_P(hash) ? &objhash : RHASH_ST_TABLE(hash)->type)

#define HASH_ASSERT(expr) RUBY_ASSERT_MESG_WHEN(HASH_DEBUG, expr, #expr)

#if HASH_DEBUG
#define hash_verify(hash) hash_verify_(hash, __FILE__, __LINE__)

void
rb_hash_dump(VALUE hash)
{
    rb_obj_info_dump(hash);

    if (RHASH_AR_TABLE_P(hash)) {
        unsigned i, n = 0, bound = RHASH_AR_TABLE_BOUND(hash);

        fprintf(stderr, "  size:%u bound:%u\n",
                RHASH_AR_TABLE_SIZE(hash), RHASH_AR_TABLE_BOUND(hash));

        for (i=0; i<bound; i++) {
            st_data_t k, v;

            if (!ar_cleared_entry(hash, i)) {
                char b1[0x100], b2[0x100];
                ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
                k = pair->key;
                v = pair->val;
                fprintf(stderr, "  %d key:%s val:%s hint:%02x\n", i,
                        rb_raw_obj_info(b1, 0x100, k),
                        rb_raw_obj_info(b2, 0x100, v),
                        ar_hint(hash, i));
                n++;
            }
            else {
                fprintf(stderr, "  %d empty\n", i);
            }
        }
    }
}

static VALUE
hash_verify_(VALUE hash, const char *file, int line)
{
    HASH_ASSERT(RB_TYPE_P(hash, T_HASH));

    if (RHASH_AR_TABLE_P(hash)) {
        unsigned i, n = 0, bound = RHASH_AR_TABLE_BOUND(hash);

        for (i=0; i<bound; i++) {
            st_data_t k, v;
            if (!ar_cleared_entry(hash, i)) {
                ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
                k = pair->key;
                v = pair->val;
                HASH_ASSERT(k != Qundef);
                HASH_ASSERT(v != Qundef);
                n++;
            }
        }
        if (n != RHASH_AR_TABLE_SIZE(hash)) {
            rb_bug("n:%u, RHASH_AR_TABLE_SIZE:%u", n, RHASH_AR_TABLE_SIZE(hash));
        }
    }
    else {
        HASH_ASSERT(RHASH_ST_TABLE(hash) != NULL);
        HASH_ASSERT(RHASH_AR_TABLE_SIZE_RAW(hash) == 0);
        HASH_ASSERT(RHASH_AR_TABLE_BOUND_RAW(hash) == 0);
    }

    if (RHASH_TRANSIENT_P(hash)) {
        volatile st_data_t MAYBE_UNUSED(key) = RHASH_AR_TABLE_REF(hash, 0)->key; /* read */
        HASH_ASSERT(RHASH_AR_TABLE(hash) != NULL);
        HASH_ASSERT(rb_transient_heap_managed_ptr_p(RHASH_AR_TABLE(hash)));
    }
    return hash;
}

#else
#define hash_verify(h) ((void)0)
#endif

static inline int
RHASH_TABLE_NULL_P(VALUE hash)
{
    if (RHASH(hash)->as.ar == NULL) {
        HASH_ASSERT(RHASH_AR_TABLE_P(hash));
        return TRUE;
    }
    else {
        return FALSE;
    }
}

static inline int
RHASH_TABLE_EMPTY_P(VALUE hash)
{
    return RHASH_SIZE(hash) == 0;
}

int
rb_hash_ar_table_p(VALUE hash)
{
    if (FL_TEST_RAW((hash), RHASH_ST_TABLE_FLAG)) {
        HASH_ASSERT(RHASH(hash)->as.st != NULL);
        return FALSE;
    }
    else {
        return TRUE;
    }
}

ar_table *
rb_hash_ar_table(VALUE hash)
{
    HASH_ASSERT(RHASH_AR_TABLE_P(hash));
    return RHASH(hash)->as.ar;
}

st_table *
rb_hash_st_table(VALUE hash)
{
    HASH_ASSERT(!RHASH_AR_TABLE_P(hash));
    return RHASH(hash)->as.st;
}

void
rb_hash_st_table_set(VALUE hash, st_table *st)
{
    HASH_ASSERT(st != NULL);
    FL_SET_RAW((hash), RHASH_ST_TABLE_FLAG);
    RHASH(hash)->as.st = st;
}

static void
hash_ar_table_set(VALUE hash, ar_table *ar)
{
    HASH_ASSERT(RHASH_AR_TABLE_P(hash));
    HASH_ASSERT((RHASH_TRANSIENT_P(hash) && ar == NULL) ? FALSE : TRUE);
    RHASH(hash)->as.ar = ar;
    hash_verify(hash);
}

#define RHASH_SET_ST_FLAG(h)          FL_SET_RAW(h, RHASH_ST_TABLE_FLAG)
#define RHASH_UNSET_ST_FLAG(h)        FL_UNSET_RAW(h, RHASH_ST_TABLE_FLAG)

static inline void
RHASH_AR_TABLE_BOUND_SET(VALUE h, st_index_t n)
{
    HASH_ASSERT(RHASH_AR_TABLE_P(h));
    HASH_ASSERT(n <= RHASH_AR_TABLE_MAX_BOUND);

    RBASIC(h)->flags &= ~RHASH_AR_TABLE_BOUND_MASK;
    RBASIC(h)->flags |= n << RHASH_AR_TABLE_BOUND_SHIFT;
}

static inline void
RHASH_AR_TABLE_SIZE_SET(VALUE h, st_index_t n)
{
    HASH_ASSERT(RHASH_AR_TABLE_P(h));
    HASH_ASSERT(n <= RHASH_AR_TABLE_MAX_SIZE);

    RBASIC(h)->flags &= ~RHASH_AR_TABLE_SIZE_MASK;
    RBASIC(h)->flags |= n << RHASH_AR_TABLE_SIZE_SHIFT;
}

static inline void
HASH_AR_TABLE_SIZE_ADD(VALUE h, st_index_t n)
{
    HASH_ASSERT(RHASH_AR_TABLE_P(h));

    RHASH_AR_TABLE_SIZE_SET(h, RHASH_AR_TABLE_SIZE(h) + n);

    hash_verify(h);
}

#define RHASH_AR_TABLE_SIZE_INC(h) HASH_AR_TABLE_SIZE_ADD(h, 1)

static inline void
RHASH_AR_TABLE_SIZE_DEC(VALUE h)
{
    HASH_ASSERT(RHASH_AR_TABLE_P(h));
    int new_size = RHASH_AR_TABLE_SIZE(h) - 1;

    if (new_size != 0) {
        RHASH_AR_TABLE_SIZE_SET(h, new_size);
    }
    else {
        RHASH_AR_TABLE_SIZE_SET(h, 0);
        RHASH_AR_TABLE_BOUND_SET(h, 0);
    }
    hash_verify(h);
}

static inline void
RHASH_AR_TABLE_CLEAR(VALUE h)
{
    RBASIC(h)->flags &= ~RHASH_AR_TABLE_SIZE_MASK;
    RBASIC(h)->flags &= ~RHASH_AR_TABLE_BOUND_MASK;

    hash_ar_table_set(h, NULL);
}

static ar_table*
ar_alloc_table(VALUE hash)
{
    ar_table *tab = (ar_table*)rb_transient_heap_alloc(hash, sizeof(ar_table));

    if (tab != NULL) {
        RHASH_SET_TRANSIENT_FLAG(hash);
    }
    else {
        RHASH_UNSET_TRANSIENT_FLAG(hash);
        tab = (ar_table*)ruby_xmalloc(sizeof(ar_table));
    }

    RHASH_AR_TABLE_SIZE_SET(hash, 0);
    RHASH_AR_TABLE_BOUND_SET(hash, 0);
    hash_ar_table_set(hash, tab);

    return tab;
}

NOINLINE(static int ar_equal(VALUE x, VALUE y));

static int
ar_equal(VALUE x, VALUE y)
{
    return rb_any_cmp(x, y) == 0;
}

static unsigned
ar_find_entry_hint(VALUE hash, ar_hint_t hint, st_data_t key)
{
    unsigned i, bound = RHASH_AR_TABLE_BOUND(hash);
    const ar_hint_t *hints = RHASH(hash)->ar_hint.ary;

    /* if table is NULL, then bound also should be 0 */

    for (i = 0; i < bound; i++) {
        if (hints[i] == hint) {
            ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
            if (ar_equal(key, pair->key)) {
                RB_DEBUG_COUNTER_INC(artable_hint_hit);
                return i;
            }
            else {
#if 0
                static int pid;
                static char fname[256];
                static FILE *fp;

                if (pid != getpid()) {
                    snprintf(fname, sizeof(fname), "/tmp/ruby-armiss.%d", pid = getpid());
                    if ((fp = fopen(fname, "w")) == NULL) rb_bug("fopen");
                }

                st_hash_t h1 = ar_do_hash(key);
                st_hash_t h2 = ar_do_hash(pair->key);

                fprintf(fp, "miss: hash_eq:%d hints[%d]:%02x hint:%02x\n"
                            "      key      :%016lx %s\n"
                            "      pair->key:%016lx %s\n",
                        h1 == h2, i, hints[i], hint,
                        h1, rb_obj_info(key), h2, rb_obj_info(pair->key));
#endif
                RB_DEBUG_COUNTER_INC(artable_hint_miss);
            }
        }
    }
    RB_DEBUG_COUNTER_INC(artable_hint_notfound);
    return RHASH_AR_TABLE_MAX_BOUND;
}

static unsigned
ar_find_entry(VALUE hash, st_hash_t hash_value, st_data_t key)
{
    ar_hint_t hint = ar_do_hash_hint(hash_value);
    return ar_find_entry_hint(hash, hint, key);
}

static inline void
ar_free_and_clear_table(VALUE hash)
{
    ar_table *tab = RHASH_AR_TABLE(hash);

    if (tab) {
        if (RHASH_TRANSIENT_P(hash)) {
            RHASH_UNSET_TRANSIENT_FLAG(hash);
        }
        else {
            ruby_xfree(RHASH_AR_TABLE(hash));
        }
        RHASH_AR_TABLE_CLEAR(hash);
    }
    HASH_ASSERT(RHASH_AR_TABLE_SIZE(hash) == 0);
    HASH_ASSERT(RHASH_AR_TABLE_BOUND(hash) == 0);
    HASH_ASSERT(RHASH_TRANSIENT_P(hash) == 0);
}

static void
ar_try_convert_table(VALUE hash)
{
    st_table *new_tab;
    const unsigned size = RHASH_AR_TABLE_SIZE(hash);
    st_index_t i;

    if (!RHASH_AR_TABLE_P(hash) || size < RHASH_AR_TABLE_MAX_SIZE) {
        return;
    }

    new_tab = st_init_table_with_size(&objhash, size * 2);

    for (i = 0; i < RHASH_AR_TABLE_MAX_BOUND; i++) {
        ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
        st_add_direct(new_tab, pair->key, pair->val);
    }
    ar_free_and_clear_table(hash);
    RHASH_ST_TABLE_SET(hash, new_tab);
    return;
}

static st_table *
ar_force_convert_table(VALUE hash, const char *file, int line)
{
    st_table *new_tab;

    if (RHASH_ST_TABLE_P(hash)) {
        return RHASH_ST_TABLE(hash);
    }

    if (RHASH_AR_TABLE(hash)) {
        unsigned i, bound = RHASH_AR_TABLE_BOUND(hash);

#if RHASH_CONVERT_TABLE_DEBUG
        rb_obj_info_dump(hash);
        fprintf(stderr, "force_convert: %s:%d\n", file, line);
        RB_DEBUG_COUNTER_INC(obj_hash_force_convert);
#endif

        new_tab = st_init_table_with_size(&objhash, RHASH_AR_TABLE_SIZE(hash));

        for (i = 0; i < bound; i++) {
            if (ar_cleared_entry(hash, i)) continue;

            ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
            st_add_direct(new_tab, pair->key, pair->val);
        }
        ar_free_and_clear_table(hash);
    }
    else {
        new_tab = st_init_table(&objhash);
    }
    RHASH_ST_TABLE_SET(hash, new_tab);

    return new_tab;
}

static ar_table *
hash_ar_table(VALUE hash)
{
    if (RHASH_TABLE_NULL_P(hash)) {
        ar_alloc_table(hash);
    }
    return RHASH_AR_TABLE(hash);
}

static int
ar_compact_table(VALUE hash)
{
    const unsigned bound = RHASH_AR_TABLE_BOUND(hash);
    const unsigned size = RHASH_AR_TABLE_SIZE(hash);

    if (size == bound) {
        return size;
    }
    else {
        unsigned i, j=0;
        ar_table_pair *pairs = RHASH_AR_TABLE(hash)->pairs;

        for (i=0; i<bound; i++) {
            if (ar_cleared_entry(hash, i)) {
                if (j <= i) j = i+1;
                for (; j<bound; j++) {
                    if (!ar_cleared_entry(hash, j)) {
                        pairs[i] = pairs[j];
                        ar_hint_set_hint(hash, i, (st_hash_t)ar_hint(hash, j));
                        ar_clear_entry(hash, j);
                        j++;
                        goto found;
                    }
                }
                /* non-empty is not found */
                goto done;
              found:;
            }
        }
      done:
        HASH_ASSERT(i<=bound);

        RHASH_AR_TABLE_BOUND_SET(hash, size);
        hash_verify(hash);
        return size;
    }
}

static int
ar_add_direct_with_hash(VALUE hash, st_data_t key, st_data_t val, st_hash_t hash_value)
{
    unsigned bin = RHASH_AR_TABLE_BOUND(hash);

    if (RHASH_AR_TABLE_SIZE(hash) >= RHASH_AR_TABLE_MAX_SIZE) {
        return 1;
    }
    else {
        if (UNLIKELY(bin >= RHASH_AR_TABLE_MAX_BOUND)) {
            bin = ar_compact_table(hash);
            hash_ar_table(hash);
        }
        HASH_ASSERT(bin < RHASH_AR_TABLE_MAX_BOUND);

        ar_set_entry(hash, bin, key, val, hash_value);
        RHASH_AR_TABLE_BOUND_SET(hash, bin+1);
        RHASH_AR_TABLE_SIZE_INC(hash);
        return 0;
    }
}

static int
ar_general_foreach(VALUE hash, st_foreach_check_callback_func *func, st_update_callback_func *replace, st_data_t arg)
{
    if (RHASH_AR_TABLE_SIZE(hash) > 0) {
        unsigned i, bound = RHASH_AR_TABLE_BOUND(hash);

        for (i = 0; i < bound; i++) {
            if (ar_cleared_entry(hash, i)) continue;

            ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
            enum st_retval retval = (*func)(pair->key, pair->val, arg, 0);
            /* pair may be not valid here because of theap */

            switch (retval) {
              case ST_CONTINUE:
                break;
              case ST_CHECK:
              case ST_STOP:
                return 0;
              case ST_REPLACE:
                if (replace) {
                    VALUE key = pair->key;
                    VALUE val = pair->val;
                    retval = (*replace)(&key, &val, arg, TRUE);

                    // TODO: pair should be same as pair before.
                    ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
                    pair->key = key;
                    pair->val = val;
                }
                break;
              case ST_DELETE:
                ar_clear_entry(hash, i);
                RHASH_AR_TABLE_SIZE_DEC(hash);
                break;
            }
        }
    }
    return 0;
}

static int
ar_foreach_with_replace(VALUE hash, st_foreach_check_callback_func *func, st_update_callback_func *replace, st_data_t arg)
{
    return ar_general_foreach(hash, func, replace, arg);
}

struct functor {
    st_foreach_callback_func *func;
    st_data_t arg;
};

static int
apply_functor(st_data_t k, st_data_t v, st_data_t d, int _)
{
    const struct functor *f = (void *)d;
    return f->func(k, v, f->arg);
}

static int
ar_foreach(VALUE hash, st_foreach_callback_func *func, st_data_t arg)
{
    const struct functor f = { func, arg };
    return ar_general_foreach(hash, apply_functor, NULL, (st_data_t)&f);
}

static int
ar_foreach_check(VALUE hash, st_foreach_check_callback_func *func, st_data_t arg,
                     st_data_t never)
{
    if (RHASH_AR_TABLE_SIZE(hash) > 0) {
        unsigned i, ret = 0, bound = RHASH_AR_TABLE_BOUND(hash);
        enum st_retval retval;
        st_data_t key;
        ar_table_pair *pair;
        ar_hint_t hint;

        for (i = 0; i < bound; i++) {
            if (ar_cleared_entry(hash, i)) continue;

            pair = RHASH_AR_TABLE_REF(hash, i);
            key = pair->key;
            hint = ar_hint(hash, i);

            retval = (*func)(key, pair->val, arg, 0);
            hash_verify(hash);

            switch (retval) {
              case ST_CHECK: {
                  if (pair->key == never) break;
                  ret = ar_find_entry_hint(hash, hint, key);
                  if (ret == RHASH_AR_TABLE_MAX_BOUND) {
                      retval = (*func)(0, 0, arg, 1);
                      return 2;
                  }
              }
              case ST_CONTINUE:
                break;
              case ST_STOP:
              case ST_REPLACE:
                return 0;
              case ST_DELETE: {
                  if (!ar_cleared_entry(hash, i)) {
                      ar_clear_entry(hash, i);
                      RHASH_AR_TABLE_SIZE_DEC(hash);
                  }
                  break;
              }
            }
        }
    }
    return 0;
}

static int
ar_update(VALUE hash, st_data_t key,
              st_update_callback_func *func, st_data_t arg)
{
    int retval, existing;
    unsigned bin = RHASH_AR_TABLE_MAX_BOUND;
    st_data_t value = 0, old_key;
    st_hash_t hash_value = ar_do_hash(key);

    if (RHASH_AR_TABLE_SIZE(hash) > 0) {
        bin = ar_find_entry(hash, hash_value, key);
        existing = (bin != RHASH_AR_TABLE_MAX_BOUND) ? TRUE : FALSE;
    }
    else {
        hash_ar_table(hash); /* allocate ltbl if needed */
        existing = FALSE;
    }

    if (existing) {
        ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, bin);
        key = pair->key;
        value = pair->val;
    }
    old_key = key;
    retval = (*func)(&key, &value, arg, existing);
    /* pair can be invalid here because of theap */

    switch (retval) {
      case ST_CONTINUE:
        if (!existing) {
            if (ar_add_direct_with_hash(hash, key, value, hash_value)) {
                return -1;
            }
        }
        else {
            ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, bin);
            if (old_key != key) {
                pair->key = key;
            }
            pair->val = value;
        }
        break;
      case ST_DELETE:
        if (existing) {
            ar_clear_entry(hash, bin);
            RHASH_AR_TABLE_SIZE_DEC(hash);
        }
        break;
    }
    return existing;
}

static int
ar_insert(VALUE hash, st_data_t key, st_data_t value)
{
    unsigned bin = RHASH_AR_TABLE_BOUND(hash);
    st_hash_t hash_value = ar_do_hash(key);

    hash_ar_table(hash); /* prepare ltbl */

    bin = ar_find_entry(hash, hash_value, key);
    if (bin == RHASH_AR_TABLE_MAX_BOUND) {
        if (RHASH_AR_TABLE_SIZE(hash) >= RHASH_AR_TABLE_MAX_SIZE) {
            return -1;
        }
        else if (bin >= RHASH_AR_TABLE_MAX_BOUND) {
            bin = ar_compact_table(hash);
            hash_ar_table(hash);
        }
        HASH_ASSERT(bin < RHASH_AR_TABLE_MAX_BOUND);

        ar_set_entry(hash, bin, key, value, hash_value);
        RHASH_AR_TABLE_BOUND_SET(hash, bin+1);
        RHASH_AR_TABLE_SIZE_INC(hash);
        return 0;
    }
    else {
        RHASH_AR_TABLE_REF(hash, bin)->val = value;
        return 1;
    }
}

static int
ar_lookup(VALUE hash, st_data_t key, st_data_t *value)
{
    if (RHASH_AR_TABLE_SIZE(hash) == 0) {
        return 0;
    }
    else {
        st_hash_t hash_value = ar_do_hash(key);
        unsigned bin = ar_find_entry(hash, hash_value, key);

        if (bin == RHASH_AR_TABLE_MAX_BOUND) {
            return 0;
        }
        else {
            HASH_ASSERT(bin < RHASH_AR_TABLE_MAX_BOUND);
            if (value != NULL) {
                *value = RHASH_AR_TABLE_REF(hash, bin)->val;
            }
            return 1;
        }
    }
}

static int
ar_delete(VALUE hash, st_data_t *key, st_data_t *value)
{
    unsigned bin;
    st_hash_t hash_value = ar_do_hash(*key);


    bin = ar_find_entry(hash, hash_value, *key);

    if (bin == RHASH_AR_TABLE_MAX_BOUND) {
        if (value != 0) *value = 0;
        return 0;
    }
    else {
        if (value != 0) {
            ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, bin);
            *value = pair->val;
        }
        ar_clear_entry(hash, bin);
        RHASH_AR_TABLE_SIZE_DEC(hash);
        return 1;
    }
}

static int
ar_shift(VALUE hash, st_data_t *key, st_data_t *value)
{
    if (RHASH_AR_TABLE_SIZE(hash) > 0) {
        unsigned i, bound = RHASH_AR_TABLE_BOUND(hash);

        for (i = 0; i < bound; i++) {
            if (!ar_cleared_entry(hash, i)) {
                ar_table_pair *pair = RHASH_AR_TABLE_REF(hash, i);
                if (value != 0) *value = pair->val;
                *key = pair->key;
                ar_clear_entry(hash, i);
                RHASH_AR_TABLE_SIZE_DEC(hash);
                return 1;
            }
        }
    }
    if (value != NULL) *value = 0;
    return 0;
}

