尝试从 C 中的 hash 表中获取值返回一些随机值而不是值

Question

我在 C 中使用 hash 表库，但似乎每当我在与输入键值对的位置不同的 function 中检索某个键的值时，我都会得到一些随机值而不是实际值。

如果我使用我放置的同一个 function 中的键从 hash 表中获取值，则一切正常。 但是，当我将键值对放入一个 function 的哈希表中，然后尝试在另一个 function 中检索它时，它每次都会吐出一个随机的新值而不是预期值。 我尝试了 3 个不同的库（GNU Hash 表、UTHash 和 Ben Hoyt 的实现，所有这些我都在下面链接），但它们都不能跨函数工作。 我认为这不是库的问题，而是我检索值的方式有问题。 我在下面附上了代码和 output 的截图，以及我用于 hash 表的当前代码。

ht.c：

#include "ht.h"

#include <assert.h>
#include <stdint.h>
#include <stdlib.h>
#include <string.h>

// Hash table entry (slot may be filled or empty).
typedef struct {
    const char* key;  // key is NULL if this slot is empty
    void* value;
} ht_entry;

#define INITIAL_CAPACITY 16  // must not be zero

// Hash table structure: create with ht_create, free with ht_destroy.
typedef struct ht {
    ht_entry* entries;  // hash slots
    size_t capacity;    // size of _entries array
    size_t length;      // number of items in hash table
};

ht* ht_create(void) {
    // Allocate space for hash table struct.
    ht* table = malloc(sizeof(ht));
    if (table == NULL) {
        return NULL;
    }
    table->length = 0;
    table->capacity = INITIAL_CAPACITY;

    // Allocate (zero'd) space for entry buckets.
    table->entries = calloc(table->capacity, sizeof(ht_entry));
    if (table->entries == NULL) {
        free(table); // error, free table before we return!
        return NULL;
    }
    return table;
}

void ht_destroy(ht* table) {
    // First free allocated keys.
    for (size_t i = 0; i < table->capacity; i++) {
        if (table->entries[i].key != NULL) {
            free((void*)table->entries[i].key);
        }
    }

    // Then free entries array and table itself.
    free(table->entries);
    free(table);
}

#define FNV_OFFSET 14695981039346656037UL
#define FNV_PRIME 1099511628211UL

// Return 64-bit FNV-1a hash for key (NUL-terminated). See description:
// https://en.wikipedia.org/wiki/Fowler–Noll–Vo_hash_function
static uint64_t hash_key(const char* key) {
    uint64_t hash = FNV_OFFSET;
    for (const char* p = key; *p; p++) {
        hash ^= (uint64_t)(unsigned char)(*p);
        hash *= FNV_PRIME;
    }
    return hash;
}

void* ht_get(ht* table, const char* key) {
    // AND hash with capacity-1 to ensure it's within entries array.
    uint64_t hash = hash_key(key);
    size_t index = (size_t)(hash & (uint64_t)(table->capacity - 1));

    // Loop till we find an empty entry.
    while (table->entries[index].key != NULL) {
        if (strcmp(key, table->entries[index].key) == 0) {
            // Found key, return value.
            return table->entries[index].value;
        }
        // Key wasn't in this slot, move to next (linear probing).
        index++;
        if (index >= table->capacity) {
            // At end of entries array, wrap around.
            index = 0;
        }
    }
    return NULL;
}

// Internal function to set an entry (without expanding table).
static const char* ht_set_entry(ht_entry* entries, size_t capacity,
        const char* key, void* value, size_t* plength) {
    // AND hash with capacity-1 to ensure it's within entries array.
    uint64_t hash = hash_key(key);
    size_t index = (size_t)(hash & (uint64_t)(capacity - 1));

    // Loop till we find an empty entry.
    while (entries[index].key != NULL) {
        if (strcmp(key, entries[index].key) == 0) {
            // Found key (it already exists), update value.
            entries[index].value = value;
            return entries[index].key;
        }
        // Key wasn't in this slot, move to next (linear probing).
        index++;
        if (index >= capacity) {
            // At end of entries array, wrap around.
            index = 0;
        }
    }

    // Didn't find key, allocate+copy if needed, then insert it.
    if (plength != NULL) {
        key = strdup(key);
        if (key == NULL) {
            return NULL;
        }
        (*plength)++;
    }
    entries[index].key = (char*)key;
    entries[index].value = value;
    return key;
}

// Expand hash table to twice its current size. Return true on success,
// false if out of memory.
static bool ht_expand(ht* table) {
    // Allocate new entries array.
    size_t new_capacity = table->capacity * 2;
    if (new_capacity < table->capacity) {
        return false;  // overflow (capacity would be too big)
    }
    ht_entry* new_entries = calloc(new_capacity, sizeof(ht_entry));
    if (new_entries == NULL) {
        return false;
    }

