一聚教程网:一个值得你收藏的教程网站

热门教程

PHP的json_encode使用分析说明

时间:2022-06-24 18:02:47 编辑:袖梨 来源:一聚教程网

json的优点就不说了,

有个习惯,我在输出json的时候,喜欢用 sprintf 拼成json格式,

前两天被朋友说不标准,必须要用json_encode生成的才是标准的json格式,我当然很郁闷啦,

用了这么多年了,刚知道 这样做不标准,既然说我不标准,那上面才是标准的json格式?

 代码如下 复制代码

{a : 'abc'} {'a' : 'abc'} {a : "abc"} {"a" : "abc"}


那都知道,只有第四种才是标准的json格式。

我这么做

 代码如下 复制代码

$ret_json='{"%s":"%s"}';

echo json_encode($ret_json,"a","abc");

必然也符合标准。

既然如此,那我就要刨根问底,json_encode生成的json格式究竟有什么不同?
上代码

 代码如下 复制代码
static PHP_FUNCTION(json_encode)
{
        zval *parameter;
        smart_str buf = {0};
        long options = 0;
 
        if (zend_parse_parameters(ZEND_NUM_ARGS() TSRMLS_CC, "z|l", ¶meter, &options) == FAILURE) {
                return;
        } 
 
        JSON_G(error_code) = PHP_JSON_ERROR_NONE;
 
        php_json_encode(&buf, parameter, options TSRMLS_CC);
 
        ZVAL_STRINGL(return_value, buf.c, buf.len, 1);
 
        smart_str_free(&buf);
}

JSON_G(error_code) = PHP_JSON_ERROR_NONE;
是定义的json错误,该错误可以通过json_last_error函数获取,你用过吗?反正我没用过。
php_json_encode是主要的操作

 代码如下 复制代码

PHP_JSON_API void php_json_encode(smart_str *buf, zval *val, int options TSRMLS_DC) /* {{{ */
{
        switch (Z_TYPE_P(val))
        {
                case IS_NULL:
                        smart_str_appendl(buf, "null", 4); //输出NULL
                        break;
 
                case IS_BOOL:
                        if (Z_BVAL_P(val)) {
                                smart_str_appendl(buf, "true", 4);//输出true
                        } else {
                                smart_str_appendl(buf, "false", 5);//输出false
                        }
                        break;
 
                case IS_LONG:
                        smart_str_append_long(buf, Z_LVAL_P(val));//输出长整形的值
                        break;
 
                case IS_DOUBLE:
                        {
                                char *d = NULL;
                                int len;
                                double dbl = Z_DVAL_P(val);
 
                                if (!zend_isinf(dbl) && !zend_isnan(dbl)) {//非无穷尽
                                        len = spprintf(&d, 0, "%.*k", (int) EG(precision), dbl);
                                        smart_str_appendl(buf, d, len);
                                        efree(d);
                                } else {
                                        php_error_docref(NULL TSRMLS_CC, E_WARNING, "double %.9g does not conform to the JSON spec, encoded as 0", dbl);
                                        smart_str_appendc(buf, '0');
                                }
                       }
                        break;
 
                case IS_STRING://字符串
                        json_escape_string(buf, Z_STRVAL_P(val), Z_STRLEN_P(val), options TSRMLS_CC);
                        break;
 
                case IS_ARRAY://数组和对象
                case IS_OBJECT:
                        json_encode_array(buf, &val, options TSRMLS_CC);
                        break;
 
                default:
                        php_error_docref(NULL TSRMLS_CC, E_WARNING, "type is unsupported, encoded as null");
                        smart_str_appendl(buf, "null", 4);
                        break;
        }
 
        return;
}

很明显,根据不同的类型,会有相应的case。
最复杂的是 字符串 、数组 、对象这三种类型,数组和对象是同一种操作。
先看看字符串吧,很长,注释直接写在代码里。

 代码如下 复制代码
/options应该是5.3版本之后才支持的,由以下常量组成的二进制掩码: JSON_HEX_QUOT, JSON_HEX_TAG, JSON_HEX_AMP, JSON_HEX_APOS, JSON_NUMERIC_CHECK, JSON_PRETTY_PRINT, JSON_UNESCAPED_SLASHES, JSON_FORCE_OBJECT, JSON_UNESCAPED_UNICODE.虽然我没用过。。。
static void json_escape_string(smart_str *buf, char *s, int len, int options TSRMLS_DC) /* {{{ */
{
        int pos = 0;
        unsigned short us;
        unsigned short *utf16;
 