static long
ar_keys(VALUE hash, st_data_t *keys, st_index_t size)
{
    unsigned i, bound = RHASH_AR_TABLE_BOUND(hash);
    st_data_t *keys_start = keys, *keys_end = keys + size;

    for (i = 0; i < bound; i++) {
        if (keys == keys_end) {
          break;
        }
        else {
            if (!ar_cleared_entry(hash, i)) {
                *keys++ = RHASH_AR_TABLE_REF(hash, i)->key;
            }
        }
    }

    return keys - keys_start;
}

static long
ar_values(VALUE hash, st_data_t *values, st_index_t size)
{
    unsigned i, bound = RHASH_AR_TABLE_BOUND(hash);
    st_data_t *values_start = values, *values_end = values + size;

    for (i = 0; i < bound; i++) {
        if (values == values_end) {
          break;
        }
        else {
            if (!ar_cleared_entry(hash, i)) {
                *values++ = RHASH_AR_TABLE_REF(hash, i)->val;
            }
        }
    }

    return values - values_start;
}

static ar_table*
ar_copy(VALUE hash1, VALUE hash2)
{
    ar_table *old_tab = RHASH_AR_TABLE(hash2);

    if (old_tab != NULL) {
        ar_table *new_tab = RHASH_AR_TABLE(hash1);
        if (new_tab == NULL) {
            new_tab = (ar_table*) rb_transient_heap_alloc(hash1, sizeof(ar_table));
            if (new_tab != NULL) {
                RHASH_SET_TRANSIENT_FLAG(hash1);
            }
            else {
                RHASH_UNSET_TRANSIENT_FLAG(hash1);
                new_tab = (ar_table*)ruby_xmalloc(sizeof(ar_table));
            }
        }
        *new_tab = *old_tab;
        RHASH(hash1)->ar_hint.word = RHASH(hash2)->ar_hint.word;
        RHASH_AR_TABLE_BOUND_SET(hash1, RHASH_AR_TABLE_BOUND(hash2));
        RHASH_AR_TABLE_SIZE_SET(hash1, RHASH_AR_TABLE_SIZE(hash2));
        hash_ar_table_set(hash1, new_tab);

        rb_gc_writebarrier_remember(hash1);
        return new_tab;
    }
    else {
        RHASH_AR_TABLE_BOUND_SET(hash1, RHASH_AR_TABLE_BOUND(hash2));
        RHASH_AR_TABLE_SIZE_SET(hash1, RHASH_AR_TABLE_SIZE(hash2));

        if (RHASH_TRANSIENT_P(hash1)) {
            RHASH_UNSET_TRANSIENT_FLAG(hash1);
        }
        else if (RHASH_AR_TABLE(hash1)) {
            ruby_xfree(RHASH_AR_TABLE(hash1));
        }

        hash_ar_table_set(hash1, NULL);

        rb_gc_writebarrier_remember(hash1);
        return old_tab;
    }
}

static void
ar_clear(VALUE hash)
{
    if (RHASH_AR_TABLE(hash) != NULL) {
        RHASH_AR_TABLE_SIZE_SET(hash, 0);
        RHASH_AR_TABLE_BOUND_SET(hash, 0);
    }
    else {
        HASH_ASSERT(RHASH_AR_TABLE_SIZE(hash) == 0);
        HASH_ASSERT(RHASH_AR_TABLE_BOUND(hash) == 0);
    }
}

#if USE_TRANSIENT_HEAP
void
rb_hash_transient_heap_evacuate(VALUE hash, int promote)
{
    if (RHASH_TRANSIENT_P(hash)) {
        ar_table *new_tab;
        ar_table *old_tab = RHASH_AR_TABLE(hash);

        if (UNLIKELY(old_tab == NULL)) {
            rb_gc_force_recycle(hash);
            return;
        }
        HASH_ASSERT(old_tab != NULL);
        if (promote) {
          promote:
            new_tab = ruby_xmalloc(sizeof(ar_table));
            RHASH_UNSET_TRANSIENT_FLAG(hash);
        }
        else {
            new_tab = rb_transient_heap_alloc(hash, sizeof(ar_table));
            if (new_tab == NULL) goto promote;
        }
        *new_tab = *old_tab;
        hash_ar_table_set(hash, new_tab);
    }
    hash_verify(hash);
}
#endif

typedef int st_foreach_func(st_data_t, st_data_t, st_data_t);

struct foreach_safe_arg {
    st_table *tbl;
    st_foreach_func *func;
    st_data_t arg;
};

static int
foreach_safe_i(st_data_t key, st_data_t value, st_data_t args, int error)
{
    int status;
    struct foreach_safe_arg *arg = (void *)args;

    if (error) return ST_STOP;
    status = (*arg->func)(key, value, arg->arg);
    if (status == ST_CONTINUE) {
	return ST_CHECK;
    }
    return status;
}

void
st_foreach_safe(st_table *table, st_foreach_func *func, st_data_t a)
{
    struct foreach_safe_arg arg;

    arg.tbl = table;
    arg.func = (st_foreach_func *)func;
    arg.arg = a;
    if (st_foreach_check(table, foreach_safe_i, (st_data_t)&arg, 0)) {
	rb_raise(rb_eRuntimeError, "hash modified during iteration");
    }
}

typedef int rb_foreach_func(VALUE, VALUE, VALUE);

struct hash_foreach_arg {
    VALUE hash;
    rb_foreach_func *func;
    VALUE arg;
};

static int
hash_ar_foreach_iter(st_data_t key, st_data_t value, st_data_t argp, int error)
{
    struct hash_foreach_arg *arg = (struct hash_foreach_arg *)argp;
    int status;

    if (error) return ST_STOP;
    status = (*arg->func)((VALUE)key, (VALUE)value, arg->arg);
    /* TODO: rehash check? rb_raise(rb_eRuntimeError, "rehash occurred during iteration"); */

    switch (status) {
      case ST_DELETE:
        return ST_DELETE;
      case ST_CONTINUE:
        break;
      case ST_STOP:
        return ST_STOP;
    }
    return ST_CHECK;
}

static int
hash_foreach_iter(st_data_t key, st_data_t value, st_data_t argp, int error)
{
    struct hash_foreach_arg *arg = (struct hash_foreach_arg *)argp;
    int status;
    st_table *tbl;

    if (error) return ST_STOP;
    tbl = RHASH_ST_TABLE(arg->hash);
    status = (*arg->func)((VALUE)key, (VALUE)value, arg->arg);
    if (RHASH_ST_TABLE(arg->hash) != tbl) {
    	rb_raise(rb_eRuntimeError, "rehash occurred during iteration");
    }
    switch (status) {
      case ST_DELETE:
	return ST_DELETE;
      case ST_CONTINUE:
	break;
      case ST_STOP:
	return ST_STOP;
    }
    return ST_CHECK;
}

static int
iter_lev_in_ivar(VALUE hash)
{
    VALUE levval = rb_ivar_get(hash, id_hash_iter_lev);
    HASH_ASSERT(FIXNUM_P(levval));
    return FIX2INT(levval);
}

void rb_ivar_set_internal(VALUE obj, ID id, VALUE val);

static void
iter_lev_in_ivar_set(VALUE hash, int lev)
{
    rb_ivar_set_internal(hash, id_hash_iter_lev, INT2FIX(lev));
}

static int
iter_lev_in_flags(VALUE hash)
{
    unsigned int u = (unsigned int)((RBASIC(hash)->flags >> RHASH_LEV_SHIFT) & RHASH_LEV_MAX);
    return (int)u;
}

static int
RHASH_ITER_LEV(VALUE hash)
{
    int lev = iter_lev_in_flags(hash);

    if (lev == RHASH_LEV_MAX) {
        return iter_lev_in_ivar(hash);
    }
    else {
        return lev;
    }
}

static void
hash_iter_lev_inc(VALUE hash)
{
    int lev = iter_lev_in_flags(hash);
    if (lev == RHASH_LEV_MAX) {
        lev = iter_lev_in_ivar(hash);
        iter_lev_in_ivar_set(hash, lev+1);
    }
    else {
        lev += 1;
        RBASIC(hash)->flags = ((RBASIC(hash)->flags & ~RHASH_LEV_MASK) | (lev << RHASH_LEV_SHIFT));
        if (lev == RHASH_LEV_MAX) {
            iter_lev_in_ivar_set(hash, lev);
        }
    }
}

static void
hash_iter_lev_dec(VALUE hash)
{
    int lev = iter_lev_in_flags(hash);
    if (lev == RHASH_LEV_MAX) {
        lev = iter_lev_in_ivar(hash);
        HASH_ASSERT(lev > 0);
        iter_lev_in_ivar_set(hash, lev-1);
    }
    else {
        HASH_ASSERT(lev > 0);
        RBASIC(hash)->flags = ((RBASIC(hash)->flags & ~RHASH_LEV_MASK) | ((lev-1) << RHASH_LEV_SHIFT));
    }
}

static VALUE
hash_foreach_ensure_rollback(VALUE hash)
{
    hash_iter_lev_inc(hash);
    return 0;
}

static VALUE
hash_foreach_ensure(VALUE hash)
{
    hash_iter_lev_dec(hash);
    return 0;
}

int
rb_hash_stlike_foreach(VALUE hash, st_foreach_callback_func *func, st_data_t arg)
{
    if (RHASH_AR_TABLE_P(hash)) {
        return ar_foreach(hash, func, arg);
    }
    else {
        return st_foreach(RHASH_ST_TABLE(hash), func, arg);
    }
}

int
rb_hash_stlike_foreach_with_replace(VALUE hash, st_foreach_check_callback_func *func, st_update_callback_func *replace, st_data_t arg)
{
    if (RHASH_AR_TABLE_P(hash)) {
        return ar_foreach_with_replace(hash, func, replace, arg);
    }
    else {
        return st_foreach_with_replace(RHASH_ST_TABLE(hash), func, replace, arg);
    }
}

static VALUE
hash_foreach_call(VALUE arg)
{
    VALUE hash = ((struct hash_foreach_arg *)arg)->hash;
    int ret = 0;
    if (RHASH_AR_TABLE_P(hash)) {
        ret = ar_foreach_check(hash, hash_ar_foreach_iter,
                                   (st_data_t)arg, (st_data_t)Qundef);
    }
    else if (RHASH_ST_TABLE_P(hash)) {
        ret = st_foreach_check(RHASH_ST_TABLE(hash), hash_foreach_iter,
                               (st_data_t)arg, (st_data_t)Qundef);
    }
    if (ret) {
        rb_raise(rb_eRuntimeError, "ret: %d, hash modified during iteration", ret);
    }
    return Qnil;
}

void
rb_hash_foreach(VALUE hash, rb_foreach_func *func, VALUE farg)
{
    struct hash_foreach_arg arg;

    if (RHASH_TABLE_EMPTY_P(hash))
        return;
    hash_iter_lev_inc(hash);
    arg.hash = hash;
    arg.func = (rb_foreach_func *)func;
    arg.arg  = farg;
    rb_ensure(hash_foreach_call, (VALUE)&arg, hash_foreach_ensure, hash);
    hash_verify(hash);
}

static VALUE
hash_alloc_flags(VALUE klass, VALUE flags, VALUE ifnone)
{
    const VALUE wb = (RGENGC_WB_PROTECTED_HASH ? FL_WB_PROTECTED : 0);
    NEWOBJ_OF(hash, struct RHash, klass, T_HASH | wb | flags);

    RHASH_SET_IFNONE((VALUE)hash, ifnone);

    return (VALUE)hash;
}

static VALUE
hash_alloc(VALUE klass)
{
    return hash_alloc_flags(klass, 0, Qnil);
}

static VALUE
empty_hash_alloc(VALUE klass)
{
    RUBY_DTRACE_CREATE_HOOK(HASH, 0);

    return hash_alloc(klass);
}

VALUE
rb_hash_new(void)
{
    return hash_alloc(rb_cHash);
}

MJIT_FUNC_EXPORTED VALUE
rb_hash_new_with_size(st_index_t size)
{
    VALUE ret = rb_hash_new();
    if (size == 0) {
        /* do nothing */
    }
    else if (size <= RHASH_AR_TABLE_MAX_SIZE) {
        ar_alloc_table(ret);
    }
    else {
        RHASH_ST_TABLE_SET(ret, st_init_table_with_size(&objhash, size));
    }
    return ret;
}

static VALUE
hash_dup(VALUE hash, VALUE klass, VALUE flags)
{
    VALUE ret = hash_alloc_flags(klass, flags,
				 RHASH_IFNONE(hash));
    if (!RHASH_EMPTY_P(hash)) {
        if (RHASH_AR_TABLE_P(hash))
            ar_copy(ret, hash);
        else if (RHASH_ST_TABLE_P(hash))
            RHASH_ST_TABLE_SET(ret, st_copy(RHASH_ST_TABLE(hash)));
    }
    return ret;
}

VALUE
rb_hash_dup(VALUE hash)
{
    const VALUE flags = RBASIC(hash)->flags;
    VALUE ret = hash_dup(hash, rb_obj_class(hash),
                         flags & (FL_EXIVAR|RHASH_PROC_DEFAULT));
    if (flags & FL_EXIVAR)
        rb_copy_generic_ivar(ret, hash);
    return ret;
}

MJIT_FUNC_EXPORTED VALUE
rb_hash_resurrect(VALUE hash)
{
    VALUE ret = hash_dup(hash, rb_cHash, 0);
    return ret;
}

static void
rb_hash_modify_check(VALUE hash)
{
    rb_check_frozen(hash);
}

MJIT_FUNC_EXPORTED struct st_table *
#if RHASH_CONVERT_TABLE_DEBUG
rb_hash_tbl_raw(VALUE hash, const char *file, int line)
{
    return ar_force_convert_table(hash, file, line);
}
#else
rb_hash_tbl_raw(VALUE hash)
{
    return ar_force_convert_table(hash, NULL, 0);
}
#endif

struct st_table *
rb_hash_tbl(VALUE hash, const char *file, int line)
{
    OBJ_WB_UNPROTECT(hash);
    return RHASH_TBL_RAW(hash);
}

static void
rb_hash_modify(VALUE hash)
{
    rb_hash_modify_check(hash);
}

NORETURN(static void no_new_key(void));
static void
no_new_key(void)
{
    rb_raise(rb_eRuntimeError, "can't add a new key into hash during iteration");
}

struct update_callback_arg {
    VALUE hash;
    st_data_t arg;
};

#define NOINSERT_UPDATE_CALLBACK(func)                                       \
static int                                                                   \
func##_noinsert(st_data_t *key, st_data_t *val, st_data_t arg, int existing) \
{                                                                            \
    if (!existing) no_new_key();                                             \
    return func(key, val, (struct update_arg *)arg, existing);               \
}                                                                            \
                                                                             \
static int                                                                   \
func##_insert(st_data_t *key, st_data_t *val, st_data_t arg, int existing)   \
{                                                                            \
    return func(key, val, (struct update_arg *)arg, existing);               \
}

struct update_arg {
    st_data_t arg;
    VALUE hash;
    VALUE new_key;
    VALUE old_key;
    VALUE new_value;
    VALUE old_value;
};

typedef int (*tbl_update_func)(st_data_t *, st_data_t *, st_data_t, int);

int
rb_hash_stlike_update(VALUE hash, st_data_t key, st_update_callback_func func, st_data_t arg)
{
    if (RHASH_AR_TABLE_P(hash)) {
        int result = ar_update(hash, (st_data_t)key, func, arg);
        if (result == -1) {
            ar_try_convert_table(hash);
        }
        else {
            return result;
        }
    }

    return st_update(RHASH_ST_TABLE(hash), (st_data_t)key, func, arg);
}

static int
tbl_update(VALUE hash, VALUE key, tbl_update_func func, st_data_t optional_arg)
{
    struct update_arg arg;
    int result;

    arg.arg = optional_arg;
    arg.hash = hash;
    arg.new_key = 0;
    arg.old_key = Qundef;
    arg.new_value = 0;
    arg.old_value = Qundef;

    result = rb_hash_stlike_update(hash, key, func, (st_data_t)&arg);

    /* write barrier */
    if (arg.new_key)   RB_OBJ_WRITTEN(hash, arg.old_key, arg.new_key);
    if (arg.new_value) RB_OBJ_WRITTEN(hash, arg.old_value, arg.new_value);

    return result;
}

#define UPDATE_CALLBACK(iter_lev, func) ((iter_lev) > 0 ? func##_noinsert : func##_insert)

#define RHASH_UPDATE_ITER(h, iter_lev, key, func, a) do {                        \
    tbl_update((h), (key), UPDATE_CALLBACK((iter_lev), func), (st_data_t)(a)); \
} while (0)

#define RHASH_UPDATE(hash, key, func, arg) \
    RHASH_UPDATE_ITER(hash, RHASH_ITER_LEV(hash), key, func, arg)

static void
set_proc_default(VALUE hash, VALUE proc)
{
    if (rb_proc_lambda_p(proc)) {
	int n = rb_proc_arity(proc);

	if (n != 2 && (n >= 0 || n < -3)) {
	    if (n < 0) n = -n-1;
	    rb_raise(rb_eTypeError, "default_proc takes two arguments (2 for %d)", n);
	}
    }

    FL_SET_RAW(hash, RHASH_PROC_DEFAULT);
    RHASH_SET_IFNONE(hash, proc);
}

/*
 *  call-seq:
 *     Hash.new                          -> new_hash
 *     Hash.new(obj)                     -> new_hash
 *     Hash.new {|hash, key| block }     -> new_hash
 *
 *  Returns a new, empty hash. If this hash is subsequently accessed by
 *  a key that doesn't correspond to a hash entry, the value returned
 *  depends on the style of <code>new</code> used to create the hash. In
 *  the first form, the access returns <code>nil</code>. If
 *  <i>obj</i> is specified, this single object will be used for
 *  all <em>default values</em>. If a block is specified, it will be
 *  called with the hash object and the key, and should return the
 *  default value. It is the block's responsibility to store the value
 *  in the hash if required.
 *
 *     h = Hash.new("Go Fish")
 *     h["a"] = 100
 *     h["b"] = 200
 *     h["a"]           #=> 100
 *     h["c"]           #=> "Go Fish"
 *     # The following alters the single default object
 *     h["c"].upcase!   #=> "GO FISH"
 *     h["d"]           #=> "GO FISH"
 *     h.keys           #=> ["a", "b"]
 *
 *     # While this creates a new default object each time
 *     h = Hash.new { |hash, key| hash[key] = "Go Fish: #{key}" }
 *     h["c"]           #=> "Go Fish: c"
 *     h["c"].upcase!   #=> "GO FISH: C"
 *     h["d"]           #=> "Go Fish: d"
 *     h.keys           #=> ["c", "d"]
 *
 */

static VALUE
rb_hash_initialize(int argc, VALUE *argv, VALUE hash)
{
    VALUE ifnone;

    rb_hash_modify(hash);
    if (rb_block_given_p()) {
	rb_check_arity(argc, 0, 0);
	ifnone = rb_block_proc();
	SET_PROC_DEFAULT(hash, ifnone);
    }
    else {
	rb_check_arity(argc, 0, 1);
	ifnone = argc == 0 ? Qnil : argv[0];
	RHASH_SET_IFNONE(hash, ifnone);
    }

    return hash;
}

/*
 *  call-seq:
 *     Hash[ key, value, ... ]         -> new_hash
 *     Hash[ [ [key, value], ... ] ]   -> new_hash
 *     Hash[ object ]                  -> new_hash
 *
 *  Creates a new hash populated with the given objects.
 *
 *  Similar to the literal <code>{ _key_ => _value_, ... }</code>. In the first
 *  form, keys and values occur in pairs, so there must be an even number of
 *  arguments.
 *
 *  The second and third form take a single argument which is either an array
 *  of key-value pairs or an object convertible to a hash.
 *
 *     Hash["a", 100, "b", 200]             #=> {"a"=>100, "b"=>200}
 *     Hash[ [ ["a", 100], ["b", 200] ] ]   #=> {"a"=>100, "b"=>200}
 *     Hash["a" => 100, "b" => 200]         #=> {"a"=>100, "b"=>200}
 */

static VALUE
rb_hash_s_create(int argc, VALUE *argv, VALUE klass)
{
    VALUE hash, tmp;

    if (argc == 1) {
        tmp = rb_hash_s_try_convert(Qnil, argv[0]);
	if (!NIL_P(tmp)) {
	    hash = hash_alloc(klass);
            if (RHASH_AR_TABLE_P(tmp)) {
                ar_copy(hash, tmp);
	    }
            else {
                RHASH_ST_TABLE_SET(hash, st_copy(RHASH_ST_TABLE(tmp)));
            }
	    return hash;
	}

	tmp = rb_check_array_type(argv[0]);
	if (!NIL_P(tmp)) {
	    long i;

	    hash = hash_alloc(klass);
	    for (i = 0; i < RARRAY_LEN(tmp); ++i) {
		VALUE e = RARRAY_AREF(tmp, i);
		VALUE v = rb_check_array_type(e);
		VALUE key, val = Qnil;

		if (NIL_P(v)) {
		    rb_raise(rb_eArgError, "wrong element type %s at %ld (expected array)",
			     rb_builtin_class_name(e), i);
		}
		switch (RARRAY_LEN(v)) {
		  default:
		    rb_raise(rb_eArgError, "invalid number of elements (%ld for 1..2)",
			     RARRAY_LEN(v));
		  case 2:
		    val = RARRAY_AREF(v, 1);
		  case 1:
		    key = RARRAY_AREF(v, 0);
		    rb_hash_aset(hash, key, val);
		}
	    }
	    return hash;
	}
    }
    if (argc % 2 != 0) {
	rb_raise(rb_eArgError, "odd number of arguments for Hash");
    }

    hash = hash_alloc(klass);
    rb_hash_bulk_insert(argc, argv, hash);
    hash_verify(hash);
    return hash;
}