    // Iterate entries, move all non-empty ones to new table's entries.
    for (size_t i = 0; i < table->capacity; i++) {
        ht_entry entry = table->entries[i];
        if (entry.key != NULL) {
            ht_set_entry(new_entries, new_capacity, entry.key,
                         entry.value, NULL);
        }
    }

    // Free old entries array and update this table's details.
    free(table->entries);
    table->entries = new_entries;
    table->capacity = new_capacity;
    return true;
}

const char* ht_set(ht* table, const char* key, void* value) {
    assert(value != NULL);
    if (value == NULL) {
        return NULL;
    }

    // If length will exceed half of current capacity, expand it.
    if (table->length >= table->capacity / 2) {
        if (!ht_expand(table)) {
            return NULL;
        }
    }

    // Set entry and update length.
    return ht_set_entry(table->entries, table->capacity, key, value,
                        &table->length);
}

size_t ht_length(ht* table) {
    return table->length;
}

hti ht_iterator(ht* table) {
    hti it;
    it._table = table;
    it._index = 0;
    return it;
}

bool ht_next(hti* it) {
    // Loop till we've hit end of entries array.
    ht* table = it->_table;
    while (it->_index < table->capacity) {
        size_t i = it->_index;
        it->_index++;
        if (table->entries[i].key != NULL) {
            // Found next non-empty item, update iterator key and value.
            ht_entry entry = table->entries[i];
            it->key = entry.key;
            it->value = entry.value;
            return true;
        }
    }
    return false;
}

ht.h:

// Simple hash table implemented in C.

#ifndef _HT_H
#define _HT_H

#include <stdbool.h>
#include <stddef.h>
#include <stdint.h>

// Hash table structure: create with ht_create, free with ht_destroy.
typedef struct ht ht;

// Create hash table and return pointer to it, or NULL if out of memory.
ht* ht_create(void);

// Free memory allocated for hash table, including allocated keys.
void ht_destroy(ht* table);

// Get item with given key (NUL-terminated) from hash table. Return
// value (which was set with ht_set), or NULL if key not found.
void* ht_get(ht* table, const char* key);

// Set item with given key (NUL-terminated) to value (which must not
// be NULL). If not already present in table, key is copied to newly
// allocated memory (keys are freed automatically when ht_destroy is
// called). Return address of copied key, or NULL if out of memory.
const char* ht_set(ht* table, const char* key, void* value);

// Return number of items in hash table.
size_t ht_length(ht* table);

// Hash table iterator: create with ht_iterator, iterate with ht_next.
typedef struct {
    const char* key;  // current key
    void* value;      // current value

    // Don't use these fields directly.
    ht* _table;       // reference to hash table being iterated
    size_t _index;    // current index into ht._entries
} hti;

// Return new hash table iterator (for use with ht_next).
hti ht_iterator(ht* table);

// Move iterator to next item in hash table, update iterator's key
// and value to current item, and return true. If there are no more
// items, return false. Don't call ht_set during iteration.
bool ht_next(hti* it);

#endif // _HT_H

GNU C： https://www.gnu.org/software/libc/manual/html_node/Hash-Search-Function.html
UTHash: https://troydhanson.github.io/uthash/
本霍伊特： https://benhoyt.com/writings/hash-table-in-c/

感谢您的任何帮助。

Answer 1

查看您的屏幕截图，我看到：

void *data = &i;

这是将存在于堆栈中的变量的地址存储在test_hashtable function 中。当您将该指针存储在 hash 表中，然后 function 退出时，您不能依赖它指向的数据。 数据没了，memory可能有别的用途。

所以你遇到的是未定义的行为。 如果您希望存储一个 int 值，那么您需要为其分配 memory 或实际将其塞入指针（对于适合指针的任何数据类型，这是一种相当普遍的做法）。

这是将 integer 存储在实际指针中的示例：

// store
const char *key = "Gate";
int i = 10;
void *data;
memcpy(&data, &i, sizeof(i));
ht_set(hashtable, key, data);

// retrieve
void *result = ht_get(hashtable, key);
int value;
memcpy(&value, &result, sizeof(value));

但是，您会在这里注意到一个问题，即由于ht_get的调用语义使用 NULL 值表示没有数据，因此您无法将 integer 0存储在表中。 所以，除非你想改变你的接口来返回一个“found”指示符和数据，那么你唯一的选择就是实际分配：

// store
const char *key = "Gate";
int i = 10;
int *data = malloc(sizeof(i));
if (data) {
    *data = i;
    ht_set(hashtable, key, data);
}

// retrieve
void *result = ht_get(hashtable, key);
if (result) {
    int value = *(int*)result;
}

// note: you'll need to `free` the data when you remove the key from your hash table.