        if (len == 0) {//如果长度为0,则直接返回 双引号 ""
                smart_str_appendl(buf, """", 2);
                return;
        }
 
        if (options & PHP_JSON_NUMERIC_CHECK) {//检测是否为0-9的数字,如果是数字,那么就会直接把数据作为long或double类型返回。
                double d;
                int type;
                long p;
 
                if ((type = is_numeric_string(s, len, &p, &d, 0)) != 0) {
                        if (type == IS_LONG) {
                                smart_str_append_long(buf, p);
                        } else if (type == IS_DOUBLE) {
                                if (!zend_isinf(d) && !zend_isnan(d)) {
                                        char *tmp;
                                        int l = spprintf(&tmp, 0, "%.*k", (int) EG(precision), d);
                                        smart_str_appendl(buf, tmp, l);
                                        efree(tmp);
                                } else {
                                        php_error_docref(NULL TSRMLS_CC, E_WARNING, "double %.9g does not conform to the JSON spec, encoded as 0", d);
                                        smart_str_appendc(buf, '0');
                                }
                        }
                        return;
                }
 
        }
 
        utf16 = (unsigned short *) safe_emalloc(len, sizeof(unsigned short), 0);
        len = utf8_to_utf16(utf16, s, len); //这里会对你输入的值一次处理转成对应的Dec码,比如1是49,a是97这样的,保存到utf16中。
        if (len <= 0) {//如果len小于0 说明出错。如果用json_encode处理GBK的编码,就会在这里挂掉。
                if (utf16) {
                        efree(utf16);
                }
                if (len < 0) {
                        JSON_G(error_code) = PHP_JSON_ERROR_UTF8;
                        if (!PG(display_errors)) {
                                php_error_docref(NULL TSRMLS_CC, E_WARNING, "Invalid UTF-8 sequence in argument");
                        }
                        smart_str_appendl(buf, "null", 4);
                } else {
                        smart_str_appendl(buf, """", 2);
                }
                return;
        }
 
        smart_str_appendc(buf, '"'); //输入 "
 
//下面这一段代码就是将一些特殊字符转义如 双引号,反斜线等等
        while (pos < len)
        {
                us = utf16[pos++];
 
                switch (us)
                {
                        case '"':
                                if (options & PHP_JSON_HEX_QUOT) {
                                        smart_str_appendl(buf, "u0022", 6);
                                } else {
                                        smart_str_appendl(buf, """, 2);
                                }
                                break;
 
                        case '':
                                smart_str_appendl(buf, "\", 2);
                                break;
case '/':
                                smart_str_appendl(buf, "/", 2);
                                break;
 
                        case 'b':
                                smart_str_appendl(buf, "b", 2);
                                break;
 
                        case 'f':
                                smart_str_appendl(buf, "f", 2);
                                break;
 
                        case 'n':
                                smart_str_appendl(buf, "n", 2);
                                break;
 
                        case 'r':
                                smart_str_appendl(buf, "r", 2);
                                break;
 
                        case 't':
                                smart_str_appendl(buf, "t", 2);
                                break;
 
                        case '<':
                                if (options & PHP_JSON_HEX_TAG) {
                                        smart_str_appendl(buf, "u003C", 6);
                                } else {
                                        smart_str_appendc(buf, '<');
                                }
                                break;
 
                        case '>':
                                if (options & PHP_JSON_HEX_TAG) {
                                        smart_str_appendl(buf, "u003E", 6);
                                } else {
                                        smart_str_appendc(buf, '>');
}
                                break;
 
                        case '&':
                                if (options & PHP_JSON_HEX_AMP) {
                                        smart_str_appendl(buf, "u0026", 6);
                                } else {
                                        smart_str_appendc(buf, '&');
                                }
                                break;
 
                        case ''':
                                if (options & PHP_JSON_HEX_APOS) {
                                        smart_str_appendl(buf, "u0027", 6);
                                } else {
                                        smart_str_appendc(buf, ''');
                                }
                                break;
 
                        default: //一直到这里,没有特殊字符就会把值append到buf中
                                if (us >= ' ' && (us & 127) == us) {
                                        smart_str_appendc(buf, (unsigned char) us);
                                } else {
                                        smart_str_appendl(buf, "u", 2);
                                        us = REVERSE16(us);
 