VALUE
rb_to_hash_type(VALUE hash)
{
    return rb_convert_type_with_id(hash, T_HASH, "Hash", idTo_hash);
}
#define to_hash rb_to_hash_type

VALUE
rb_check_hash_type(VALUE hash)
{
    return rb_check_convert_type_with_id(hash, T_HASH, "Hash", idTo_hash);
}

/*
 *  call-seq:
 *     Hash.try_convert(obj) -> hash or nil
 *
 *  Try to convert <i>obj</i> into a hash, using to_hash method.
 *  Returns converted hash or nil if <i>obj</i> cannot be converted
 *  for any reason.
 *
 *     Hash.try_convert({1=>2})   # => {1=>2}
 *     Hash.try_convert("1=>2")   # => nil
 */
static VALUE
rb_hash_s_try_convert(VALUE dummy, VALUE hash)
{
    return rb_check_hash_type(hash);
}

struct rehash_arg {
    VALUE hash;
    st_table *tbl;
};

static int
rb_hash_rehash_i(VALUE key, VALUE value, VALUE arg)
{
    if (RHASH_AR_TABLE_P(arg)) {
        ar_insert(arg, (st_data_t)key, (st_data_t)value);
    }
    else {
        st_insert(RHASH_ST_TABLE(arg), (st_data_t)key, (st_data_t)value);
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.rehash -> hsh
 *
 *  Rebuilds the hash based on the current hash values for each key. If
 *  values of key objects have changed since they were inserted, this
 *  method will reindex <i>hsh</i>. If Hash#rehash is
 *  called while an iterator is traversing the hash, a
 *  RuntimeError will be raised in the iterator.
 *
 *     a = [ "a", "b" ]
 *     c = [ "c", "d" ]
 *     h = { a => 100, c => 300 }
 *     h[a]       #=> 100
 *     a[0] = "z"
 *     h[a]       #=> nil
 *     h.rehash   #=> {["z", "b"]=>100, ["c", "d"]=>300}
 *     h[a]       #=> 100
 */

VALUE
rb_hash_rehash(VALUE hash)
{
    VALUE tmp;
    st_table *tbl;

    if (RHASH_ITER_LEV(hash) > 0) {
	rb_raise(rb_eRuntimeError, "rehash during iteration");
    }
    rb_hash_modify_check(hash);
    if (RHASH_AR_TABLE_P(hash)) {
        tmp = hash_alloc(0);
        ar_alloc_table(tmp);
        rb_hash_foreach(hash, rb_hash_rehash_i, (VALUE)tmp);
        ar_free_and_clear_table(hash);
        ar_copy(hash, tmp);
        ar_free_and_clear_table(tmp);
    }
    else if (RHASH_ST_TABLE_P(hash)) {
        st_table *old_tab = RHASH_ST_TABLE(hash);
        tmp = hash_alloc(0);
        tbl = st_init_table_with_size(old_tab->type, old_tab->num_entries);
        RHASH_ST_TABLE_SET(tmp, tbl);
        rb_hash_foreach(hash, rb_hash_rehash_i, (VALUE)tmp);
        st_free_table(old_tab);
        RHASH_ST_TABLE_SET(hash, tbl);
        RHASH_ST_CLEAR(tmp);
    }
    hash_verify(hash);
    return hash;
}

VALUE
rb_hash_default_value(VALUE hash, VALUE key)
{
    if (rb_method_basic_definition_p(CLASS_OF(hash), id_default)) {
	VALUE ifnone = RHASH_IFNONE(hash);
        if (!FL_TEST(hash, RHASH_PROC_DEFAULT)) return ifnone;
	if (key == Qundef) return Qnil;
	return rb_funcall(ifnone, id_yield, 2, hash, key);
    }
    else {
	return rb_funcall(hash, id_default, 1, key);
    }
}

static inline int
hash_stlike_lookup(VALUE hash, st_data_t key, st_data_t *pval)
{
    hash_verify(hash);

    if (RHASH_AR_TABLE_P(hash)) {
        return ar_lookup(hash, key, pval);
    }
    else {
        return st_lookup(RHASH_ST_TABLE(hash), key, pval);
    }
}

MJIT_FUNC_EXPORTED int
rb_hash_stlike_lookup(VALUE hash, st_data_t key, st_data_t *pval)
{
    return hash_stlike_lookup(hash, key, pval);
}

/*
 *  call-seq:
 *     hsh[key]    ->  value
 *
 *  Element Reference---Retrieves the <i>value</i> object corresponding
 *  to the <i>key</i> object. If not found, returns the default value (see
 *  Hash::new for details).
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h["a"]   #=> 100
 *     h["c"]   #=> nil
 *
 */

VALUE
rb_hash_aref(VALUE hash, VALUE key)
{
    st_data_t val;

    if (hash_stlike_lookup(hash, key, &val)) {
        return (VALUE)val;
    }
    else {
        return rb_hash_default_value(hash, key);
    }
}

VALUE
rb_hash_lookup2(VALUE hash, VALUE key, VALUE def)
{
    st_data_t val;

    if (hash_stlike_lookup(hash, key, &val)) {
        return (VALUE)val;
    }
    else {
        return def; /* without Hash#default */
    }
}

VALUE
rb_hash_lookup(VALUE hash, VALUE key)
{
    return rb_hash_lookup2(hash, key, Qnil);
}

/*
 *  call-seq:
 *     hsh.fetch(key [, default] )       -> obj
 *     hsh.fetch(key) {| key | block }   -> obj
 *
 *  Returns a value from the hash for the given key. If the key can't be
 *  found, there are several options: With no other arguments, it will
 *  raise a KeyError exception; if <i>default</i> is given,
 *  then that will be returned; if the optional code block is specified,
 *  then that will be run and its result returned.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.fetch("a")                            #=> 100
 *     h.fetch("z", "go fish")                 #=> "go fish"
 *     h.fetch("z") { |el| "go fish, #{el}"}   #=> "go fish, z"
 *
 *  The following example shows that an exception is raised if the key
 *  is not found and a default value is not supplied.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.fetch("z")
 *
 *  <em>produces:</em>
 *
 *     prog.rb:2:in `fetch': key not found (KeyError)
 *      from prog.rb:2
 *
 */

static VALUE
rb_hash_fetch_m(int argc, VALUE *argv, VALUE hash)
{
    VALUE key;
    st_data_t val;
    long block_given;

    rb_check_arity(argc, 1, 2);
    key = argv[0];

    block_given = rb_block_given_p();
    if (block_given && argc == 2) {
	rb_warn("block supersedes default value argument");
    }

    if (hash_stlike_lookup(hash, key, &val)) {
        return (VALUE)val;
    }
    else {
        if (block_given) {
            return rb_yield(key);
        }
        else if (argc == 1) {
            VALUE desc = rb_protect(rb_inspect, key, 0);
            if (NIL_P(desc)) {
                desc = rb_any_to_s(key);
            }
            desc = rb_str_ellipsize(desc, 65);
            rb_key_err_raise(rb_sprintf("key not found: %"PRIsVALUE, desc), hash, key);
        }
        else {
            return argv[1];
        }
    }
}

VALUE
rb_hash_fetch(VALUE hash, VALUE key)
{
    return rb_hash_fetch_m(1, &key, hash);
}

/*
 *  call-seq:
 *     hsh.default(key=nil)   -> obj
 *
 *  Returns the default value, the value that would be returned by
 *  <i>hsh</i>[<i>key</i>] if <i>key</i> did not exist in <i>hsh</i>.
 *  See also Hash::new and Hash#default=.
 *
 *     h = Hash.new                            #=> {}
 *     h.default                               #=> nil
 *     h.default(2)                            #=> nil
 *
 *     h = Hash.new("cat")                     #=> {}
 *     h.default                               #=> "cat"
 *     h.default(2)                            #=> "cat"
 *
 *     h = Hash.new {|h,k| h[k] = k.to_i*10}   #=> {}
 *     h.default                               #=> nil
 *     h.default(2)                            #=> 20
 */

static VALUE
rb_hash_default(int argc, VALUE *argv, VALUE hash)
{
    VALUE args[2], ifnone;

    rb_check_arity(argc, 0, 1);
    ifnone = RHASH_IFNONE(hash);
    if (FL_TEST(hash, RHASH_PROC_DEFAULT)) {
	if (argc == 0) return Qnil;
	args[0] = hash;
	args[1] = argv[0];
	return rb_funcallv(ifnone, id_yield, 2, args);
    }
    return ifnone;
}

/*
 *  call-seq:
 *     hsh.default = obj     -> obj
 *
 *  Sets the default value, the value returned for a key that does not
 *  exist in the hash. It is not possible to set the default to a
 *  Proc that will be executed on each key lookup.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.default = "Go fish"
 *     h["a"]     #=> 100
 *     h["z"]     #=> "Go fish"
 *     # This doesn't do what you might hope...
 *     h.default = proc do |hash, key|
 *       hash[key] = key + key
 *     end
 *     h[2]       #=> #<Proc:0x401b3948@-:6>
 *     h["cat"]   #=> #<Proc:0x401b3948@-:6>
 */

static VALUE
rb_hash_set_default(VALUE hash, VALUE ifnone)
{
    rb_hash_modify_check(hash);
    SET_DEFAULT(hash, ifnone);
    return ifnone;
}

/*
 *  call-seq:
 *     hsh.default_proc -> anObject
 *
 *  If Hash::new was invoked with a block, return that
 *  block, otherwise return <code>nil</code>.
 *
 *     h = Hash.new {|h,k| h[k] = k*k }   #=> {}
 *     p = h.default_proc                 #=> #<Proc:0x401b3d08@-:1>
 *     a = []                             #=> []
 *     p.call(a, 2)
 *     a                                  #=> [nil, nil, 4]
 */


static VALUE
rb_hash_default_proc(VALUE hash)
{
    if (FL_TEST(hash, RHASH_PROC_DEFAULT)) {
	return RHASH_IFNONE(hash);
    }
    return Qnil;
}

/*
 *  call-seq:
 *     hsh.default_proc = proc_obj or nil
 *
 *  Sets the default proc to be executed on each failed key lookup.
 *
 *     h.default_proc = proc do |hash, key|
 *       hash[key] = key + key
 *     end
 *     h[2]       #=> 4
 *     h["cat"]   #=> "catcat"
 */

VALUE
rb_hash_set_default_proc(VALUE hash, VALUE proc)
{
    VALUE b;

    rb_hash_modify_check(hash);
    if (NIL_P(proc)) {
	SET_DEFAULT(hash, proc);
	return proc;
    }
    b = rb_check_convert_type_with_id(proc, T_DATA, "Proc", idTo_proc);
    if (NIL_P(b) || !rb_obj_is_proc(b)) {
	rb_raise(rb_eTypeError,
		 "wrong default_proc type %s (expected Proc)",
		 rb_obj_classname(proc));
    }
    proc = b;
    SET_PROC_DEFAULT(hash, proc);
    return proc;
}

static int
key_i(VALUE key, VALUE value, VALUE arg)
{
    VALUE *args = (VALUE *)arg;

    if (rb_equal(value, args[0])) {
	args[1] = key;
	return ST_STOP;
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.key(value)    -> key
 *
 *  Returns the key of an occurrence of a given value. If the value is
 *  not found, returns <code>nil</code>.
 *
 *     h = { "a" => 100, "b" => 200, "c" => 300, "d" => 300 }
 *     h.key(200)   #=> "b"
 *     h.key(300)   #=> "c"
 *     h.key(999)   #=> nil
 *
 */

static VALUE
rb_hash_key(VALUE hash, VALUE value)
{
    VALUE args[2];

    args[0] = value;
    args[1] = Qnil;

    rb_hash_foreach(hash, key_i, (VALUE)args);

    return args[1];
}

/* :nodoc: */
static VALUE
rb_hash_index(VALUE hash, VALUE value)
{
    rb_warn("Hash#index is deprecated; use Hash#key");
    return rb_hash_key(hash, value);
}

int
rb_hash_stlike_delete(VALUE hash, st_data_t *pkey, st_data_t *pval)
{
    if (RHASH_AR_TABLE_P(hash)) {
        return ar_delete(hash, pkey, pval);
    }
    else {
        return st_delete(RHASH_ST_TABLE(hash), pkey, pval);
    }
}

/*
 * delete a specified entry a given key.
 * if there is the corresponding entry, return a value of the entry.
 * if there is no corresponding entry, return Qundef.
 */
VALUE
rb_hash_delete_entry(VALUE hash, VALUE key)
{
    st_data_t ktmp = (st_data_t)key, val;

    if (rb_hash_stlike_delete(hash, &ktmp, &val)) {
        return (VALUE)val;
    }
    else {
        return Qundef;
    }
}

/*
 * delete a specified entry by a given key.
 * if there is the corresponding entry, return a value of the entry.
 * if there is no corresponding entry, return Qnil.
 */
VALUE
rb_hash_delete(VALUE hash, VALUE key)
{
    VALUE deleted_value = rb_hash_delete_entry(hash, key);

    if (deleted_value != Qundef) { /* likely pass */
	return deleted_value;
    }
    else {
	return Qnil;
    }
}

/*
 *  call-seq:
 *     hsh.delete(key)                   -> value
 *     hsh.delete(key) {| key | block }  -> value
 *
 *  Deletes the key-value pair and returns the value from <i>hsh</i> whose
 *  key is equal to <i>key</i>. If the key is not found, it returns
 *  <em>nil</em>. If the optional code block is given and the
 *  key is not found, pass in the key and return the result of
 *  <i>block</i>.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.delete("a")                              #=> 100
 *     h.delete("z")                              #=> nil
 *     h.delete("z") { |el| "#{el} not found" }   #=> "z not found"
 *
 */

static VALUE
rb_hash_delete_m(VALUE hash, VALUE key)
{
    VALUE val;

    rb_hash_modify_check(hash);
    val = rb_hash_delete_entry(hash, key);

    if (val != Qundef) {
	return val;
    }
    else {
	if (rb_block_given_p()) {
	    return rb_yield(key);
	}
	else {
	    return Qnil;
	}
    }
}

struct shift_var {
    VALUE key;
    VALUE val;
};

static int
shift_i_safe(VALUE key, VALUE value, VALUE arg)
{
    struct shift_var *var = (struct shift_var *)arg;

    var->key = key;
    var->val = value;
    return ST_STOP;
}

/*
 *  call-seq:
 *     hsh.shift -> anArray or obj
 *
 *  Removes a key-value pair from <i>hsh</i> and returns it as the
 *  two-item array <code>[</code> <i>key, value</i> <code>]</code>, or
 *  the hash's default value if the hash is empty.
 *
 *     h = { 1 => "a", 2 => "b", 3 => "c" }
 *     h.shift   #=> [1, "a"]
 *     h         #=> {2=>"b", 3=>"c"}
 */

static VALUE
rb_hash_shift(VALUE hash)
{
    struct shift_var var;

    rb_hash_modify_check(hash);
    if (RHASH_AR_TABLE_P(hash)) {
	var.key = Qundef;
	if (RHASH_ITER_LEV(hash) == 0) {
            if (ar_shift(hash, &var.key, &var.val)) {
		return rb_assoc_new(var.key, var.val);
	    }
	}
	else {
            rb_hash_foreach(hash, shift_i_safe, (VALUE)&var);
            if (var.key != Qundef) {
                rb_hash_delete_entry(hash, var.key);
                return rb_assoc_new(var.key, var.val);
            }
        }
    }
    if (RHASH_ST_TABLE_P(hash)) {
        var.key = Qundef;
        if (RHASH_ITER_LEV(hash) == 0) {
            if (st_shift(RHASH_ST_TABLE(hash), &var.key, &var.val)) {
                return rb_assoc_new(var.key, var.val);
            }
        }
        else {
	    rb_hash_foreach(hash, shift_i_safe, (VALUE)&var);
	    if (var.key != Qundef) {
		rb_hash_delete_entry(hash, var.key);
		return rb_assoc_new(var.key, var.val);
	    }
	}
    }
    return rb_hash_default_value(hash, Qnil);
}

static int
delete_if_i(VALUE key, VALUE value, VALUE hash)
{
    if (RTEST(rb_yield_values(2, key, value))) {
	return ST_DELETE;
    }
    return ST_CONTINUE;
}

static VALUE
hash_enum_size(VALUE hash, VALUE args, VALUE eobj)
{
    return rb_hash_size(hash);
}

/*
 *  call-seq:
 *     hsh.delete_if {| key, value | block }  -> hsh
 *     hsh.delete_if                          -> an_enumerator
 *
 *  Deletes every key-value pair from <i>hsh</i> for which <i>block</i>
 *  evaluates to <code>true</code>.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *     h = { "a" => 100, "b" => 200, "c" => 300 }
 *     h.delete_if {|key, value| key >= "b" }   #=> {"a"=>100}
 *
 */

VALUE
rb_hash_delete_if(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_modify_check(hash);
    if (!RHASH_TABLE_EMPTY_P(hash)) {
        rb_hash_foreach(hash, delete_if_i, hash);
    }
    return hash;
}

/*
 *  call-seq:
 *     hsh.reject! {| key, value | block }  -> hsh or nil
 *     hsh.reject!                          -> an_enumerator
 *
 *  Equivalent to Hash#delete_if, but returns
 *  <code>nil</code> if no changes were made.
 */

VALUE
rb_hash_reject_bang(VALUE hash)
{
    st_index_t n;

    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_modify(hash);
    n = RHASH_SIZE(hash);
    if (!n) return Qnil;
    rb_hash_foreach(hash, delete_if_i, hash);
    if (n == RHASH_SIZE(hash)) return Qnil;
    return hash;
}

static int
reject_i(VALUE key, VALUE value, VALUE result)
{
    if (!RTEST(rb_yield_values(2, key, value))) {
	rb_hash_aset(result, key, value);
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.reject {|key, value| block}   -> a_hash
 *     hsh.reject                        -> an_enumerator
 *
 *  Returns a new hash consisting of entries for which the block returns false.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *     h = { "a" => 100, "b" => 200, "c" => 300 }
 *     h.reject {|k,v| k < "b"}  #=> {"b" => 200, "c" => 300}
 *     h.reject {|k,v| v > 100}  #=> {"a" => 100}
 */

VALUE
rb_hash_reject(VALUE hash)
{
    VALUE result;

    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    if (RTEST(ruby_verbose)) {
	VALUE klass;
	if (HAS_EXTRA_STATES(hash, klass)) {
	    rb_warn("extra states are no longer copied: %+"PRIsVALUE, hash);
	}
    }
    result = rb_hash_new();
    if (!RHASH_EMPTY_P(hash)) {
	rb_hash_foreach(hash, reject_i, result);
    }
    return result;
}

/*
 *  call-seq:
 *     hsh.slice(*keys) -> a_hash
 *
 *  Returns a hash containing only the given keys and their values.
 *
 *     h = { a: 100, b: 200, c: 300 }
 *     h.slice(:a)           #=> {:a=>100}
 *     h.slice(:b, :c, :d)   #=> {:b=>200, :c=>300}
 */

static VALUE
rb_hash_slice(int argc, VALUE *argv, VALUE hash)
{
    int i;
    VALUE key, value, result;

    if (argc == 0 || RHASH_EMPTY_P(hash)) {
	return rb_hash_new();
    }
    result = rb_hash_new_with_size(argc);

    for (i = 0; i < argc; i++) {
	key = argv[i];
	value = rb_hash_lookup2(hash, key, Qundef);
	if (value != Qundef)
	    rb_hash_aset(result, key, value);
    }

    return result;
}

/*
 * call-seq:
 *   hsh.values_at(key, ...)   -> array
 *
 * Return an array containing the values associated with the given keys.
 * Also see Hash.select.
 *
 *   h = { "cat" => "feline", "dog" => "canine", "cow" => "bovine" }
 *   h.values_at("cow", "cat")  #=> ["bovine", "feline"]
 */

VALUE
rb_hash_values_at(int argc, VALUE *argv, VALUE hash)
{
    VALUE result = rb_ary_new2(argc);
    long i;

    for (i=0; i<argc; i++) {
	rb_ary_push(result, rb_hash_aref(hash, argv[i]));
    }
    return result;
}

/*
 * call-seq:
 *   hsh.fetch_values(key, ...)                 -> array
 *   hsh.fetch_values(key, ...) { |key| block } -> array
 *
 * Returns an array containing the values associated with the given keys
 * but also raises KeyError when one of keys can't be found.
 * Also see Hash#values_at and Hash#fetch.
 *
 *   h = { "cat" => "feline", "dog" => "canine", "cow" => "bovine" }
 *
 *   h.fetch_values("cow", "cat")                   #=> ["bovine", "feline"]
 *   h.fetch_values("cow", "bird")                  # raises KeyError
 *   h.fetch_values("cow", "bird") { |k| k.upcase } #=> ["bovine", "BIRD"]
 */

VALUE
rb_hash_fetch_values(int argc, VALUE *argv, VALUE hash)
{
    VALUE result = rb_ary_new2(argc);
    long i;

    for (i=0; i<argc; i++) {
	rb_ary_push(result, rb_hash_fetch(hash, argv[i]));
    }
    return result;
}

static int
select_i(VALUE key, VALUE value, VALUE result)
{
    if (RTEST(rb_yield_values(2, key, value))) {
	rb_hash_aset(result, key, value);
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.select {|key, value| block}   -> a_hash
 *     hsh.select                        -> an_enumerator
 *     hsh.filter {|key, value| block}   -> a_hash
 *     hsh.filter                        -> an_enumerator
 *
 *  Returns a new hash consisting of entries for which the block returns true.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *     h = { "a" => 100, "b" => 200, "c" => 300 }
 *     h.select {|k,v| k > "a"}  #=> {"b" => 200, "c" => 300}
 *     h.select {|k,v| v < 200}  #=> {"a" => 100}
 *
 *  Hash#filter is an alias for Hash#select.
 */

