PHP的json_encode使用分析說明

本文章來給大家介紹json_encode使用分析，興趣了解php源碼的朋友可嘗試參考哦。

json的優點就不說了，

有個習慣，我在輸出json的時候，喜歡用 sprintf 拼成json格式，

前兩天被朋友說不標准，必須要用json_encode生成的才是標准的json格式，我當然很郁悶啦，

用了這麼多年了，剛知道 這樣做不標准，既然說我不標准，那上面才是標准的json格式？

 代碼如下 復制代碼

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

那都知道，只有第四種才是標准的json格式。

我這麼做

 代碼如下 復制代碼

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

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

必然也符合標准。

既然如此，那我就要刨根問底，json_encode生成的json格式究竟有什麼不同？
上代碼

 代碼如下 復制代碼 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", &parameter, &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;
是定義的json錯誤，該錯誤可以通過json_last_error函數獲取，你用過嗎？反正我沒用過。
php_json_encode是主要的操作

 代碼如下 復制代碼

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); //輸出NULL
                        break;
 
                case IS_BOOL:
                        if (Z_BVAL_P(val)) {
                                smart_str_appendl(buf, "true", 4);//輸出true
                        } else {
                                smart_str_appendl(buf, "false", 5);//輸出false
                        }
                        break;
 
                case IS_LONG:
                        smart_str_append_long(buf, Z_LVAL_P(val));//輸出長整形的值
                        break;
 
                case IS_DOUBLE:
                        {
                                char *d = NULL;
                                int len;
                                double dbl = Z_DVAL_P(val);
 
                                if (!zend_isinf(dbl) && !zend_isnan(dbl)) {//非無窮盡
                                        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://數組和對象
                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;
}

很明顯，根據不同的類型，會有相應的case。
最復雜的是 字符串 、數組 、對象這三種類型，數組和對象是同一種操作。
先看看字符串吧，很長，注釋直接寫在代碼裡。

 代碼如下 復制代碼 /options應該是5.3版本之後才支持的，由以下常量組成的二進制掩碼： 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.雖然我沒用過。。。
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) {//如果長度為0，則直接返回 雙引號 ""
                smart_str_appendl(buf, """", 2);
                return;
        }
 
        if (options & PHP_JSON_NUMERIC_CHECK) {//檢測是否為0-9的數字，如果是數字，那麼就會直接把數據作為long或double類型返回。
                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); //這裡會對你輸入的值一次處理轉成對應的Dec碼，比如1是49，a是97這樣的，保存到utf16中。
        if (len <= 0) {//如果len小於0 說明出錯。如果用json_encode處理GBK的編碼，就會在這裡掛掉。
                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, '"'); //輸入 "
 
//下面這一段代碼就是將一些特殊字符轉義如 雙引號，反斜線等等
        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: //一直到這裡，沒有特殊字符就會把值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, '"'); //結束 雙引號。
        efree(utf16);
}

再來看看數組和對象，也很簡單，

 代碼如下 復制代碼 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;
        }
//開始標簽
        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;
                                        }
//將值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--;
                                }
                        }
                }
        }
//結束標簽
        if (r == PHP_JSON_OUTPUT_ARRAY) {
                smart_str_appendc(buf, ']');
        } else {
                smart_str_appendc(buf, '}');
        }
}

通過簡單分析，證明了一個問題，跟我上面用sprintf的方法其實是一樣的，都是拼接字符串，

而且 為了性能，更應該鼓勵用sprintf來拼接json格式，

因為 json_encode會進行很多 循環操作，而且所消耗的性能是線性的 O(n^2)。