                                        smart_str_appendc(buf, digits[us & ((1 << 4) - 1)]);
                                        us >>= 4;
                                        smart_str_appendc(buf, digits[us & ((1 << 4) - 1)]);
                                        us >>= 4;
                                        smart_str_appendc(buf, digits[us & ((1 << 4) - 1)]);
                                        us >>= 4;
                                        smart_str_appendc(buf, digits[us & ((1 << 4) - 1)]);
                                }
                                break;
                }
        }
        smart_str_appendc(buf, '"'); //结束 双引号。
        efree(utf16);
}

再来看看数组和对象,也很简单,

 代码如下 复制代码
static void json_encode_array(smart_str *buf, zval **val, int options TSRMLS_DC) /* {{{ */
{
        int i, r;
        HashTable *myht;
 
        if (Z_TYPE_PP(val) == IS_ARRAY) {
                myht = HASH_OF(*val);
                r = (options & PHP_JSON_FORCE_OBJECT) ? PHP_JSON_OUTPUT_OBJECT : json_determine_array_type(val TSRMLS_CC);
        } else {
                myht = Z_OBJPROP_PP(val);
                r = PHP_JSON_OUTPUT_OBJECT;
        } 
 
        if (myht && myht->nApplyCount > 1) {
                php_error_docref(NULL TSRMLS_CC, E_WARNING, "recursion detected");
                smart_str_appendl(buf, "null", 4);
                return;
        }
//开始标签
        if (r == PHP_JSON_OUTPUT_ARRAY) {
                smart_str_appendc(buf, '[');
        } else {
                smart_str_appendc(buf, '{');
        } 
 
        i = myht ? zend_hash_num_elements(myht) : 0;
 
        if (i > 0)
        {
                char *key;
                zval **data;
                ulong index;
                uint key_len;
                HashPosition pos;
                HashTable *tmp_ht;
                int need_comma = 0;
 
                zend_hash_internal_pointer_reset_ex(myht, &pos);
//便利哈希表
                for (;; zend_hash_move_forward_ex(myht, &pos)) {
                        i = zend_hash_get_current_key_ex(myht, &key, &key_len, &index, 0, &pos);
                        if (i == HASH_KEY_NON_EXISTANT)
                                break;
 
                        if (zend_hash_get_current_data_ex(myht, (void **) &data, &pos) == SUCCESS) {
                                tmp_ht = HASH_OF(*data);
                                if (tmp_ht) {
                                        tmp_ht->nApplyCount++;
                                }
 
                                if (r == PHP_JSON_OUTPUT_ARRAY) {
                                        if (need_comma) {
                                                smart_str_appendc(buf, ',');
                                        } else {
                                                need_comma = 1;
                                        }
//将值append到 buf中
                                        php_json_encode(buf, *data, options TSRMLS_CC);
                                } else if (r == PHP_JSON_OUTPUT_OBJECT) {
                                        if (i == HASH_KEY_IS_STRING) {
                                                if (key[0] == '' && Z_TYPE_PP(val) == IS_OBJECT) {
                                                        /* Skip protected and private members. */
                                                        if (tmp_ht) {
                                                                tmp_ht->nApplyCount--;
                                                        }
                                                        continue;
                                                }
 
                                                if (need_comma) {
                                                        smart_str_appendc(buf, ',');
                                                } else {
                                                        need_comma = 1;
                                                }
 
                                                json_escape_string(buf, key, key_len - 1, options & ~PHP_JSON_NUMERIC_CHECK TSRMLS_CC);
                                                smart_str_appendc(buf, ':');
 
                                                php_json_encode(buf, *data, options TSRMLS_CC);
                                        } else {
                                                if (need_comma) {
                                                        smart_str_appendc(buf, ',');
                                                } else {
                                                        need_comma = 1;
                                                }
 
                                                smart_str_appendc(buf, '"');
                                                smart_str_append_long(buf, (long) index);
                                                smart_str_appendc(buf, '"');
                                                smart_str_appendc(buf, ':');
 
                                                php_json_encode(buf, *data, options TSRMLS_CC);
                                        }
                                }
 
                                if (tmp_ht) {
                                        tmp_ht->nApplyCount--;
                                }
                        }
                }
        }
//结束标签
        if (r == PHP_JSON_OUTPUT_ARRAY) {
                smart_str_appendc(buf, ']');
        } else {
                smart_str_appendc(buf, '}');
        }
}

通过简单分析,证明了一个问题,跟我上面用sprintf的方法其实是一样的,都是拼接字符串,

而且 为了性能,更应该鼓励用sprintf来拼接json格式,

因为 json_encode会进行很多 循环操作,而且所消耗的性能是线性的 O(n^2)。

热门栏目