VALUE
rb_hash_select(VALUE hash)
{
    VALUE result;

    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    result = rb_hash_new();
    if (!RHASH_EMPTY_P(hash)) {
	rb_hash_foreach(hash, select_i, result);
    }
    return result;
}

static int
keep_if_i(VALUE key, VALUE value, VALUE hash)
{
    if (!RTEST(rb_yield_values(2, key, value))) {
	return ST_DELETE;
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.select! {| key, value | block }  -> hsh or nil
 *     hsh.select!                          -> an_enumerator
 *     hsh.filter! {| key, value | block }  -> hsh or nil
 *     hsh.filter!                          -> an_enumerator
 *
 *  Equivalent to Hash#keep_if, but returns
 *  +nil+ if no changes were made.
 *
 *  Hash#filter! is an alias for Hash#select!.
 */

VALUE
rb_hash_select_bang(VALUE hash)
{
    st_index_t n;

    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_modify_check(hash);
    n = RHASH_SIZE(hash);
    if (!n) return Qnil;
    rb_hash_foreach(hash, keep_if_i, hash);
    if (n == RHASH_SIZE(hash)) return Qnil;
    return hash;
}

/*
 *  call-seq:
 *     hsh.keep_if {| key, value | block }  -> hsh
 *     hsh.keep_if                          -> an_enumerator
 *
 *  Deletes every key-value pair from <i>hsh</i> for which <i>block</i>
 *  evaluates to +false+.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *  See also Hash#select!.
 */

VALUE
rb_hash_keep_if(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_modify_check(hash);
    if (!RHASH_TABLE_EMPTY_P(hash)) {
        rb_hash_foreach(hash, keep_if_i, hash);
    }
    return hash;
}

static int
clear_i(VALUE key, VALUE value, VALUE dummy)
{
    return ST_DELETE;
}

/*
 *  call-seq:
 *     hsh.clear -> hsh
 *
 *  Removes all key-value pairs from <i>hsh</i>.
 *
 *     h = { "a" => 100, "b" => 200 }   #=> {"a"=>100, "b"=>200}
 *     h.clear                          #=> {}
 *
 */

VALUE
rb_hash_clear(VALUE hash)
{
    rb_hash_modify_check(hash);

    if (RHASH_ITER_LEV(hash) > 0) {
        rb_hash_foreach(hash, clear_i, 0);
    }
    else if (RHASH_AR_TABLE_P(hash)) {
        ar_clear(hash);
    }
    else {
        st_clear(RHASH_ST_TABLE(hash));
    }

    return hash;
}

static int
hash_aset(st_data_t *key, st_data_t *val, struct update_arg *arg, int existing)
{
    if (existing) {
	arg->new_value = arg->arg;
	arg->old_value = *val;
    }
    else {
	arg->new_key = *key;
	arg->new_value = arg->arg;
    }
    *val = arg->arg;
    return ST_CONTINUE;
}

VALUE
rb_hash_key_str(VALUE key)
{
    if (!RB_FL_ANY_RAW(key, FL_EXIVAR) && RBASIC_CLASS(key) == rb_cString) {
        return rb_fstring(key);
    }
    else {
	return rb_str_new_frozen(key);
    }
}

static int
hash_aset_str(st_data_t *key, st_data_t *val, struct update_arg *arg, int existing)
{
    if (!existing && !RB_OBJ_FROZEN(*key)) {
	*key = rb_hash_key_str(*key);
    }
    return hash_aset(key, val, arg, existing);
}

NOINSERT_UPDATE_CALLBACK(hash_aset)
NOINSERT_UPDATE_CALLBACK(hash_aset_str)

/*
 *  call-seq:
 *     hsh[key] = value        -> value
 *     hsh.store(key, value)   -> value
 *
 *  == Element Assignment
 *
 *  Associates the value given by +value+ with the key given by +key+.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h["a"] = 9
 *     h["c"] = 4
 *     h   #=> {"a"=>9, "b"=>200, "c"=>4}
 *     h.store("d", 42) #=> 42
 *     h   #=> {"a"=>9, "b"=>200, "c"=>4, "d"=>42}
 *
 *  +key+ should not have its value changed while it is in use as a key (an
 *  <tt>unfrozen String</tt> passed as a key will be duplicated and frozen).
 *
 *     a = "a"
 *     b = "b".freeze
 *     h = { a => 100, b => 200 }
 *     h.key(100).equal? a #=> false
 *     h.key(200).equal? b #=> true
 *
 */

VALUE
rb_hash_aset(VALUE hash, VALUE key, VALUE val)
{
    int iter_lev = RHASH_ITER_LEV(hash);

    rb_hash_modify(hash);

    if (RHASH_TABLE_NULL_P(hash)) {
	if (iter_lev > 0) no_new_key();
        ar_alloc_table(hash);
    }

    if (RHASH_TYPE(hash) == &identhash || rb_obj_class(key) != rb_cString) {
	RHASH_UPDATE_ITER(hash, iter_lev, key, hash_aset, val);
    }
    else {
	RHASH_UPDATE_ITER(hash, iter_lev, key, hash_aset_str, val);
    }
    return val;
}

/*
 *  call-seq:
 *     hsh.replace(other_hash) -> hsh
 *
 *  Replaces the contents of <i>hsh</i> with the contents of
 *  <i>other_hash</i>.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.replace({ "c" => 300, "d" => 400 })   #=> {"c"=>300, "d"=>400}
 *
 */

static VALUE
rb_hash_replace(VALUE hash, VALUE hash2)
{
    rb_hash_modify_check(hash);
    if (hash == hash2) return hash;
    if (RHASH_ITER_LEV(hash) > 0) {
        rb_raise(rb_eRuntimeError, "can't replace hash during iteration");
    }
    hash2 = to_hash(hash2);

    COPY_DEFAULT(hash, hash2);

    if (RHASH_AR_TABLE_P(hash)) {
        if (RHASH_AR_TABLE_P(hash2)) {
            ar_clear(hash);
        }
        else {
            ar_free_and_clear_table(hash);
            RHASH_ST_TABLE_SET(hash, st_init_table_with_size(RHASH_TYPE(hash2), RHASH_SIZE(hash2)));
        }
    }
    else {
        if (RHASH_AR_TABLE_P(hash2)) {
            st_free_table(RHASH_ST_TABLE(hash));
            RHASH_ST_CLEAR(hash);
        }
        else {
            st_clear(RHASH_ST_TABLE(hash));
            RHASH_TBL_RAW(hash)->type = RHASH_ST_TABLE(hash2)->type;
        }
    }
    rb_hash_foreach(hash2, rb_hash_rehash_i, (VALUE)hash);

    rb_gc_writebarrier_remember(hash);

    return hash;
}

/*
 *  call-seq:
 *     hsh.length    ->  integer
 *     hsh.size      ->  integer
 *
 *  Returns the number of key-value pairs in the hash.
 *
 *     h = { "d" => 100, "a" => 200, "v" => 300, "e" => 400 }
 *     h.size          #=> 4
 *     h.delete("a")   #=> 200
 *     h.size          #=> 3
 *     h.length        #=> 3
 *
 *  Hash#length is an alias for Hash#size.
 */

VALUE
rb_hash_size(VALUE hash)
{
    return INT2FIX(RHASH_SIZE(hash));
}

size_t
rb_hash_size_num(VALUE hash)
{
    return (long)RHASH_SIZE(hash);
}

/*
 *  call-seq:
 *     hsh.empty?    -> true or false
 *
 *  Returns <code>true</code> if <i>hsh</i> contains no key-value pairs.
 *
 *     {}.empty?   #=> true
 *
 */

static VALUE
rb_hash_empty_p(VALUE hash)
{
    return RHASH_EMPTY_P(hash) ? Qtrue : Qfalse;
}

static int
each_value_i(VALUE key, VALUE value, VALUE _)
{
    rb_yield(value);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.each_value {| value | block } -> hsh
 *     hsh.each_value                    -> an_enumerator
 *
 *  Calls <i>block</i> once for each key in <i>hsh</i>, passing the
 *  value as a parameter.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.each_value {|value| puts value }
 *
 *  <em>produces:</em>
 *
 *     100
 *     200
 */

static VALUE
rb_hash_each_value(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_foreach(hash, each_value_i, 0);
    return hash;
}

static int
each_key_i(VALUE key, VALUE value, VALUE _)
{
    rb_yield(key);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.each_key {| key | block } -> hsh
 *     hsh.each_key                  -> an_enumerator
 *
 *  Calls <i>block</i> once for each key in <i>hsh</i>, passing the key
 *  as a parameter.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.each_key {|key| puts key }
 *
 *  <em>produces:</em>
 *
 *     a
 *     b
 */
static VALUE
rb_hash_each_key(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_foreach(hash, each_key_i, 0);
    return hash;
}

static int
each_pair_i(VALUE key, VALUE value, VALUE _)
{
    rb_yield(rb_assoc_new(key, value));
    return ST_CONTINUE;
}

static int
each_pair_i_fast(VALUE key, VALUE value, VALUE _)
{
    VALUE argv[2];
    argv[0] = key;
    argv[1] = value;
    rb_yield_values2(2, argv);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.each      {| key, value | block } -> hsh
 *     hsh.each_pair {| key, value | block } -> hsh
 *     hsh.each                              -> an_enumerator
 *     hsh.each_pair                         -> an_enumerator
 *
 *  Calls <i>block</i> once for each key in <i>hsh</i>, passing the key-value
 *  pair as parameters.
 *
 *  If no block is given, an enumerator is returned instead.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.each {|key, value| puts "#{key} is #{value}" }
 *
 *  <em>produces:</em>
 *
 *     a is 100
 *     b is 200
 *
 */

static VALUE
rb_hash_each_pair(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    if (rb_block_arity() > 1)
	rb_hash_foreach(hash, each_pair_i_fast, 0);
    else
	rb_hash_foreach(hash, each_pair_i, 0);
    return hash;
}

static int
transform_keys_i(VALUE key, VALUE value, VALUE result)
{
    VALUE new_key = rb_yield(key);
    rb_hash_aset(result, new_key, value);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.transform_keys {|key| block } -> new_hash
 *     hsh.transform_keys                -> an_enumerator
 *
 *  Returns a new hash with the results of running the block once for
 *  every key.
 *  This method does not change the values.
 *
 *     h = { a: 1, b: 2, c: 3 }
 *     h.transform_keys {|k| k.to_s }  #=> { "a" => 1, "b" => 2, "c" => 3 }
 *     h.transform_keys(&:to_s)        #=> { "a" => 1, "b" => 2, "c" => 3 }
 *     h.transform_keys.with_index {|k, i| "#{k}.#{i}" }
 *                                     #=> { "a.0" => 1, "b.1" => 2, "c.2" => 3 }
 *
 *  If no block is given, an enumerator is returned instead.
 */
static VALUE
rb_hash_transform_keys(VALUE hash)
{
    VALUE result;

    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    result = rb_hash_new();
    if (!RHASH_EMPTY_P(hash)) {
        rb_hash_foreach(hash, transform_keys_i, result);
    }

    return result;
}

static VALUE rb_hash_flatten(int argc, VALUE *argv, VALUE hash);

/*
 *  call-seq:
 *     hsh.transform_keys! {|key| block } -> hsh
 *     hsh.transform_keys!                -> an_enumerator
 *
 *  Invokes the given block once for each key in <i>hsh</i>, replacing it
 *  with the new key returned by the block, and then returns <i>hsh</i>.
 *  This method does not change the values.
 *
 *     h = { a: 1, b: 2, c: 3 }
 *     h.transform_keys! {|k| k.to_s }  #=> { "a" => 1, "b" => 2, "c" => 3 }
 *     h.transform_keys!(&:to_sym)      #=> { a: 1, b: 2, c: 3 }
 *     h.transform_keys!.with_index {|k, i| "#{k}.#{i}" }
 *                                      #=> { "a.0" => 1, "b.1" => 2, "c.2" => 3 }
 *
 *  If no block is given, an enumerator is returned instead.
 */
static VALUE
rb_hash_transform_keys_bang(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_modify_check(hash);
    if (!RHASH_TABLE_EMPTY_P(hash)) {
        long i;
        VALUE pairs = rb_hash_flatten(0, NULL, hash);
        rb_hash_clear(hash);
        for (i = 0; i < RARRAY_LEN(pairs); i += 2) {
            VALUE key = RARRAY_AREF(pairs, i), new_key = rb_yield(key),
                  val = RARRAY_AREF(pairs, i+1);
            rb_hash_aset(hash, new_key, val);
        }
    }
    return hash;
}

static int
transform_values_foreach_func(st_data_t key, st_data_t value, st_data_t argp, int error)
{
    return ST_REPLACE;
}

static int
transform_values_foreach_replace(st_data_t *key, st_data_t *value, st_data_t argp, int existing)
{
    VALUE new_value = rb_yield((VALUE)*value);
    *value = new_value;
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.transform_values {|value| block } -> new_hash
 *     hsh.transform_values                  -> an_enumerator
 *
 *  Returns a new hash with the results of running the block once for
 *  every value.
 *  This method does not change the keys.
 *
 *     h = { a: 1, b: 2, c: 3 }
 *     h.transform_values {|v| v * v + 1 }  #=> { a: 2, b: 5, c: 10 }
 *     h.transform_values(&:to_s)           #=> { a: "1", b: "2", c: "3" }
 *     h.transform_values.with_index {|v, i| "#{v}.#{i}" }
 *                                          #=> { a: "1.0", b: "2.1", c: "3.2" }
 *
 *  If no block is given, an enumerator is returned instead.
 */
static VALUE
rb_hash_transform_values(VALUE hash)
{
    VALUE result;

    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    result = hash_dup(hash, rb_cHash, 0);

    if (!RHASH_EMPTY_P(hash)) {
        rb_hash_stlike_foreach_with_replace(result, transform_values_foreach_func, transform_values_foreach_replace, 0);
    }

    return result;
}

/*
 *  call-seq:
 *     hsh.transform_values! {|value| block } -> hsh
 *     hsh.transform_values!                  -> an_enumerator
 *
 *  Invokes the given block once for each value in <i>hsh</i>, replacing it
 *  with the new value returned by the block, and then returns <i>hsh</i>.
 *  This method does not change the keys.
 *
 *     h = { a: 1, b: 2, c: 3 }
 *     h.transform_values! {|v| v * v + 1 }  #=> { a: 2, b: 5, c: 10 }
 *     h.transform_values!(&:to_s)           #=> { a: "2", b: "5", c: "10" }
 *     h.transform_values!.with_index {|v, i| "#{v}.#{i}" }
 *                                           #=> { a: "2.0", b: "5.1", c: "10.2" }
 *
 *  If no block is given, an enumerator is returned instead.
 */
static VALUE
rb_hash_transform_values_bang(VALUE hash)
{
    RETURN_SIZED_ENUMERATOR(hash, 0, 0, hash_enum_size);
    rb_hash_modify_check(hash);

    if (!RHASH_TABLE_EMPTY_P(hash)) {
        rb_hash_stlike_foreach_with_replace(hash, transform_values_foreach_func, transform_values_foreach_replace, 0);
    }

    return hash;
}

static int
to_a_i(VALUE key, VALUE value, VALUE ary)
{
    rb_ary_push(ary, rb_assoc_new(key, value));
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.to_a -> array
 *
 *  Converts <i>hsh</i> to a nested array of <code>[</code> <i>key,
 *  value</i> <code>]</code> arrays.
 *
 *     h = { "c" => 300, "a" => 100, "d" => 400, "c" => 300  }
 *     h.to_a   #=> [["c", 300], ["a", 100], ["d", 400]]
 */

static VALUE
rb_hash_to_a(VALUE hash)
{
    VALUE ary;

    ary = rb_ary_new_capa(RHASH_SIZE(hash));
    rb_hash_foreach(hash, to_a_i, ary);

    return ary;
}

static int
inspect_i(VALUE key, VALUE value, VALUE str)
{
    VALUE str2;

    str2 = rb_inspect(key);
    if (RSTRING_LEN(str) > 1) {
	rb_str_buf_cat_ascii(str, ", ");
    }
    else {
	rb_enc_copy(str, str2);
    }
    rb_str_buf_append(str, str2);
    rb_str_buf_cat_ascii(str, "=>");
    str2 = rb_inspect(value);
    rb_str_buf_append(str, str2);

    return ST_CONTINUE;
}

static VALUE
inspect_hash(VALUE hash, VALUE dummy, int recur)
{
    VALUE str;

    if (recur) return rb_usascii_str_new2("{...}");
    str = rb_str_buf_new2("{");
    rb_hash_foreach(hash, inspect_i, str);
    rb_str_buf_cat2(str, "}");

    return str;
}

/*
 * call-seq:
 *   hsh.to_s     -> string
 *   hsh.inspect  -> string
 *
 * Return the contents of this hash as a string.
 *
 *     h = { "c" => 300, "a" => 100, "d" => 400, "c" => 300  }
 *     h.to_s   #=> "{\"c\"=>300, \"a\"=>100, \"d\"=>400}"
 */

static VALUE
rb_hash_inspect(VALUE hash)
{
    if (RHASH_EMPTY_P(hash))
	return rb_usascii_str_new2("{}");
    return rb_exec_recursive(inspect_hash, hash, 0);
}

/*
 * call-seq:
 *    hsh.to_hash   => hsh
 *
 * Returns +self+.
 */

static VALUE
rb_hash_to_hash(VALUE hash)
{
    return hash;
}

VALUE
rb_hash_set_pair(VALUE hash, VALUE arg)
{
    VALUE pair;

    pair = rb_check_array_type(arg);
    if (NIL_P(pair)) {
        rb_raise(rb_eTypeError, "wrong element type %s (expected array)",
                 rb_builtin_class_name(arg));
    }
    if (RARRAY_LEN(pair) != 2) {
        rb_raise(rb_eArgError, "element has wrong array length (expected 2, was %ld)",
                 RARRAY_LEN(pair));
    }
    rb_hash_aset(hash, RARRAY_AREF(pair, 0), RARRAY_AREF(pair, 1));
    return hash;
}

static int
to_h_i(VALUE key, VALUE value, VALUE hash)
{
    rb_hash_set_pair(hash, rb_yield_values(2, key, value));
    return ST_CONTINUE;
}

static VALUE
rb_hash_to_h_block(VALUE hash)
{
    VALUE h = rb_hash_new_with_size(RHASH_SIZE(hash));
    rb_hash_foreach(hash, to_h_i, h);
    return h;
}

/*
 *  call-seq:
 *     hsh.to_h                         -> hsh or new_hash
 *     hsh.to_h {|key, value| block }   -> new_hash
 *
 *  Returns +self+. If called on a subclass of Hash, converts
 *  the receiver to a Hash object.
 *
 *  If a block is given, the results of the block on each pair of
 *  the receiver will be used as pairs.
 */

static VALUE
rb_hash_to_h(VALUE hash)
{
    if (rb_block_given_p()) {
        return rb_hash_to_h_block(hash);
    }
    if (rb_obj_class(hash) != rb_cHash) {
	const VALUE flags = RBASIC(hash)->flags;
        hash = hash_dup(hash, rb_cHash, flags & RHASH_PROC_DEFAULT);
    }
    return hash;
}

static int
keys_i(VALUE key, VALUE value, VALUE ary)
{
    rb_ary_push(ary, key);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.keys    -> array
 *
 *  Returns a new array populated with the keys from this hash. See also
 *  Hash#values.
 *
 *     h = { "a" => 100, "b" => 200, "c" => 300, "d" => 400 }
 *     h.keys   #=> ["a", "b", "c", "d"]
 *
 */

MJIT_FUNC_EXPORTED VALUE
rb_hash_keys(VALUE hash)
{
    st_index_t size = RHASH_SIZE(hash);
    VALUE keys =  rb_ary_new_capa(size);

    if (size == 0) return keys;

    if (ST_DATA_COMPATIBLE_P(VALUE)) {
        RARRAY_PTR_USE_TRANSIENT(keys, ptr, {
            if (RHASH_AR_TABLE_P(hash)) {
                size = ar_keys(hash, ptr, size);
            }
            else {
                st_table *table = RHASH_ST_TABLE(hash);
                size = st_keys(table, ptr, size);
            }
        });
        rb_gc_writebarrier_remember(keys);
	rb_ary_set_len(keys, size);
    }
    else {
	rb_hash_foreach(hash, keys_i, keys);
    }

    return keys;
}

static int
values_i(VALUE key, VALUE value, VALUE ary)
{
    rb_ary_push(ary, value);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.values    -> array
 *
 *  Returns a new array populated with the values from <i>hsh</i>. See
 *  also Hash#keys.
 *
 *     h = { "a" => 100, "b" => 200, "c" => 300 }
 *     h.values   #=> [100, 200, 300]
 *
 */

VALUE
rb_hash_values(VALUE hash)
{
    VALUE values;
    st_index_t size = RHASH_SIZE(hash);

    values = rb_ary_new_capa(size);
    if (size == 0) return values;

    if (ST_DATA_COMPATIBLE_P(VALUE)) {
        if (RHASH_AR_TABLE_P(hash)) {
            rb_gc_writebarrier_remember(values);
            RARRAY_PTR_USE_TRANSIENT(values, ptr, {
                size = ar_values(hash, ptr, size);
            });
        }
        else if (RHASH_ST_TABLE_P(hash)) {
            st_table *table = RHASH_ST_TABLE(hash);
            rb_gc_writebarrier_remember(values);
            RARRAY_PTR_USE_TRANSIENT(values, ptr, {
                size = st_values(table, ptr, size);
            });
        }
	rb_ary_set_len(values, size);
    }
    else {
	rb_hash_foreach(hash, values_i, values);
    }

    return values;
}

/*
 *  call-seq:
 *     hsh.has_key?(key)    -> true or false
 *     hsh.include?(key)    -> true or false
 *     hsh.key?(key)        -> true or false
 *     hsh.member?(key)     -> true or false
 *
 *  Returns <code>true</code> if the given key is present in <i>hsh</i>.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.has_key?("a")   #=> true
 *     h.has_key?("z")   #=> false
 *
 *  Note that #include? and #member? do not test member
 *  equality using <code>==</code> as do other Enumerables.
 *
 *  See also Enumerable#include?
 */

MJIT_FUNC_EXPORTED VALUE
rb_hash_has_key(VALUE hash, VALUE key)
{
    if (hash_stlike_lookup(hash, key, NULL)) {
        return Qtrue;
    }
    else {
        return Qfalse;
    }
}

static int
rb_hash_search_value(VALUE key, VALUE value, VALUE arg)
{
    VALUE *data = (VALUE *)arg;

    if (rb_equal(value, data[1])) {
	data[0] = Qtrue;
	return ST_STOP;
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.has_value?(value)    -> true or false
 *     hsh.value?(value)        -> true or false
 *
 *  Returns <code>true</code> if the given value is present for some key
 *  in <i>hsh</i>.
 *
 *     h = { "a" => 100, "b" => 200 }
 *     h.value?(100)   #=> true
 *     h.value?(999)   #=> false
 */

static VALUE
rb_hash_has_value(VALUE hash, VALUE val)
{
    VALUE data[2];

    data[0] = Qfalse;
    data[1] = val;
    rb_hash_foreach(hash, rb_hash_search_value, (VALUE)data);
    return data[0];
}

struct equal_data {
    VALUE result;
    VALUE hash;
    int eql;
};

static int
eql_i(VALUE key, VALUE val1, VALUE arg)
{
    struct equal_data *data = (struct equal_data *)arg;
    st_data_t val2;

    if (!hash_stlike_lookup(data->hash, key, &val2)) {
        data->result = Qfalse;
        return ST_STOP;
    }
    else {
        if (!(data->eql ? rb_eql(val1, (VALUE)val2) : (int)rb_equal(val1, (VALUE)val2))) {
            data->result = Qfalse;
            return ST_STOP;
        }
        return ST_CONTINUE;
    }
}

static VALUE
recursive_eql(VALUE hash, VALUE dt, int recur)
{
    struct equal_data *data;

    if (recur) return Qtrue;	/* Subtle! */
    data = (struct equal_data*)dt;
    data->result = Qtrue;
    rb_hash_foreach(hash, eql_i, dt);

    return data->result;
}

static VALUE
hash_equal(VALUE hash1, VALUE hash2, int eql)
{
    struct equal_data data;

    if (hash1 == hash2) return Qtrue;
    if (!RB_TYPE_P(hash2, T_HASH)) {
	if (!rb_respond_to(hash2, idTo_hash)) {
	    return Qfalse;
	}
	if (eql) {
	    if (rb_eql(hash2, hash1)) {
		return Qtrue;
	    }
	    else {
		return Qfalse;
	    }
	}
	else {
	    return rb_equal(hash2, hash1);
	}
    }
    if (RHASH_SIZE(hash1) != RHASH_SIZE(hash2))
	return Qfalse;
    if (!RHASH_TABLE_EMPTY_P(hash1) && !RHASH_TABLE_EMPTY_P(hash2)) {
        if (RHASH_TYPE(hash1) != RHASH_TYPE(hash2)) {
            return Qfalse;
        }
        else {
            data.hash = hash2;
            data.eql = eql;
            return rb_exec_recursive_paired(recursive_eql, hash1, hash2, (VALUE)&data);
        }
    }

#if 0
    if (!(rb_equal(RHASH_IFNONE(hash1), RHASH_IFNONE(hash2)) &&
          FL_TEST(hash1, RHASH_PROC_DEFAULT) == FL_TEST(hash2, RHASH_PROC_DEFAULT)))
	return Qfalse;
#endif
    return Qtrue;
}

/*
 *  call-seq:
 *     hsh == other_hash    -> true or false
 *
 *  Equality---Two hashes are equal if they each contain the same number
 *  of keys and if each key-value pair is equal to (according to
 *  Object#==) the corresponding elements in the other hash.
 *
 *     h1 = { "a" => 1, "c" => 2 }
 *     h2 = { 7 => 35, "c" => 2, "a" => 1 }
 *     h3 = { "a" => 1, "c" => 2, 7 => 35 }
 *     h4 = { "a" => 1, "d" => 2, "f" => 35 }
 *     h1 == h2   #=> false
 *     h2 == h3   #=> true
 *     h3 == h4   #=> false
 *
 *  The orders of each hashes are not compared.
 *
 *     h1 = { "a" => 1, "c" => 2 }
 *     h2 = { "c" => 2, "a" => 1 }
 *     h1 == h2   #=> true
 *
 */

static VALUE
rb_hash_equal(VALUE hash1, VALUE hash2)
{
    return hash_equal(hash1, hash2, FALSE);
}

/*
 *  call-seq:
 *     hash.eql?(other)  -> true or false
 *
 *  Returns <code>true</code> if <i>hash</i> and <i>other</i> are
 *  both hashes with the same content.
 *  The orders of each hashes are not compared.
 */

static VALUE
rb_hash_eql(VALUE hash1, VALUE hash2)
{
    return hash_equal(hash1, hash2, TRUE);
}

static int
hash_i(VALUE key, VALUE val, VALUE arg)
{
    st_index_t *hval = (st_index_t *)arg;
    st_index_t hdata[2];

    hdata[0] = rb_hash(key);
    hdata[1] = rb_hash(val);
    *hval ^= st_hash(hdata, sizeof(hdata), 0);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.hash   -> integer
 *
 *  Compute a hash-code for this hash. Two hashes with the same content
 *  will have the same hash code (and will compare using <code>eql?</code>).
 *
 *  See also Object#hash.
 */

static VALUE
rb_hash_hash(VALUE hash)
{
    st_index_t size = RHASH_SIZE(hash);
    st_index_t hval = rb_hash_start(size);
    hval = rb_hash_uint(hval, (st_index_t)rb_hash_hash);
    if (size) {
	rb_hash_foreach(hash, hash_i, (VALUE)&hval);
    }
    hval = rb_hash_end(hval);
    return ST2FIX(hval);
}

static int
rb_hash_invert_i(VALUE key, VALUE value, VALUE hash)
{
    rb_hash_aset(hash, value, key);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.invert -> new_hash
 *
 *  Returns a new hash created by using <i>hsh</i>'s values as keys, and
 *  the keys as values.
 *  If a key with the same value already exists in the <i>hsh</i>, then
 *  the last one defined will be used, the earlier value(s) will be discarded.
 *
 *     h = { "n" => 100, "m" => 100, "y" => 300, "d" => 200, "a" => 0 }
 *     h.invert   #=> {0=>"a", 100=>"m", 200=>"d", 300=>"y"}
 *
 *  If there is no key with the same value, Hash#invert is involutive.
 *
 *    h = { a: 1, b: 3, c: 4 }
 *    h.invert.invert == h #=> true
 *
 *  The condition, no key with the same value, can be tested by comparing
 *  the size of inverted hash.
 *
 *    # no key with the same value
 *    h = { a: 1, b: 3, c: 4 }
 *    h.size == h.invert.size #=> true
 *
 *    # two (or more) keys has the same value
 *    h = { a: 1, b: 3, c: 1 }
 *    h.size == h.invert.size #=> false
 *
 */

static VALUE
rb_hash_invert(VALUE hash)
{
    VALUE h = rb_hash_new_with_size(RHASH_SIZE(hash));

    rb_hash_foreach(hash, rb_hash_invert_i, h);
    return h;
}

static int
rb_hash_update_callback(st_data_t *key, st_data_t *value, struct update_arg *arg, int existing)
{
    if (existing) {
	arg->old_value = *value;
	arg->new_value = arg->arg;
    }
    else {
	arg->new_key = *key;
	arg->new_value = arg->arg;
    }
    *value = arg->arg;
    return ST_CONTINUE;
}

NOINSERT_UPDATE_CALLBACK(rb_hash_update_callback)

static int
rb_hash_update_i(VALUE key, VALUE value, VALUE hash)
{
    RHASH_UPDATE(hash, key, rb_hash_update_callback, value);
    return ST_CONTINUE;
}

static int
rb_hash_update_block_callback(st_data_t *key, st_data_t *value, struct update_arg *arg, int existing)
{
    VALUE newvalue = (VALUE)arg->arg;

    if (existing) {
	newvalue = rb_yield_values(3, (VALUE)*key, (VALUE)*value, newvalue);
	arg->old_value = *value;
    }
    else {
	arg->new_key = *key;
    }
    arg->new_value = newvalue;
    *value = newvalue;
    return ST_CONTINUE;
}

NOINSERT_UPDATE_CALLBACK(rb_hash_update_block_callback)

static int
rb_hash_update_block_i(VALUE key, VALUE value, VALUE hash)
{
    RHASH_UPDATE(hash, key, rb_hash_update_block_callback, value);
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.merge!(other_hash1, other_hash2, ...)              -> hsh
 *     hsh.update(other_hash1, other_hash2, ...)              -> hsh
 *     hsh.merge!(other_hash1, other_hash2, ...) {|key, oldval, newval| block}
 *                                                            -> hsh
 *     hsh.update(other_hash1, other_hash2, ...) {|key, oldval, newval| block}
 *                                                            -> hsh
 *
 *  Adds the contents of the given hashes to the receiver.
 *
 *  If no block is given, entries with duplicate keys are overwritten
 *  with the values from each +other_hash+ successively,
 *  otherwise the value for each duplicate key is determined by
 *  calling the block with the key, its value in the receiver and
 *  its value in each +other_hash+.
 *
 *     h1 = { "a" => 100, "b" => 200 }
 *     h1.merge!          #=> {"a"=>100, "b"=>200}
 *     h1                 #=> {"a"=>100, "b"=>200}
 *
 *     h1 = { "a" => 100, "b" => 200 }
 *     h2 = { "b" => 246, "c" => 300 }
 *     h1.merge!(h2)      #=> {"a"=>100, "b"=>246, "c"=>300}
 *     h1                 #=> {"a"=>100, "b"=>246, "c"=>300}
 *
 *     h1 = { "a" => 100, "b" => 200 }
 *     h2 = { "b" => 246, "c" => 300 }
 *     h3 = { "b" => 357, "d" => 400 }
 *     h1.merge!(h2, h3)
 *                        #=> {"a"=>100, "b"=>357, "c"=>300, "d"=>400}
 *     h1                 #=> {"a"=>100, "b"=>357, "c"=>300, "d"=>400}
 *
 *     h1 = { "a" => 100, "b" => 200 }
 *     h2 = { "b" => 246, "c" => 300 }
 *     h3 = { "b" => 357, "d" => 400 }
 *     h1.merge!(h2, h3) {|key, v1, v2| v1 }
 *                        #=> {"a"=>100, "b"=>200, "c"=>300, "d"=>400}
 *     h1                 #=> {"a"=>100, "b"=>200, "c"=>300, "d"=>400}
 *
 *  Hash#update is an alias for Hash#merge!.
 */

static VALUE
rb_hash_update(int argc, VALUE *argv, VALUE self)
{
    int i;
    bool block_given = rb_block_given_p();

    rb_hash_modify(self);
    for (i = 0; i < argc; i++){
       VALUE hash = to_hash(argv[i]);
       if (block_given) {
           rb_hash_foreach(hash, rb_hash_update_block_i, self);
       }
       else {
           rb_hash_foreach(hash, rb_hash_update_i, self);
       }
    }
    return self;
}

struct update_func_arg {
    VALUE hash;
    VALUE value;
    rb_hash_update_func *func;
};

static int
rb_hash_update_func_callback(st_data_t *key, st_data_t *value, struct update_arg *arg, int existing)
{
    struct update_func_arg *uf_arg = (struct update_func_arg *)arg->arg;
    VALUE newvalue = uf_arg->value;

    if (existing) {
	newvalue = (*uf_arg->func)((VALUE)*key, (VALUE)*value, newvalue);
	arg->old_value = *value;
    }
    else {
	arg->new_key = *key;
    }
    arg->new_value = newvalue;
    *value = newvalue;
    return ST_CONTINUE;
}

NOINSERT_UPDATE_CALLBACK(rb_hash_update_func_callback)

static int
rb_hash_update_func_i(VALUE key, VALUE value, VALUE arg0)
{
    struct update_func_arg *arg = (struct update_func_arg *)arg0;
    VALUE hash = arg->hash;

    arg->value = value;
    RHASH_UPDATE(hash, key, rb_hash_update_func_callback, (VALUE)arg);
    return ST_CONTINUE;
}

VALUE
rb_hash_update_by(VALUE hash1, VALUE hash2, rb_hash_update_func *func)
{
    rb_hash_modify(hash1);
    hash2 = to_hash(hash2);
    if (func) {
	struct update_func_arg arg;
	arg.hash = hash1;
	arg.func = func;
	rb_hash_foreach(hash2, rb_hash_update_func_i, (VALUE)&arg);
    }
    else {
	rb_hash_foreach(hash2, rb_hash_update_i, hash1);
    }
    return hash1;
}

/*
 *  call-seq:
 *     hsh.merge(other_hash1, other_hash2, ...)           -> new_hash
 *     hsh.merge(other_hash1, other_hash2, ...) {|key, oldval, newval| block}
 *                                                        -> new_hash
 *
 *  Returns a new hash that combines the contents of the receiver and
 *  the contents of the given hashes.
 *
 *  If no block is given, entries with duplicate keys are overwritten
 *  with the values from each +other_hash+ successively,
 *  otherwise the value for each duplicate key is determined by
 *  calling the block with the key, its value in the receiver and
 *  its value in each +other_hash+.
 *
 *  When called without any argument, returns a copy of the receiver.
 *
 *     h1 = { "a" => 100, "b" => 200 }
 *     h2 = { "b" => 246, "c" => 300 }
 *     h3 = { "b" => 357, "d" => 400 }
 *     h1.merge          #=> {"a"=>100, "b"=>200}
 *     h1.merge(h2)      #=> {"a"=>100, "b"=>246, "c"=>300}
 *     h1.merge(h2, h3)  #=> {"a"=>100, "b"=>357, "c"=>300, "d"=>400}
 *     h1.merge(h2) {|key, oldval, newval| newval - oldval}
 *                       #=> {"a"=>100, "b"=>46,  "c"=>300}
 *     h1.merge(h2, h3) {|key, oldval, newval| newval - oldval}
 *                       #=> {"a"=>100, "b"=>311, "c"=>300, "d"=>400}
 *     h1                #=> {"a"=>100, "b"=>200}
 *
 */

static VALUE
rb_hash_merge(int argc, VALUE *argv, VALUE self)
{
    return rb_hash_update(argc, argv, rb_hash_dup(self));
}

static int
assoc_cmp(VALUE a, VALUE b)
{
    return !RTEST(rb_equal(a, b));
}

static VALUE
lookup2_call(VALUE arg)
{
    VALUE *args = (VALUE *)arg;
    return rb_hash_lookup2(args[0], args[1], Qundef);
}

struct reset_hash_type_arg {
    VALUE hash;
    const struct st_hash_type *orighash;
};

static VALUE
reset_hash_type(VALUE arg)
{
    struct reset_hash_type_arg *p = (struct reset_hash_type_arg *)arg;
    HASH_ASSERT(RHASH_ST_TABLE_P(p->hash));
    RHASH_ST_TABLE(p->hash)->type = p->orighash;
    return Qundef;
}

static int
assoc_i(VALUE key, VALUE val, VALUE arg)
{
    VALUE *args = (VALUE *)arg;

    if (RTEST(rb_equal(args[0], key))) {
	args[1] = rb_assoc_new(key, val);
	return ST_STOP;
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hash.assoc(obj)   ->  an_array  or  nil
 *
 *  Searches through the hash comparing _obj_ with the key using <code>==</code>.
 *  Returns the key-value pair (two elements array) or +nil+
 *  if no match is found.  See Array#assoc.
 *
 *     h = {"colors"  => ["red", "blue", "green"],
 *          "letters" => ["a", "b", "c" ]}
 *     h.assoc("letters")  #=> ["letters", ["a", "b", "c"]]
 *     h.assoc("foo")      #=> nil
 */

VALUE
rb_hash_assoc(VALUE hash, VALUE key)
{
    st_table *table;
    const struct st_hash_type *orighash;
    VALUE args[2];

    if (RHASH_EMPTY_P(hash)) return Qnil;

    ar_force_convert_table(hash, __FILE__, __LINE__);
    HASH_ASSERT(RHASH_ST_TABLE_P(hash));
    table = RHASH_ST_TABLE(hash);
    orighash = table->type;

    if (orighash != &identhash) {
	VALUE value;
	struct reset_hash_type_arg ensure_arg;
	struct st_hash_type assochash;

	assochash.compare = assoc_cmp;
	assochash.hash = orighash->hash;
        table->type = &assochash;
	args[0] = hash;
	args[1] = key;
	ensure_arg.hash = hash;
	ensure_arg.orighash = orighash;
	value = rb_ensure(lookup2_call, (VALUE)&args, reset_hash_type, (VALUE)&ensure_arg);
	if (value != Qundef) return rb_assoc_new(key, value);
    }

    args[0] = key;
    args[1] = Qnil;
    rb_hash_foreach(hash, assoc_i, (VALUE)args);
    return args[1];
}

static int
rassoc_i(VALUE key, VALUE val, VALUE arg)
{
    VALUE *args = (VALUE *)arg;

    if (RTEST(rb_equal(args[0], val))) {
	args[1] = rb_assoc_new(key, val);
	return ST_STOP;
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hash.rassoc(obj) -> an_array or nil
 *
 *  Searches through the hash comparing _obj_ with the value using <code>==</code>.
 *  Returns the first key-value pair (two-element array) that matches. See
 *  also Array#rassoc.
 *
 *     a = {1=> "one", 2 => "two", 3 => "three", "ii" => "two"}
 *     a.rassoc("two")    #=> [2, "two"]
 *     a.rassoc("four")   #=> nil
 */

VALUE
rb_hash_rassoc(VALUE hash, VALUE obj)
{
    VALUE args[2];

    args[0] = obj;
    args[1] = Qnil;
    rb_hash_foreach(hash, rassoc_i, (VALUE)args);
    return args[1];
}

static int
flatten_i(VALUE key, VALUE val, VALUE ary)
{
    VALUE pair[2];

    pair[0] = key;
    pair[1] = val;
    rb_ary_cat(ary, pair, 2);

    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hash.flatten -> an_array
 *     hash.flatten(level) -> an_array
 *
 *  Returns a new array that is a one-dimensional flattening of this
 *  hash. That is, for every key or value that is an array, extract
 *  its elements into the new array.  Unlike Array#flatten, this
 *  method does not flatten recursively by default.  The optional
 *  <i>level</i> argument determines the level of recursion to flatten.
 *
 *     a =  {1=> "one", 2 => [2,"two"], 3 => "three"}
 *     a.flatten    # => [1, "one", 2, [2, "two"], 3, "three"]
 *     a.flatten(2) # => [1, "one", 2, 2, "two", 3, "three"]
 */

static VALUE
rb_hash_flatten(int argc, VALUE *argv, VALUE hash)
{
    VALUE ary;

    rb_check_arity(argc, 0, 1);

    if (argc) {
	int level = NUM2INT(argv[0]);

	if (level == 0) return rb_hash_to_a(hash);

	ary = rb_ary_new_capa(RHASH_SIZE(hash) * 2);
	rb_hash_foreach(hash, flatten_i, ary);
	level--;

	if (level > 0) {
	    VALUE ary_flatten_level = INT2FIX(level);
	    rb_funcallv(ary, id_flatten_bang, 1, &ary_flatten_level);
	}
	else if (level < 0) {
	    /* flatten recursively */
	    rb_funcallv(ary, id_flatten_bang, 0, 0);
	}
    }
    else {
	ary = rb_ary_new_capa(RHASH_SIZE(hash) * 2);
	rb_hash_foreach(hash, flatten_i, ary);
    }

    return ary;
}

static int
delete_if_nil(VALUE key, VALUE value, VALUE hash)
{
    if (NIL_P(value)) {
	return ST_DELETE;
    }
    return ST_CONTINUE;
}

static int
set_if_not_nil(VALUE key, VALUE value, VALUE hash)
{
    if (!NIL_P(value)) {
	rb_hash_aset(hash, key, value);
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.compact -> new_hash
 *
 *  Returns a new hash with the nil values/key pairs removed
 *
 *     h = { a: 1, b: false, c: nil }
 *     h.compact     #=> { a: 1, b: false }
 *     h             #=> { a: 1, b: false, c: nil }
 *
 */

static VALUE
rb_hash_compact(VALUE hash)
{
    VALUE result = rb_hash_new();
    if (!RHASH_EMPTY_P(hash)) {
	rb_hash_foreach(hash, set_if_not_nil, result);
    }
    return result;
}

/*
 *  call-seq:
 *     hsh.compact! -> hsh or nil
 *
 *  Removes all nil values from the hash.
 *  Returns nil if no changes were made, otherwise returns the hash.
 *
 *     h = { a: 1, b: false, c: nil }
 *     h.compact!     #=> { a: 1, b: false }
 *
 */

static VALUE
rb_hash_compact_bang(VALUE hash)
{
    st_index_t n;
    rb_hash_modify_check(hash);
    n = RHASH_SIZE(hash);
    if (n) {
	rb_hash_foreach(hash, delete_if_nil, hash);
        if (n != RHASH_SIZE(hash))
	    return hash;
    }
    return Qnil;
}

static st_table *rb_init_identtable_with_size(st_index_t size);

/*
 *  call-seq:
 *     hsh.compare_by_identity -> hsh
 *
 *  Makes <i>hsh</i> compare its keys by their identity, i.e. it
 *  will consider exact same objects as same keys.
 *
 *     h1 = { "a" => 100, "b" => 200, :c => "c" }
 *     h1["a"]        #=> 100
 *     h1.compare_by_identity
 *     h1.compare_by_identity? #=> true
 *     h1["a".dup]    #=> nil  # different objects.
 *     h1[:c]         #=> "c"  # same symbols are all same.
 *
 */

static VALUE
rb_hash_compare_by_id(VALUE hash)
{
    VALUE tmp;
    st_table *identtable;

    if (rb_hash_compare_by_id_p(hash)) return hash;

    rb_hash_modify_check(hash);
    ar_force_convert_table(hash, __FILE__, __LINE__);
    HASH_ASSERT(RHASH_ST_TABLE_P(hash));

    tmp = hash_alloc(0);
    identtable = rb_init_identtable_with_size(RHASH_SIZE(hash));
    RHASH_ST_TABLE_SET(tmp, identtable);
    rb_hash_foreach(hash, rb_hash_rehash_i, (VALUE)tmp);
    st_free_table(RHASH_ST_TABLE(hash));
    RHASH_ST_TABLE_SET(hash, identtable);
    RHASH_ST_CLEAR(tmp);
    rb_gc_force_recycle(tmp);

    return hash;
}

/*
 *  call-seq:
 *     hsh.compare_by_identity? -> true or false
 *
 *  Returns <code>true</code> if <i>hsh</i> will compare its keys by
 *  their identity.  Also see Hash#compare_by_identity.
 *
 */

MJIT_FUNC_EXPORTED VALUE
rb_hash_compare_by_id_p(VALUE hash)
{
    if (RHASH_ST_TABLE_P(hash) && RHASH_ST_TABLE(hash)->type == &identhash) {
	return Qtrue;
    }
    else {
        return Qfalse;
    }
}

VALUE
rb_ident_hash_new(void)
{
    VALUE hash = rb_hash_new();
    RHASH_ST_TABLE_SET(hash, st_init_table(&identhash));
    return hash;
}

st_table *
rb_init_identtable(void)
{
    return st_init_table(&identhash);
}

static st_table *
rb_init_identtable_with_size(st_index_t size)
{
    return st_init_table_with_size(&identhash, size);
}

static int
any_p_i(VALUE key, VALUE value, VALUE arg)
{
    VALUE ret = rb_yield(rb_assoc_new(key, value));
    if (RTEST(ret)) {
	*(VALUE *)arg = Qtrue;
	return ST_STOP;
    }
    return ST_CONTINUE;
}

static int
any_p_i_fast(VALUE key, VALUE value, VALUE arg)
{
    VALUE ret = rb_yield_values(2, key, value);
    if (RTEST(ret)) {
	*(VALUE *)arg = Qtrue;
	return ST_STOP;
    }
    return ST_CONTINUE;
}

static int
any_p_i_pattern(VALUE key, VALUE value, VALUE arg)
{
    VALUE ret = rb_funcall(((VALUE *)arg)[1], idEqq, 1, rb_assoc_new(key, value));
    if (RTEST(ret)) {
	*(VALUE *)arg = Qtrue;
	return ST_STOP;
    }
    return ST_CONTINUE;
}

/*
 *  call-seq:
 *     hsh.any? [{ |(key, value)| block }]   -> true or false
 *     hsh.any?(pattern)                     -> true or false
 *
 *  See also Enumerable#any?
 */

static VALUE
rb_hash_any_p(int argc, VALUE *argv, VALUE hash)
{
    VALUE args[2];
    args[0] = Qfalse;

    rb_check_arity(argc, 0, 1);
    if (RHASH_EMPTY_P(hash)) return Qfalse;
    if (argc) {
        if (rb_block_given_p()) {
            rb_warn("given block not used");
        }
	args[1] = argv[0];

	rb_hash_foreach(hash, any_p_i_pattern, (VALUE)args);
    }
    else {
	if (!rb_block_given_p()) {
	    /* yields pairs, never false */
	    return Qtrue;
	}
	if (rb_block_arity() > 1)
	    rb_hash_foreach(hash, any_p_i_fast, (VALUE)args);
	else
	    rb_hash_foreach(hash, any_p_i, (VALUE)args);
    }
    return args[0];
}

/*
 * call-seq:
 *   hsh.dig(key, ...)                 -> object
 *
 * Extracts the nested value specified by the sequence of <i>key</i>
 * objects by calling +dig+ at each step, returning +nil+ if any
 * intermediate step is +nil+.
 *
 *   h = { foo: {bar: {baz: 1}}}
 *
 *   h.dig(:foo, :bar, :baz)     #=> 1
 *   h.dig(:foo, :zot, :xyz)     #=> nil
 *
 *   g = { foo: [10, 11, 12] }
 *   g.dig(:foo, 1)              #=> 11
 *   g.dig(:foo, 1, 0)           #=> TypeError: Integer does not have #dig method
 *   g.dig(:foo, :bar)           #=> TypeError: no implicit conversion of Symbol into Integer
 */

static VALUE
rb_hash_dig(int argc, VALUE *argv, VALUE self)
{
    rb_check_arity(argc, 1, UNLIMITED_ARGUMENTS);
    self = rb_hash_aref(self, *argv);
    if (!--argc) return self;
    ++argv;
    return rb_obj_dig(argc, argv, self, Qnil);
}

static int
hash_le_i(VALUE key, VALUE value, VALUE arg)
{
    VALUE *args = (VALUE *)arg;
    VALUE v = rb_hash_lookup2(args[0], key, Qundef);
    if (v != Qundef && rb_equal(value, v)) return ST_CONTINUE;
    args[1] = Qfalse;
    return ST_STOP;
}

static VALUE
hash_le(VALUE hash1, VALUE hash2)
{
    VALUE args[2];
    args[0] = hash2;
    args[1] = Qtrue;
    rb_hash_foreach(hash1, hash_le_i, (VALUE)args);
    return args[1];
}

/*
 * call-seq:
 *   hash <= other -> true or false
 *
 * Returns <code>true</code> if <i>hash</i> is subset of
 * <i>other</i> or equals to <i>other</i>.
 *
 *    h1 = {a:1, b:2}
 *    h2 = {a:1, b:2, c:3}
 *    h1 <= h2   #=> true
 *    h2 <= h1   #=> false
 *    h1 <= h1   #=> true
 */
static VALUE
rb_hash_le(VALUE hash, VALUE other)
{
    other = to_hash(other);
    if (RHASH_SIZE(hash) > RHASH_SIZE(other)) return Qfalse;
    return hash_le(hash, other);
}

/*
 * call-seq:
 *   hash < other -> true or false
 *
 * Returns <code>true</code> if <i>hash</i> is subset of
 * <i>other</i>.
 *
 *    h1 = {a:1, b:2}
 *    h2 = {a:1, b:2, c:3}
 *    h1 < h2    #=> true
 *    h2 < h1    #=> false
 *    h1 < h1    #=> false
 */
static VALUE
rb_hash_lt(VALUE hash, VALUE other)
{
    other = to_hash(other);
    if (RHASH_SIZE(hash) >= RHASH_SIZE(other)) return Qfalse;
    return hash_le(hash, other);
}

/*
 * call-seq:
 *   hash >= other -> true or false
 *
 * Returns <code>true</code> if <i>other</i> is subset of
 * <i>hash</i> or equals to <i>hash</i>.
 *
 *    h1 = {a:1, b:2}
 *    h2 = {a:1, b:2, c:3}
 *    h1 >= h2   #=> false
 *    h2 >= h1   #=> true
 *    h1 >= h1   #=> true
 */
static VALUE
rb_hash_ge(VALUE hash, VALUE other)
{
    other = to_hash(other);
    if (RHASH_SIZE(hash) < RHASH_SIZE(other)) return Qfalse;
    return hash_le(other, hash);
}

/*
 * call-seq:
 *   hash > other -> true or false
 *
 * Returns <code>true</code> if <i>other</i> is subset of
 * <i>hash</i>.
 *
 *    h1 = {a:1, b:2}
 *    h2 = {a:1, b:2, c:3}
 *    h1 > h2    #=> false
 *    h2 > h1    #=> true
 *    h1 > h1    #=> false
 */
static VALUE
rb_hash_gt(VALUE hash, VALUE other)
{
    other = to_hash(other);
    if (RHASH_SIZE(hash) <= RHASH_SIZE(other)) return Qfalse;
    return hash_le(other, hash);
}

static VALUE
hash_proc_call(RB_BLOCK_CALL_FUNC_ARGLIST(key, hash))
{
    rb_check_arity(argc, 1, 1);
    return rb_hash_aref(hash, *argv);
}

/*
 * call-seq:
 *   hash.to_proc -> proc
 *
 * Returns a Proc which maps keys to values.
 *
 *   h = {a:1, b:2}
 *   hp = h.to_proc
 *   hp.call(:a)          #=> 1
 *   hp.call(:b)          #=> 2
 *   hp.call(:c)          #=> nil
 *   [:a, :b, :c].map(&h) #=> [1, 2, nil]
 */
static VALUE
rb_hash_to_proc(VALUE hash)
{
    return rb_func_proc_new(hash_proc_call, hash);
}

static VALUE
rb_hash_deconstruct_keys(VALUE hash, VALUE keys)
{
    return hash;
}

static int
add_new_i(st_data_t *key, st_data_t *val, st_data_t arg, int existing)
{
    VALUE *args = (VALUE *)arg;
    if (existing) return ST_STOP;
    RB_OBJ_WRITTEN(args[0], Qundef, (VALUE)*key);
    RB_OBJ_WRITE(args[0], (VALUE *)val, args[1]);
    return ST_CONTINUE;
}

/*
 * add +key+ to +val+ pair if +hash+ does not contain +key+.
 * returns non-zero if +key+ was contained.
 */
int
rb_hash_add_new_element(VALUE hash, VALUE key, VALUE val)
{
    st_table *tbl;
    int ret = 0;
    VALUE args[2];
    args[0] = hash;
    args[1] = val;

    if (RHASH_AR_TABLE_P(hash)) {
        hash_ar_table(hash);

        ret = ar_update(hash, (st_data_t)key, add_new_i, (st_data_t)args);
        if (ret != -1) {
            return ret;
        }
        ar_try_convert_table(hash);
    }
    tbl = RHASH_TBL_RAW(hash);
    return st_update(tbl, (st_data_t)key, add_new_i, (st_data_t)args);

}

static st_data_t
key_stringify(VALUE key)
{
    return (rb_obj_class(key) == rb_cString && !RB_OBJ_FROZEN(key)) ?
        rb_hash_key_str(key) : key;
}

static void
ar_bulk_insert(VALUE hash, long argc, const VALUE *argv)
{
    long i;
    for (i = 0; i < argc; ) {
        st_data_t k = key_stringify(argv[i++]);
        st_data_t v = argv[i++];
        ar_insert(hash, k, v);
        RB_OBJ_WRITTEN(hash, Qundef, k);
        RB_OBJ_WRITTEN(hash, Qundef, v);
    }
}

void
rb_hash_bulk_insert(long argc, const VALUE *argv, VALUE hash)
{
    HASH_ASSERT(argc % 2 == 0);
    if (argc > 0) {
        st_index_t size = argc / 2;

        if (RHASH_TABLE_NULL_P(hash)) {
            if (size <= RHASH_AR_TABLE_MAX_SIZE) {
                hash_ar_table(hash);
            }
            else {
                RHASH_TBL_RAW(hash);
            }
        }

        if (RHASH_AR_TABLE_P(hash) &&
            (RHASH_AR_TABLE_SIZE(hash) + size <= RHASH_AR_TABLE_MAX_SIZE)) {
            ar_bulk_insert(hash, argc, argv);
        }
        else {
            rb_hash_bulk_insert_into_st_table(argc, argv, hash);
        }
    }
}

static char **origenviron;
#ifdef _WIN32
#define GET_ENVIRON(e) ((e) = rb_w32_get_environ())
#define FREE_ENVIRON(e) rb_w32_free_environ(e)
static char **my_environ;
#undef environ
#define environ my_environ
#undef getenv
static char *(*w32_getenv)(const char*);
static char *
w32_getenv_unknown(const char *name)
{
    char *(*func)(const char*);
    if (rb_locale_encindex() == rb_ascii8bit_encindex()) {
	func = rb_w32_getenv;
    }
    else {
	func = rb_w32_ugetenv;
    }
    /* atomic assignment in flat memory model */
    return (w32_getenv = func)(name);
}
static char *(*w32_getenv)(const char*) = w32_getenv_unknown;
#define getenv(n) w32_getenv(n)
#elif defined(__APPLE__)
#undef environ
#define environ (*_NSGetEnviron())
#define GET_ENVIRON(e) (e)
#define FREE_ENVIRON(e)
#else
extern char **environ;
#define GET_ENVIRON(e) (e)
#define FREE_ENVIRON(e)
#endif
#ifdef ENV_IGNORECASE
#define ENVMATCH(s1, s2) (STRCASECMP((s1), (s2)) == 0)
#define ENVNMATCH(s1, s2, n) (STRNCASECMP((s1), (s2), (n)) == 0)
#else
#define ENVMATCH(n1, n2) (strcmp((n1), (n2)) == 0)
#define ENVNMATCH(s1, s2, n) (memcmp((s1), (s2), (n)) == 0)
#endif

static VALUE
env_enc_str_new(const char *ptr, long len, rb_encoding *enc)
{
#ifdef _WIN32
    rb_encoding *internal = rb_default_internal_encoding();
    const int ecflags = ECONV_INVALID_REPLACE | ECONV_UNDEF_REPLACE;
    rb_encoding *utf8 = rb_utf8_encoding();
    VALUE str = rb_enc_str_new(NULL, 0, (internal ? internal : enc));
    if (NIL_P(rb_str_cat_conv_enc_opts(str, 0, ptr, len, utf8, ecflags, Qnil))) {
        rb_str_initialize(str, ptr, len, NULL);
    }
#else
    VALUE str = rb_external_str_new_with_enc(ptr, len, enc);
#endif

    rb_obj_freeze(str);
    return str;
}

static VALUE
env_enc_str_new_cstr(const char *ptr, rb_encoding *enc)
{
    return env_enc_str_new(ptr, strlen(ptr), enc);
}

static VALUE
env_str_new(const char *ptr, long len)
{
    return env_enc_str_new(ptr, len, rb_locale_encoding());
}

static VALUE
env_str_new2(const char *ptr)
{
    if (!ptr) return Qnil;
    return env_str_new(ptr, strlen(ptr));
}

static const char TZ_ENV[] = "TZ";
extern bool ruby_tz_uptodate_p;

static rb_encoding *
env_encoding_for(const char *name, const char *ptr)
{
    if (ENVMATCH(name, PATH_ENV)) {
	return rb_filesystem_encoding();
    }
    else {
	return rb_locale_encoding();
    }
}

static VALUE
env_name_new(const char *name, const char *ptr)
{
    return env_enc_str_new_cstr(ptr, env_encoding_for(name, ptr));
}

static void *
get_env_cstr(
#ifdef _WIN32
    volatile VALUE *pstr,
#else
    VALUE str,
#endif
    const char *name)
{
#ifdef _WIN32
    VALUE str = *pstr;
#endif
    char *var;
    rb_encoding *enc = rb_enc_get(str);
    if (!rb_enc_asciicompat(enc)) {
	rb_raise(rb_eArgError, "bad environment variable %s: ASCII incompatible encoding: %s",
		 name, rb_enc_name(enc));
    }
#ifdef _WIN32
    if (!rb_enc_str_asciionly_p(str)) {
	*pstr = str = rb_str_conv_enc(str, NULL, rb_utf8_encoding());
    }
#endif
    var = RSTRING_PTR(str);
    if (memchr(var, '\0', RSTRING_LEN(str))) {
	rb_raise(rb_eArgError, "bad environment variable %s: contains null byte", name);
    }
    return rb_str_fill_terminator(str, 1); /* ASCII compatible */
}

#ifdef _WIN32
#define get_env_ptr(var, val) \
    (var = get_env_cstr(&(val), #var))
#else
#define get_env_ptr(var, val) \
    (var = get_env_cstr(val, #var))
#endif

static inline const char *
env_name(volatile VALUE *s)
{
    const char *name;
    SafeStringValue(*s);
    get_env_ptr(name, *s);
    return name;
}

#define env_name(s) env_name(&(s))

static VALUE env_aset(VALUE nm, VALUE val);

static VALUE
env_delete(VALUE name)
{
    const char *nam, *val;

    nam = env_name(name);
    val = getenv(nam);

    /*
     * ENV['TZ'] = nil has a special meaning.
     * TZ is no longer considered up-to-date and ruby call tzset() as needed.
     * It could be useful if sysadmin change /etc/localtime.
     * This hack might works only on Linux glibc.
     */
    if (ENVMATCH(nam, TZ_ENV)) {
        ruby_tz_uptodate_p = FALSE;
    }

    if (val) {
	VALUE value = env_str_new2(val);

	ruby_setenv(nam, 0);
	if (ENVMATCH(nam, PATH_ENV)) {
	    RB_GC_GUARD(name);
	}
	return value;
    }
    return Qnil;
}

/*
 * call-seq:
 *   ENV.delete(name)                  -> value
 *   ENV.delete(name) { |name| block } -> value
 *
 * Deletes the environment variable with +name+ if it exists and returns its value:
 *   ENV['foo'] = '0'
 *   ENV.delete('foo') # => '0'
 * Returns +nil+ if the named environment variable does not exist:
 *   ENV.delete('foo') # => nil
 * If a block given and the environment variable does not exist,
 * yields +name+ to the block and returns +nil+:
 *   ENV.delete('foo') { |name| puts name } # => nil
 *   foo
 * If a block given and the environment variable exists,
 * deletes the environment variable and returns its value (ignoring the block):
 *   ENV['foo'] = '0'
 *   ENV.delete('foo') { |name| fail 'ignored' } # => "0"
 * Raises TypeError if +name+ is not a String and cannot be coerced with \#to_str:
 *   ENV.delete(Object.new) # => TypeError raised
 */
static VALUE
env_delete_m(VALUE obj, VALUE name)
{
    VALUE val;

    val = env_delete(name);
    if (NIL_P(val) && rb_block_given_p()) rb_yield(name);
    return val;
}

/*
 * call-seq:
 *   ENV[name] -> value
 *
 * Returns the value for the environment variable +name+ if it exists:
 *   ENV['foo'] = 'bar'
 *   ENV['foo'] # => "bar"
 * Returns nil if the named variable does not exist:
 *   ENV.delete('foo')
 *   ENV['foo'] # => nil
 * Raises TypeError if +name+ is not a String and cannot be coerced with \#to_str:
 *   ENV.delete(Object.new) # => TypeError raised
 */
static VALUE
rb_f_getenv(VALUE obj, VALUE name)
{
    const char *nam, *env;

    nam = env_name(name);
    env = getenv(nam);
    if (env) {
	return env_name_new(nam, env);
    }
    return Qnil;
}

/*
 * call-seq:
 *   ENV.fetch(name)                  -> value
 *   ENV.fetch(name, default)         -> value
 *   ENV.fetch(name) { |name| block } -> value
 *
 * If +name+ is the name of an environment variable, returns its value:
 *   ENV['foo'] = 'bar'
 *   ENV.fetch('foo') # => "bar"
 * Otherwise if a block is given (but not a default value),
 * yields +name+ to the block and returns the block's return value:
 *   ENV.fetch('foo') { |name| :need_not_return_a_string } # => :need_not_return_a_string
 * Otherwise if a default value is given (but not a block), returns the default value:
 *   ENV.delete('foo')
 *   ENV.fetch('foo', :default_need_not_be_a_string) # => :default_need_not_be_a_string
 * If the environment variable does not exist and both default and block are given,
 * issues a warning ("warning: block supersedes default value argument"),
 * yields +name+ to the block, and returns the block's return value:
 *   ENV.fetch('foo', :default) { |name| :block_return } # => :block_return
 * Raises TypeError if +name+ is not a String and cannot be coerced with \#to_str:
 *   ENV.delete(Object.new) # => TypeError raised
 * Raises KeyError if +name+ is a String, but is not found,
 * and neither default value nor block is given:
 *   ENV.fetch('foo') # => KeyError raised
 */
static VALUE
env_fetch(int argc, VALUE *argv, VALUE _)
{
    VALUE key;
    long block_given;
    const char *nam, *env;

    rb_check_arity(argc, 1, 2);
    key = argv[0];
    block_given = rb_block_given_p();
    if (block_given && argc == 2) {
	rb_warn("block supersedes default value argument");
    }
    nam = env_name(key);
    env = getenv(nam);
    if (!env) {
	if (block_given) return rb_yield(key);
	if (argc == 1) {
	    rb_key_err_raise(rb_sprintf("key not found: \"%"PRIsVALUE"\"", key), envtbl, key);
	}
	return argv[1];
    }
    return env_name_new(nam, env);
}

int
rb_env_path_tainted(void)
{
    rb_warning("rb_env_path_tainted is deprecated and will be removed in Ruby 3.2.");
    return 0;
}

#if defined(_WIN32) || (defined(HAVE_SETENV) && defined(HAVE_UNSETENV))
#elif defined __sun
static int
in_origenv(const char *str)
{
    char **env;
    for (env = origenviron; *env; ++env) {
	if (*env == str) return 1;
    }
    return 0;
}
#else
static int
envix(const char *nam)
{
    register int i, len = strlen(nam);
    char **env;

    env = GET_ENVIRON(environ);
    for (i = 0; env[i]; i++) {
	if (ENVNMATCH(env[i],nam,len) && env[i][len] == '=')
	    break;			/* memcmp must come first to avoid */
    }					/* potential SEGV's */
    FREE_ENVIRON(environ);
    return i;
}
#endif

#if defined(_WIN32)
static size_t
getenvsize(const WCHAR* p)
{
    const WCHAR* porg = p;
    while (*p++) p += lstrlenW(p) + 1;
    return p - porg + 1;
}

static size_t
getenvblocksize(void)
{
#ifdef _MAX_ENV
    return _MAX_ENV;
#else
    return 32767;
#endif
}

static int
check_envsize(size_t n)
{
    if (_WIN32_WINNT < 0x0600 && rb_w32_osver() < 6) {
	/* https://msdn.microsoft.com/en-us/library/windows/desktop/ms682653(v=vs.85).aspx */
	/* Windows Server 2003 and Windows XP: The maximum size of the
	 * environment block for the process is 32,767 characters. */
	WCHAR* p = GetEnvironmentStringsW();
	if (!p) return -1; /* never happen */
	n += getenvsize(p);
	FreeEnvironmentStringsW(p);
	if (n >= getenvblocksize()) {
	    return -1;
	}
    }
    return 0;
}
#endif

#if defined(_WIN32) || \
  (defined(__sun) && !(defined(HAVE_SETENV) && defined(HAVE_UNSETENV)))

NORETURN(static void invalid_envname(const char *name));

static void
invalid_envname(const char *name)
{
    rb_syserr_fail_str(EINVAL, rb_sprintf("ruby_setenv(%s)", name));
}

static const char *
check_envname(const char *name)
{
    if (strchr(name, '=')) {
	invalid_envname(name);
    }
    return name;
}
#endif

void
ruby_setenv(const char *name, const char *value)
{
#if defined(_WIN32)
# if defined(MINGW_HAS_SECURE_API) || RUBY_MSVCRT_VERSION >= 80
#   define HAVE__WPUTENV_S 1
# endif
    VALUE buf;
    WCHAR *wname;
    WCHAR *wvalue = 0;
    int failed = 0;
    int len;
    check_envname(name);
    len = MultiByteToWideChar(CP_UTF8, 0, name, -1, NULL, 0);
    if (value) {
	int len2;
	len2 = MultiByteToWideChar(CP_UTF8, 0, value, -1, NULL, 0);
	if (check_envsize((size_t)len + len2)) { /* len and len2 include '\0' */
	    goto fail;  /* 2 for '=' & '\0' */
	}
	wname = ALLOCV_N(WCHAR, buf, len + len2);
	wvalue = wname + len;
	MultiByteToWideChar(CP_UTF8, 0, name, -1, wname, len);
	MultiByteToWideChar(CP_UTF8, 0, value, -1, wvalue, len2);
#ifndef HAVE__WPUTENV_S
	wname[len-1] = L'=';
#endif
    }
    else {
	wname = ALLOCV_N(WCHAR, buf, len + 1);
	MultiByteToWideChar(CP_UTF8, 0, name, -1, wname, len);
	wvalue = wname + len;
	*wvalue = L'\0';
#ifndef HAVE__WPUTENV_S
	wname[len-1] = L'=';
#endif
    }
#ifndef HAVE__WPUTENV_S
    failed = _wputenv(wname);
#else
    failed = _wputenv_s(wname, wvalue);
#endif
    ALLOCV_END(buf);
    /* even if putenv() failed, clean up and try to delete the
     * variable from the system area. */
    if (!value || !*value) {
	/* putenv() doesn't handle empty value */
	if (!SetEnvironmentVariable(name, value) &&
	    GetLastError() != ERROR_ENVVAR_NOT_FOUND) goto fail;
    }
    if (failed) {
      fail:
	invalid_envname(name);
    }
#elif defined(HAVE_SETENV) && defined(HAVE_UNSETENV)
    if (value) {
	if (setenv(name, value, 1))
	    rb_sys_fail_str(rb_sprintf("setenv(%s)", name));
    }
    else {
#ifdef VOID_UNSETENV
	unsetenv(name);
#else
	if (unsetenv(name))
	    rb_sys_fail_str(rb_sprintf("unsetenv(%s)", name));
#endif
    }
#elif defined __sun
    /* Solaris 9 (or earlier) does not have setenv(3C) and unsetenv(3C). */
    /* The below code was tested on Solaris 10 by:
         % ./configure ac_cv_func_setenv=no ac_cv_func_unsetenv=no
    */
    size_t len, mem_size;
    char **env_ptr, *str, *mem_ptr;

    check_envname(name);
    len = strlen(name);
    if (value) {
	mem_size = len + strlen(value) + 2;
	mem_ptr = malloc(mem_size);
	if (mem_ptr == NULL)
	    rb_sys_fail_str(rb_sprintf("malloc("PRIuSIZE")", mem_size));
	snprintf(mem_ptr, mem_size, "%s=%s", name, value);
    }
    for (env_ptr = GET_ENVIRON(environ); (str = *env_ptr) != 0; ++env_ptr) {
	if (!strncmp(str, name, len) && str[len] == '=') {
	    if (!in_origenv(str)) free(str);
	    while ((env_ptr[0] = env_ptr[1]) != 0) env_ptr++;
	    break;
	}
    }
    if (value) {
	if (putenv(mem_ptr)) {
	    free(mem_ptr);
	    rb_sys_fail_str(rb_sprintf("putenv(%s)", name));
	}
    }
#else  /* WIN32 */
    size_t len;
    int i;

    i=envix(name);		        /* where does it go? */

    if (environ == origenviron) {	/* need we copy environment? */
	int j;
	int max;
	char **tmpenv;

	for (max = i; environ[max]; max++) ;
	tmpenv = ALLOC_N(char*, max+2);
	for (j=0; j<max; j++)		/* copy environment */
	    tmpenv[j] = ruby_strdup(environ[j]);
	tmpenv[max] = 0;
	environ = tmpenv;		/* tell exec where it is now */
    }
    if (environ[i]) {
	char **envp = origenviron;
	while (*envp && *envp != environ[i]) envp++;
	if (!*envp)
	    xfree(environ[i]);
	if (!value) {
	    while (environ[i]) {
		environ[i] = environ[i+1];
		i++;
	    }
	    return;
	}
    }
    else {			/* does not exist yet */
	if (!value) return;
	REALLOC_N(environ, char*, i+2);	/* just expand it a bit */
	environ[i+1] = 0;	/* make sure it's null terminated */
    }
    len = strlen(name) + strlen(value) + 2;
    environ[i] = ALLOC_N(char, len);
    snprintf(environ[i],len,"%s=%s",name,value); /* all that work just for this */
#endif /* WIN32 */
}

void
ruby_unsetenv(const char *name)
{
    ruby_setenv(name, 0);
}

/*
 * call-seq:
 *   ENV[name] = value
 *   ENV.store(name, value) -> value
 *
 * Sets the environment variable +name+ to +value+.  If the value given is
 * +nil+ the environment variable is deleted.
 * +name+ must be a string.
 *
 */
static VALUE
env_aset_m(VALUE obj, VALUE nm, VALUE val)
{
    return env_aset(nm, val);
}

static VALUE
env_aset(VALUE nm, VALUE val)
{
    char *name, *value;

    if (NIL_P(val)) {
        env_delete(nm);
	return Qnil;
    }
    SafeStringValue(nm);
    SafeStringValue(val);
    /* nm can be modified in `val.to_str`, don't get `name` before
     * check for `val` */
    get_env_ptr(name, nm);
    get_env_ptr(value, val);

    ruby_setenv(name, value);
    if (ENVMATCH(name, PATH_ENV)) {
	RB_GC_GUARD(nm);
    }
    else if (ENVMATCH(name, TZ_ENV)) {
	ruby_tz_uptodate_p = FALSE;
    }
    return val;
}

static VALUE
env_keys(void)
{
    char **env;
    VALUE ary;

    ary = rb_ary_new();
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s) {
	    rb_ary_push(ary, env_str_new(*env, s-*env));
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return ary;
}

/*
 * call-seq:
 *   ENV.keys -> Array
 *
 * Returns every environment variable name in an Array
 */

static VALUE
env_f_keys(VALUE _)
{
    return env_keys();
}

static VALUE
rb_env_size(VALUE ehash, VALUE args, VALUE eobj)
{
    char **env;
    long cnt = 0;

    env = GET_ENVIRON(environ);
    for (; *env ; ++env) {
	if (strchr(*env, '=')) {
	    cnt++;
	}
    }
    FREE_ENVIRON(environ);
    return LONG2FIX(cnt);
}

/*
 * call-seq:
 *   ENV.each_key { |name| block } -> ENV
 *   ENV.each_key                  -> Enumerator
 *
 * Yields each environment variable name.
 *
 * An Enumerator is returned if no block is given.
 */
static VALUE
env_each_key(VALUE ehash)
{
    VALUE keys;
    long i;

    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    keys = env_keys();
    for (i=0; i<RARRAY_LEN(keys); i++) {
	rb_yield(RARRAY_AREF(keys, i));
    }
    return ehash;
}

static VALUE
env_values(void)
{
    VALUE ary;
    char **env;

    ary = rb_ary_new();
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s) {
	    rb_ary_push(ary, env_str_new2(s+1));
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return ary;
}

/*
 * call-seq:
 *   ENV.values -> Array
 *
 * Returns every environment variable value as an Array
 */

static VALUE
env_f_values(VALUE _)
{
    return env_values();
}

/*
 * call-seq:
 *   ENV.each_value { |value| block } -> ENV
 *   ENV.each_value                   -> Enumerator
 *
 * Yields each environment variable +value+.
 *
 * An Enumerator is returned if no block was given.
 */
static VALUE
env_each_value(VALUE ehash)
{
    VALUE values;
    long i;

    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    values = env_values();
    for (i=0; i<RARRAY_LEN(values); i++) {
	rb_yield(RARRAY_AREF(values, i));
    }
    return ehash;
}

/*
 * call-seq:
 *   ENV.each      { |name, value| block } -> ENV
 *   ENV.each                              -> Enumerator
 *   ENV.each_pair { |name, value| block } -> ENV
 *   ENV.each_pair                         -> Enumerator
 *
 * Yields each environment variable +name+ and +value+.
 *
 * If no block is given an Enumerator is returned.
 */
static VALUE
env_each_pair(VALUE ehash)
{
    char **env;
    VALUE ary;
    long i;

    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);

    ary = rb_ary_new();
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s) {
	    rb_ary_push(ary, env_str_new(*env, s-*env));
	    rb_ary_push(ary, env_str_new2(s+1));
	}
	env++;
    }
    FREE_ENVIRON(environ);

    if (rb_block_arity() > 1) {
	for (i=0; i<RARRAY_LEN(ary); i+=2) {
	    rb_yield_values(2, RARRAY_AREF(ary, i), RARRAY_AREF(ary, i+1));
	}
    }
    else {
	for (i=0; i<RARRAY_LEN(ary); i+=2) {
	    rb_yield(rb_assoc_new(RARRAY_AREF(ary, i), RARRAY_AREF(ary, i+1)));
	}
    }
    return ehash;
}

/*
 * call-seq:
 *   ENV.reject! { |name, value| block } -> ENV or nil
 *   ENV.reject!                         -> Enumerator
 *
 * Equivalent to ENV.delete_if but returns +nil+ if no changes were made.
 *
 * Returns an Enumerator if no block was given.
 */
static VALUE
env_reject_bang(VALUE ehash)
{
    VALUE keys;
    long i;
    int del = 0;

    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    keys = env_keys();
    RBASIC_CLEAR_CLASS(keys);
    for (i=0; i<RARRAY_LEN(keys); i++) {
	VALUE val = rb_f_getenv(Qnil, RARRAY_AREF(keys, i));
	if (!NIL_P(val)) {
	    if (RTEST(rb_yield_values(2, RARRAY_AREF(keys, i), val))) {
                env_delete(RARRAY_AREF(keys, i));
		del++;
	    }
	}
    }
    RB_GC_GUARD(keys);
    if (del == 0) return Qnil;
    return envtbl;
}

/*
 * call-seq:
 *   ENV.delete_if { |name, value| block } -> ENV
 *   ENV.delete_if                         -> Enumerator
 *
 * Deletes every environment variable for which the block evaluates to +true+.
 *
 * If no block is given an enumerator is returned instead.
 */
static VALUE
env_delete_if(VALUE ehash)
{
    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    env_reject_bang(ehash);
    return envtbl;
}

/*
 * call-seq:
 *   ENV.values_at(name, ...) -> Array
 *
 * Returns an array containing the environment variable values associated with
 * the given names.  See also ENV.select.
 */
static VALUE
env_values_at(int argc, VALUE *argv, VALUE _)
{
    VALUE result;
    long i;

    result = rb_ary_new();
    for (i=0; i<argc; i++) {
	rb_ary_push(result, rb_f_getenv(Qnil, argv[i]));
    }
    return result;
}

/*
 * call-seq:
 *   ENV.select { |name, value| block } -> Hash
 *   ENV.select                         -> Enumerator
 *   ENV.filter { |name, value| block } -> Hash
 *   ENV.filter                         -> Enumerator
 *
 * Returns a copy of the environment for entries where the block returns true.
 *
 * Returns an Enumerator if no block was given.
 *
 * ENV.filter is an alias for ENV.select.
 */
static VALUE
env_select(VALUE ehash)
{
    VALUE result;
    VALUE keys;
    long i;

    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    result = rb_hash_new();
    keys = env_keys();
    for (i = 0; i < RARRAY_LEN(keys); ++i) {
	VALUE key = RARRAY_AREF(keys, i);
	VALUE val = rb_f_getenv(Qnil, key);
	if (!NIL_P(val)) {
	    if (RTEST(rb_yield_values(2, key, val))) {
		rb_hash_aset(result, key, val);
	    }
	}
    }
    RB_GC_GUARD(keys);

    return result;
}

/*
 * call-seq:
 *   ENV.select! { |name, value| block } -> ENV or nil
 *   ENV.select!                         -> Enumerator
 *   ENV.filter! { |name, value| block } -> ENV or nil
 *   ENV.filter!                         -> Enumerator
 *
 * Equivalent to ENV.keep_if but returns +nil+ if no changes were made.
 *
 * ENV.filter! is an alias for ENV.select!.
 */
static VALUE
env_select_bang(VALUE ehash)
{
    VALUE keys;
    long i;
    int del = 0;

    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    keys = env_keys();
    RBASIC_CLEAR_CLASS(keys);
    for (i=0; i<RARRAY_LEN(keys); i++) {
	VALUE val = rb_f_getenv(Qnil, RARRAY_AREF(keys, i));
	if (!NIL_P(val)) {
	    if (!RTEST(rb_yield_values(2, RARRAY_AREF(keys, i), val))) {
                env_delete(RARRAY_AREF(keys, i));
		del++;
	    }
	}
    }
    RB_GC_GUARD(keys);
    if (del == 0) return Qnil;
    return envtbl;
}

/*
 * call-seq:
 *   ENV.keep_if { |name, value| block } -> ENV
 *   ENV.keep_if                         -> Enumerator
 *
 * Deletes every environment variable where the block evaluates to +false+.
 *
 * Returns an enumerator if no block was given.
 */
static VALUE
env_keep_if(VALUE ehash)
{
    RETURN_SIZED_ENUMERATOR(ehash, 0, 0, rb_env_size);
    env_select_bang(ehash);
    return envtbl;
}

/*
 *  call-seq:
 *     ENV.slice(*keys) -> a_hash
 *
 *  Returns a hash containing only the given keys from ENV and their values.
 *
 *     ENV.slice("TERM","HOME")  #=> {"TERM"=>"xterm-256color", "HOME"=>"/Users/rhc"}
 */
static VALUE
env_slice(int argc, VALUE *argv, VALUE _)
{
    int i;
    VALUE key, value, result;

    if (argc == 0) {
        return rb_hash_new();
    }
    result = rb_hash_new_with_size(argc);

    for (i = 0; i < argc; i++) {
        key = argv[i];
        value = rb_f_getenv(Qnil, key);
        if (value != Qnil)
            rb_hash_aset(result, key, value);
    }

    return result;
}

VALUE
rb_env_clear(void)
{
    VALUE keys;
    long i;

    keys = env_keys();
    for (i=0; i<RARRAY_LEN(keys); i++) {
	VALUE val = rb_f_getenv(Qnil, RARRAY_AREF(keys, i));
	if (!NIL_P(val)) {
            env_delete(RARRAY_AREF(keys, i));
	}
    }
    RB_GC_GUARD(keys);
    return envtbl;
}

/*
 * call-seq:
 *   ENV.clear
 *
 * Removes every environment variable.
 */
static VALUE
env_clear(VALUE _)
{
    return rb_env_clear();
}

/*
 * call-seq:
 *   ENV.to_s -> "ENV"
 *
 * Returns "ENV"
 */
static VALUE
env_to_s(VALUE _)
{
    return rb_usascii_str_new2("ENV");
}

/*
 * call-seq:
 *   ENV.inspect -> string
 *
 * Returns the contents of the environment as a String.
 */
static VALUE
env_inspect(VALUE _)
{
    char **env;
    VALUE str, i;

    str = rb_str_buf_new2("{");
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');

	if (env != environ) {
	    rb_str_buf_cat2(str, ", ");
	}
	if (s) {
	    rb_str_buf_cat2(str, "\"");
	    rb_str_buf_cat(str, *env, s-*env);
	    rb_str_buf_cat2(str, "\"=>");
	    i = rb_inspect(rb_str_new2(s+1));
	    rb_str_buf_append(str, i);
	}
	env++;
    }
    FREE_ENVIRON(environ);
    rb_str_buf_cat2(str, "}");

    return str;
}

/*
 * call-seq:
 *   ENV.to_a -> Array
 *
 * Converts the environment variables into an array of names and value arrays.
 *
 *   ENV.to_a # => [["TERM", "xterm-color"], ["SHELL", "/bin/bash"], ...]
 *
 */
static VALUE
env_to_a(VALUE _)
{
    char **env;
    VALUE ary;

    ary = rb_ary_new();
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s) {
	    rb_ary_push(ary, rb_assoc_new(env_str_new(*env, s-*env),
					  env_str_new2(s+1)));
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return ary;
}

/*
 * call-seq:
 *   ENV.rehash
 *
 * Re-hashing the environment variables does nothing.  It is provided for
 * compatibility with Hash.
 */
static VALUE
env_none(VALUE _)
{
    return Qnil;
}

/*
 * call-seq:
 *   ENV.length
 *   ENV.size
 *
 * Returns the number of environment variables.
 */
static VALUE
env_size(VALUE _)
{
    int i;
    char **env;

    env = GET_ENVIRON(environ);
    for (i=0; env[i]; i++)
	;
    FREE_ENVIRON(environ);
    return INT2FIX(i);
}

/*
 * call-seq:
 *   ENV.empty? -> true or false
 *
 * Returns true when there are no environment variables
 */
static VALUE
env_empty_p(VALUE _)
{
    char **env;

    env = GET_ENVIRON(environ);
    if (env[0] == 0) {
	FREE_ENVIRON(environ);
	return Qtrue;
    }
    FREE_ENVIRON(environ);
    return Qfalse;
}

/*
 * call-seq:
 *   ENV.key?(name)     -> true or false
 *   ENV.include?(name) -> true or false
 *   ENV.has_key?(name) -> true or false
 *   ENV.member?(name)  -> true or false
 *
 * Returns +true+ if there is an environment variable with the given +name+.
 */
static VALUE
env_has_key(VALUE env, VALUE key)
{
    const char *s;

    s = env_name(key);
    if (getenv(s)) return Qtrue;
    return Qfalse;
}

/*
 * call-seq:
 *   ENV.assoc(name) -> Array or nil
 *
 * Returns an Array of the name and value of the environment variable with
 * +name+ or +nil+ if the name cannot be found.
 */
static VALUE
env_assoc(VALUE env, VALUE key)
{
    const char *s, *e;

    s = env_name(key);
    e = getenv(s);
    if (e) return rb_assoc_new(key, env_str_new2(e));
    return Qnil;
}

/*
 * call-seq:
 *   ENV.value?(value) -> true or false
 *   ENV.has_value?(value) -> true or false
 *
 * Returns +true+ if there is an environment variable with the given +value+.
 */
static VALUE
env_has_value(VALUE dmy, VALUE obj)
{
    char **env;

    obj = rb_check_string_type(obj);
    if (NIL_P(obj)) return Qnil;
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s++) {
	    long len = strlen(s);
	    if (RSTRING_LEN(obj) == len && strncmp(s, RSTRING_PTR(obj), len) == 0) {
		FREE_ENVIRON(environ);
		return Qtrue;
	    }
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return Qfalse;
}

/*
 * call-seq:
 *   ENV.rassoc(value)
 *
 * Returns an Array of the name and value of the environment variable with
 * +value+ or +nil+ if the value cannot be found.
 */
static VALUE
env_rassoc(VALUE dmy, VALUE obj)
{
    char **env;

    obj = rb_check_string_type(obj);
    if (NIL_P(obj)) return Qnil;
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s++) {
	    long len = strlen(s);
	    if (RSTRING_LEN(obj) == len && strncmp(s, RSTRING_PTR(obj), len) == 0) {
                VALUE result = rb_assoc_new(rb_str_new(*env, s-*env-1), obj);
		FREE_ENVIRON(environ);
		return result;
	    }
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return Qnil;
}

/*
 * call-seq:
 *   ENV.key(value) -> name
 *
 * Returns the name of the environment variable with +value+.  If the value is
 * not found +nil+ is returned.
 */
static VALUE
env_key(VALUE dmy, VALUE value)
{
    char **env;
    VALUE str;

    SafeStringValue(value);
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s++) {
	    long len = strlen(s);
	    if (RSTRING_LEN(value) == len && strncmp(s, RSTRING_PTR(value), len) == 0) {
		str = env_str_new(*env, s-*env-1);
		FREE_ENVIRON(environ);
		return str;
	    }
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return Qnil;
}

/*
 * call-seq:
 *   ENV.index(value) -> key
 *
 * Deprecated method that is equivalent to ENV.key
 */
static VALUE
env_index(VALUE dmy, VALUE value)
{
    rb_warn("ENV.index is deprecated; use ENV.key");
    return env_key(dmy, value);
}

static VALUE
env_to_hash(void)
{
    char **env;
    VALUE hash;

    hash = rb_hash_new();
    env = GET_ENVIRON(environ);
    while (*env) {
	char *s = strchr(*env, '=');
	if (s) {
	    rb_hash_aset(hash, env_str_new(*env, s-*env),
			       env_str_new2(s+1));
	}
	env++;
    }
    FREE_ENVIRON(environ);
    return hash;
}

/*
 * call-seq:
 *   ENV.to_hash -> hash
 *
 * Creates a hash with a copy of the environment variables.
 *
 */

static VALUE
env_f_to_hash(VALUE _)
{
    return env_to_hash();
}

/*
 * call-seq:
 *   ENV.to_h                        -> hash
 *   ENV.to_h {|name, value| block } -> hash
 *
 * Creates a hash with a copy of the environment variables.
 *
 */
static VALUE
env_to_h(VALUE _)
{
    VALUE hash = env_to_hash();
    if (rb_block_given_p()) {
        hash = rb_hash_to_h_block(hash);
    }
    return hash;
}

/*
 * call-seq:
 *   ENV.reject { |name, value| block } -> Hash
 *   ENV.reject                         -> Enumerator
 *
 * Same as ENV.delete_if, but works on (and returns) a copy of the
 * environment.
 */
static VALUE
env_reject(VALUE _)
{
    return rb_hash_delete_if(env_to_hash());
}

/*
 * call-seq:
 *   ENV.freeze -> raises TypeError
 *
 * Ruby does not allow ENV to be frozen, so calling ENV.freeze
 * raises TypeError.
 */
static VALUE
env_freeze(VALUE self)
{
    rb_raise(rb_eTypeError, "cannot freeze ENV");
    return self; /* Not reached */
}

/*
 * call-seq:
 *   ENV.shift -> Array or nil
 *
 * Removes an environment variable name-value pair from ENV and returns it as
 * an Array.  Returns +nil+ if when the environment is empty.
 */
static VALUE
env_shift(VALUE _)
{
    char **env;
    VALUE result = Qnil;

    env = GET_ENVIRON(environ);
    if (*env) {
	char *s = strchr(*env, '=');
	if (s) {
	    VALUE key = env_str_new(*env, s-*env);
	    VALUE val = env_str_new2(getenv(RSTRING_PTR(key)));
            env_delete(key);
	    result = rb_assoc_new(key, val);
	}
    }
    FREE_ENVIRON(environ);
    return result;
}

/*
 * call-seq:
 *   ENV.invert -> Hash
 *
 * Returns a new hash created by using environment variable names as values
 * and values as names.
 */
static VALUE
env_invert(VALUE _)
{
    return rb_hash_invert(env_to_hash());
}

static int
env_replace_i(VALUE key, VALUE val, VALUE keys)
{
    env_aset(key, val);
    if (rb_ary_includes(keys, key)) {
	rb_ary_delete(keys, key);
    }
    return ST_CONTINUE;
}

/*
 * call-seq:
 *   ENV.replace(hash) -> env
 *
 * Replaces the contents of the environment variables with the contents of
 * +hash+.
 */
static VALUE
env_replace(VALUE env, VALUE hash)
{
    VALUE keys;
    long i;

    keys = env_keys();
    if (env == hash) return env;
    hash = to_hash(hash);
    rb_hash_foreach(hash, env_replace_i, keys);

    for (i=0; i<RARRAY_LEN(keys); i++) {
        env_delete(RARRAY_AREF(keys, i));
    }
    RB_GC_GUARD(keys);
    return env;
}

static int
env_update_i(VALUE key, VALUE val, VALUE _)
{
    env_aset(key, val);
    return ST_CONTINUE;
}

static int
env_update_block_i(VALUE key, VALUE val, VALUE _)
{
    VALUE oldval = rb_f_getenv(Qnil, key);
    if (!NIL_P(oldval)) {
	val = rb_yield_values(3, key, oldval, val);
    }
    env_aset(key, val);
    return ST_CONTINUE;
}

/*
 * call-seq:
 *   ENV.update(hash)                                        -> ENV
 *   ENV.update(hash) { |name, old_value, new_value| block } -> ENV
 *   ENV.merge!(hash)                                        -> ENV
 *   ENV.merge!(hash) { |name, old_value, new_value| block } -> ENV
 *
 * Adds the contents of +hash+ to the environment variables.  If no block is
 * specified entries with duplicate keys are overwritten, otherwise the value
 * of each duplicate name is determined by calling the block with the key, its
 * value from the environment and its value from the hash.
 */
static VALUE
env_update(VALUE env, VALUE hash)
{
    if (env == hash) return env;
    hash = to_hash(hash);
    rb_foreach_func *func = rb_block_given_p() ?
        env_update_block_i : env_update_i;
    rb_hash_foreach(hash, func, 0);
    return env;
}

/*
 *  A Hash is a dictionary-like collection of unique keys and their values.
 *  Also called associative arrays, they are similar to Arrays, but where an
 *  Array uses integers as its index, a Hash allows you to use any object
 *  type.
 *
 *  Hashes enumerate their values in the order that the corresponding keys
 *  were inserted.
 *
 *  A Hash can be easily created by using its implicit form:
 *
 *    grades = { "Jane Doe" => 10, "Jim Doe" => 6 }
 *
 *  Hashes allow an alternate syntax for keys that are symbols.
 *  Instead of
 *
 *    options = { :font_size => 10, :font_family => "Arial" }
 *
 *  You could write it as:
 *
 *    options = { font_size: 10, font_family: "Arial" }
 *
 *  Each named key is a symbol you can access in hash:
 *
 *    options[:font_size]  # => 10
 *
 *  A Hash can also be created through its ::new method:
 *
 *    grades = Hash.new
 *    grades["Dorothy Doe"] = 9
 *
 *  Hashes have a <em>default value</em> that is returned when accessing
 *  keys that do not exist in the hash. If no default is set +nil+ is used.
 *  You can set the default value by sending it as an argument to Hash.new:
 *
 *    grades = Hash.new(0)
 *
 *  Or by using the #default= method:
 *
 *    grades = {"Timmy Doe" => 8}
 *    grades.default = 0
 *
 *  Accessing a value in a Hash requires using its key:
 *
 *    puts grades["Jane Doe"] # => 0
 *
 *  === Common Uses
 *
 *  Hashes are an easy way to represent data structures, such as
 *
 *    books         = {}
 *    books[:matz]  = "The Ruby Programming Language"
 *    books[:black] = "The Well-Grounded Rubyist"
 *
 *  Hashes are also commonly used as a way to have named parameters in
 *  functions. Note that no brackets are used below. If a hash is the last
 *  argument on a method call, no braces are needed, thus creating a really
 *  clean interface:
 *
 *    Person.create(name: "John Doe", age: 27)
 *
 *    def self.create(params)
 *      @name = params[:name]
 *      @age  = params[:age]
 *    end
 *
 *  === Hash Keys
 *
 *  Two objects refer to the same hash key when their <code>hash</code> value
 *  is identical and the two objects are <code>eql?</code> to each other.
 *
 *  A user-defined class may be used as a hash key if the <code>hash</code>
 *  and <code>eql?</code> methods are overridden to provide meaningful
 *  behavior.  By default, separate instances refer to separate hash keys.
 *
 *  A typical implementation of <code>hash</code> is based on the
 *  object's data while <code>eql?</code> is usually aliased to the overridden
 *  <code>==</code> method:
 *
 *    class Book
 *      attr_reader :author, :title
 *
 *      def initialize(author, title)
 *        @author = author
 *        @title = title
 *      end
 *
 *      def ==(other)
 *        self.class === other and
 *          other.author == @author and
 *          other.title == @title
 *      end
 *
 *      alias eql? ==
 *
 *      def hash
 *        @author.hash ^ @title.hash # XOR
 *      end
 *    end
 *
 *    book1 = Book.new 'matz', 'Ruby in a Nutshell'
 *    book2 = Book.new 'matz', 'Ruby in a Nutshell'
 *
 *    reviews = {}
 *
 *    reviews[book1] = 'Great reference!'
 *    reviews[book2] = 'Nice and compact!'
 *
 *    reviews.length #=> 1
 *
 *  See also Object#hash and Object#eql?
 */

void
Init_Hash(void)
{
#undef rb_intern
#define rb_intern(str) rb_intern_const(str)
    id_hash = rb_intern("hash");
    id_yield = rb_intern("yield");
    id_default = rb_intern("default");
    id_flatten_bang = rb_intern("flatten!");
    id_hash_iter_lev = rb_make_internal_id();

    rb_cHash = rb_define_class("Hash", rb_cObject);

    rb_include_module(rb_cHash, rb_mEnumerable);

    rb_define_alloc_func(rb_cHash, empty_hash_alloc);
    rb_define_singleton_method(rb_cHash, "[]", rb_hash_s_create, -1);
    rb_define_singleton_method(rb_cHash, "try_convert", rb_hash_s_try_convert, 1);
    rb_define_method(rb_cHash, "initialize", rb_hash_initialize, -1);
    rb_define_method(rb_cHash, "initialize_copy", rb_hash_replace, 1);
    rb_define_method(rb_cHash, "rehash", rb_hash_rehash, 0);

    rb_define_method(rb_cHash, "to_hash", rb_hash_to_hash, 0);
    rb_define_method(rb_cHash, "to_h", rb_hash_to_h, 0);
    rb_define_method(rb_cHash, "to_a", rb_hash_to_a, 0);
    rb_define_method(rb_cHash, "inspect", rb_hash_inspect, 0);
    rb_define_alias(rb_cHash, "to_s", "inspect");
    rb_define_method(rb_cHash, "to_proc", rb_hash_to_proc, 0);

    rb_define_method(rb_cHash, "==", rb_hash_equal, 1);
    rb_define_method(rb_cHash, "[]", rb_hash_aref, 1);
    rb_define_method(rb_cHash, "hash", rb_hash_hash, 0);
    rb_define_method(rb_cHash, "eql?", rb_hash_eql, 1);
    rb_define_method(rb_cHash, "fetch", rb_hash_fetch_m, -1);
    rb_define_method(rb_cHash, "[]=", rb_hash_aset, 2);
    rb_define_method(rb_cHash, "store", rb_hash_aset, 2);
    rb_define_method(rb_cHash, "default", rb_hash_default, -1);
    rb_define_method(rb_cHash, "default=", rb_hash_set_default, 1);
    rb_define_method(rb_cHash, "default_proc", rb_hash_default_proc, 0);
    rb_define_method(rb_cHash, "default_proc=", rb_hash_set_default_proc, 1);
    rb_define_method(rb_cHash, "key", rb_hash_key, 1);
    rb_define_method(rb_cHash, "index", rb_hash_index, 1);
    rb_define_method(rb_cHash, "size", rb_hash_size, 0);
    rb_define_method(rb_cHash, "length", rb_hash_size, 0);
    rb_define_method(rb_cHash, "empty?", rb_hash_empty_p, 0);

    rb_define_method(rb_cHash, "each_value", rb_hash_each_value, 0);
    rb_define_method(rb_cHash, "each_key", rb_hash_each_key, 0);
    rb_define_method(rb_cHash, "each_pair", rb_hash_each_pair, 0);
    rb_define_method(rb_cHash, "each", rb_hash_each_pair, 0);

    rb_define_method(rb_cHash, "transform_keys", rb_hash_transform_keys, 0);
    rb_define_method(rb_cHash, "transform_keys!", rb_hash_transform_keys_bang, 0);
    rb_define_method(rb_cHash, "transform_values", rb_hash_transform_values, 0);
    rb_define_method(rb_cHash, "transform_values!", rb_hash_transform_values_bang, 0);

    rb_define_method(rb_cHash, "keys", rb_hash_keys, 0);
    rb_define_method(rb_cHash, "values", rb_hash_values, 0);
    rb_define_method(rb_cHash, "values_at", rb_hash_values_at, -1);
    rb_define_method(rb_cHash, "fetch_values", rb_hash_fetch_values, -1);

    rb_define_method(rb_cHash, "shift", rb_hash_shift, 0);
    rb_define_method(rb_cHash, "delete", rb_hash_delete_m, 1);
    rb_define_method(rb_cHash, "delete_if", rb_hash_delete_if, 0);
    rb_define_method(rb_cHash, "keep_if", rb_hash_keep_if, 0);
    rb_define_method(rb_cHash, "select", rb_hash_select, 0);
    rb_define_method(rb_cHash, "select!", rb_hash_select_bang, 0);
    rb_define_method(rb_cHash, "filter", rb_hash_select, 0);
    rb_define_method(rb_cHash, "filter!", rb_hash_select_bang, 0);
    rb_define_method(rb_cHash, "reject", rb_hash_reject, 0);
    rb_define_method(rb_cHash, "reject!", rb_hash_reject_bang, 0);
    rb_define_method(rb_cHash, "slice", rb_hash_slice, -1);
    rb_define_method(rb_cHash, "clear", rb_hash_clear, 0);
    rb_define_method(rb_cHash, "invert", rb_hash_invert, 0);
    rb_define_method(rb_cHash, "update", rb_hash_update, -1);
    rb_define_method(rb_cHash, "replace", rb_hash_replace, 1);
    rb_define_method(rb_cHash, "merge!", rb_hash_update, -1);
    rb_define_method(rb_cHash, "merge", rb_hash_merge, -1);
    rb_define_method(rb_cHash, "assoc", rb_hash_assoc, 1);
    rb_define_method(rb_cHash, "rassoc", rb_hash_rassoc, 1);
    rb_define_method(rb_cHash, "flatten", rb_hash_flatten, -1);
    rb_define_method(rb_cHash, "compact", rb_hash_compact, 0);
    rb_define_method(rb_cHash, "compact!", rb_hash_compact_bang, 0);

    rb_define_method(rb_cHash, "include?", rb_hash_has_key, 1);
    rb_define_method(rb_cHash, "member?", rb_hash_has_key, 1);
    rb_define_method(rb_cHash, "has_key?", rb_hash_has_key, 1);
    rb_define_method(rb_cHash, "has_value?", rb_hash_has_value, 1);
    rb_define_method(rb_cHash, "key?", rb_hash_has_key, 1);
    rb_define_method(rb_cHash, "value?", rb_hash_has_value, 1);

    rb_define_method(rb_cHash, "compare_by_identity", rb_hash_compare_by_id, 0);
    rb_define_method(rb_cHash, "compare_by_identity?", rb_hash_compare_by_id_p, 0);

    rb_define_method(rb_cHash, "any?", rb_hash_any_p, -1);
    rb_define_method(rb_cHash, "dig", rb_hash_dig, -1);

    rb_define_method(rb_cHash, "<=", rb_hash_le, 1);
    rb_define_method(rb_cHash, "<", rb_hash_lt, 1);
    rb_define_method(rb_cHash, ">=", rb_hash_ge, 1);
    rb_define_method(rb_cHash, ">", rb_hash_gt, 1);

    rb_define_method(rb_cHash, "deconstruct_keys", rb_hash_deconstruct_keys, 1);

    /* Document-class: ENV
     *
     * ENV is a hash-like accessor for environment variables.
     *
     * === Interaction with the Operating System
     * The ENV object interacts with the operating system's environment variables:
     * - When you get the value for a name in ENV, the value is retrieved from among the current environment variables.
     * - When you create or set a name-value pair in ENV, the name and value are immediately set in the environment variables.
     * - When you delete a name-value pair in ENV, it is immediately deleted from the environment variables.
     *
     * === Names and Values
     * Generally speaking, each name or value is a String.
     *
     * Strictly speaking:
     * - Each name or value must be one of the following:
     *   - A String.
     *   - An object that responds to #to_str by returning a String, which will be used as the name or value.
     * - A name may not:
     *   - Be the empty string.
     *   - Contain character <code>=</code> or the NUL character (<code>"\0"</code>).
     *   - Have an ASCII-incompatible encoding (e.g., UTF-16LE, ISO-2022-JP).
     *
     * === About Ordering
     * ENV presents its content in the order found
     * in the operating system's environment variables.
     * Therefore the ordering of ENV content is OS-dependent, and may be indeterminate.
     *
     * This will be seen in:
     * - A Hash returned by an ENV method.
     * - An Enumerator returned by an ENV method.
     * - An Array returned by ENV.keys, ENV.values, or ENV.to_a.
     * - The String returned by ENV.inspect.
     */

    /*
     * Hack to get RDoc to regard ENV as a class:
     * envtbl = rb_define_class("ENV", rb_cObject);
     */
    origenviron = environ;
    envtbl = rb_obj_alloc(rb_cObject);
    rb_extend_object(envtbl, rb_mEnumerable);

    rb_define_singleton_method(envtbl, "[]", rb_f_getenv, 1);
    rb_define_singleton_method(envtbl, "fetch", env_fetch, -1);
    rb_define_singleton_method(envtbl, "[]=", env_aset_m, 2);
    rb_define_singleton_method(envtbl, "store", env_aset_m, 2);
    rb_define_singleton_method(envtbl, "each", env_each_pair, 0);
    rb_define_singleton_method(envtbl, "each_pair", env_each_pair, 0);
    rb_define_singleton_method(envtbl, "each_key", env_each_key, 0);
    rb_define_singleton_method(envtbl, "each_value", env_each_value, 0);
    rb_define_singleton_method(envtbl, "delete", env_delete_m, 1);
    rb_define_singleton_method(envtbl, "delete_if", env_delete_if, 0);
    rb_define_singleton_method(envtbl, "keep_if", env_keep_if, 0);
    rb_define_singleton_method(envtbl, "slice", env_slice, -1);
    rb_define_singleton_method(envtbl, "clear", env_clear, 0);
    rb_define_singleton_method(envtbl, "reject", env_reject, 0);
    rb_define_singleton_method(envtbl, "reject!", env_reject_bang, 0);
    rb_define_singleton_method(envtbl, "select", env_select, 0);
    rb_define_singleton_method(envtbl, "select!", env_select_bang, 0);
    rb_define_singleton_method(envtbl, "filter", env_select, 0);
    rb_define_singleton_method(envtbl, "filter!", env_select_bang, 0);
    rb_define_singleton_method(envtbl, "shift", env_shift, 0);
    rb_define_singleton_method(envtbl, "freeze", env_freeze, 0);
    rb_define_singleton_method(envtbl, "invert", env_invert, 0);
    rb_define_singleton_method(envtbl, "replace", env_replace, 1);
    rb_define_singleton_method(envtbl, "update", env_update, 1);
    rb_define_singleton_method(envtbl, "merge!", env_update, 1);
    rb_define_singleton_method(envtbl, "inspect", env_inspect, 0);
    rb_define_singleton_method(envtbl, "rehash", env_none, 0);
    rb_define_singleton_method(envtbl, "to_a", env_to_a, 0);
    rb_define_singleton_method(envtbl, "to_s", env_to_s, 0);
    rb_define_singleton_method(envtbl, "key", env_key, 1);
    rb_define_singleton_method(envtbl, "index", env_index, 1);
    rb_define_singleton_method(envtbl, "size", env_size, 0);
    rb_define_singleton_method(envtbl, "length", env_size, 0);
    rb_define_singleton_method(envtbl, "empty?", env_empty_p, 0);
    rb_define_singleton_method(envtbl, "keys", env_f_keys, 0);
    rb_define_singleton_method(envtbl, "values", env_f_values, 0);
    rb_define_singleton_method(envtbl, "values_at", env_values_at, -1);
    rb_define_singleton_method(envtbl, "include?", env_has_key, 1);
    rb_define_singleton_method(envtbl, "member?", env_has_key, 1);
    rb_define_singleton_method(envtbl, "has_key?", env_has_key, 1);
    rb_define_singleton_method(envtbl, "has_value?", env_has_value, 1);
    rb_define_singleton_method(envtbl, "key?", env_has_key, 1);
    rb_define_singleton_method(envtbl, "value?", env_has_value, 1);
    rb_define_singleton_method(envtbl, "to_hash", env_f_to_hash, 0);
    rb_define_singleton_method(envtbl, "to_h", env_to_h, 0);
    rb_define_singleton_method(envtbl, "assoc", env_assoc, 1);
    rb_define_singleton_method(envtbl, "rassoc", env_rassoc, 1);

    /*
     * ENV is a Hash-like accessor for environment variables.
     *
     * See ENV (the class) for more details.
     */
    rb_define_global_const("ENV", envtbl);

    /* for callcc */
    ruby_register_rollback_func_for_ensure(hash_foreach_ensure, hash_foreach_ensure_rollback);

    HASH_ASSERT(sizeof(ar_hint_t) * RHASH_AR_TABLE_MAX_SIZE == sizeof(VALUE));
}