nginx的配置文件解析：ngx_conf_read_token函數

該函數獲取配置文件nginx.conf中的配置行或者配置塊起始處的token.將這些token放在cf->args動態數組中, 並返回這些token所在配置文件中的位置. 
例如一個配置行的話. 就返回NGX_OK. 配置行以分號結尾. 如果是一個配置塊的起始處, 即以'{'結尾, 則返回NGX_CONF_BLOCK_START.
對於token的定義爲: 連續非空白字符. 單雙引號包圍的字符. 單雙引號內允許轉義序列如\t, \r, \n.
'#'之後爲註釋. 其後行內的所有字符全部忽略. 

理解該函數的要點只需理解幾個標識的具體含義即可.
found: 爲1表明發現一個token. 然後將該token壓入cf->args數組中.
need_space: token被單雙引號包圍時, 最後一個分號後面必須是空白字符、分號、坐花括號或者右括號(左括號匹配在代碼中沒看見？). 否則出錯.
last_space: 爲1表明前面掃描的字符均爲空白字符. 即還未發現token起始位置. 爲0表明已找到token的起始位置.
sharp_comment: 掃描到'#'字符後置1, 其後所有字符都會忽略直至一行結束.
quoted: 轉義序列, '\'之後的字符被跳過.然後quoted置0.
s_quoted, d_quoted: token以單或雙引號起始, 必須以單或雙引號結束.

例如:
worker_processes  1;
解析後cf->args數組裏有兩個ngx_str_t結構. 兩個字符串分別是"worker_processes"和"1", 返回值爲NGX_OK。
event {
     worker_connections     1024;
}
解析時cf_args數組裏有一個ngx_str_t結構.字符串爲"event". 返回值爲NGX_CONF_BLOCK_START.

點擊(此處)摺疊或打開

static ngx_int_t
ngx_conf_read_token(ngx_conf_t *cf)
{
    u_char *start, ch, *src, *dst;
    off_t file_size;
    size_t len;
    ssize_t n, size;
    ngx_uint_t found, need_space, last_space, sharp_comment, variable;
    ngx_uint_t quoted, s_quoted, d_quoted, start_line;
    ngx_str_t *word;
    ngx_buf_t *b;

    found = 0;
    need_space = 0;
    last_space = 1;
    sharp_comment = 0;
    variable = 0;
    quoted = 0;
    s_quoted = 0;
    d_quoted = 0;

    cf->args->nelts = 0;
    b = cf->conf_file->buffer;
    start = b->pos;
    start_line = cf->conf_file->line;

    file_size = ngx_file_size(&cf->conf_file->file.info);

    for ( ;; ) {
//已處理完緩衝區中的字符.
        if (b->pos >= b->last) {

            if (cf->conf_file->file.offset >= file_size) {
//文件已處理完. 但是已解析出的token數大於0或者已發現一個token的
//起始位置. 配置行不完整. 配置語法出錯。
                if (cf->args->nelts > 0 || !last_space) {

                    if (cf->conf_file->file.fd == NGX_INVALID_FILE) {
                        ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                           "unexpected end of parameter, "
                                           "expecting \";\"");
                        return NGX_ERROR;
                    }

                    ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                  "unexpected end of file, "
                                  "expecting \";\"
or \"}\"");
                    return NGX_ERROR;
                }
//處理完配置文件.
                return NGX_CONF_FILE_DONE;
            }
//未處理的緩衝區字符長度.
            len = b->pos - start;
//若長度等於緩衝區長度. 則配置出錯.
//token長度不能大於NGX_CONF_BUFFER.
            if (len == NGX_CONF_BUFFER) {
                cf->conf_file->line = start_line;

                if (d_quoted) {
                    ch = '"';

                } else if (s_quoted) {
                    ch = '\'';

                } else {
                    ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                       "too long parameter \"%*s...\" started",
                                       10, start);
                    return NGX_ERROR;
                }

                ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                   "too long parameter, probably "
                                   "missing terminating \"%c\" character",
ch);
                return NGX_ERROR;
            }
//將未處理完的字符拷貝到緩衝區開始出. 這些字符是一個token的一部分
            if (len) {
                ngx_memmove(b->start, start, len);
            }
// 配置文件未處理的長度.
            size = (ssize_t) (file_size - cf->conf_file->file.offset);
//若超過可用緩衝區長度則一次讀取可用緩衝區長度的字符.
            if (size > b->end - (b->start + len)) {
                size = b->end - (b->start + len);
            }
//從上次讀取結束的地方開始讀取size個字符繼續處理.
            n = ngx_read_file(&cf->conf_file->file, b->start + len, size,
                              cf->conf_file->file.offset);

            if (n == NGX_ERROR) {
                return NGX_ERROR;
            }

            if (n != size) {
                ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                   ngx_read_file_n " returned "
                                   "only %z bytes instead of %z",
                                   n, size);
                return NGX_ERROR;
            }
//[b->pos, b->last) 之間是需要處理的字符.
            b->pos = b->start + len;
            b->last = b->pos + n;
            start = b->start;
        }

        ch = *b->pos++;
//當前字符是換行符. 行計數器增1, 若之前掃描到該行首是註釋符, 則註釋標識清
//0。
        if (ch == LF) {
            cf->conf_file->line++;

            if (sharp_comment) {
                sharp_comment = 0;
            }
        }
//當前行是註釋行, 吃掉該行後續所有字符.
        if (sharp_comment) {
            continue;
        }
//跳過\後面的字符. 即轉義序列.
        if (quoted) {
            quoted = 0;
            continue;
        }
//當token被單或雙引號包圍是need_space會被置1.
        if (need_space) {

            if (ch == ' ' || ch == '\t' || ch == CR || ch == LF) {
                last_space = 1;
                need_space = 0;
                continue;
            }


//配置行以分號結尾.
            if (ch == ';') {
                return NGX_OK;
            }
//配置塊以{開始
            if (ch == '{') {
                return NGX_CONF_BLOCK_START;
            }

            if (ch == ')') {
                last_space = 1;
                need_space = 0;

            } else {
                 ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                    "unexpected \"%c\"", ch);
                 return NGX_ERROR;
            }
        }
//last_space爲1表明未發現token起始位置, 吃掉所有空白字符.
        if (last_space) {
            if (ch == ' ' || ch == '\t' || ch == CR || ch == LF) {
                continue;
            }
//非空白字符是一個token的起始位置.
            start = b->pos - 1;
            start_line = cf->conf_file->line;

            switch (ch) {

            case ';':
            case '{':
//在分號和左括號之前必定會出現至少一個token. 否則是配置語法錯誤
                if (cf->args->nelts == 0) {
                    ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                       "unexpected \"%c\"", ch);
                    return NGX_ERROR;
                }
//配置塊起始位置.
                if (ch == '{') {
                    return NGX_CONF_BLOCK_START;
                }
//配置行結束.
                return NGX_OK;
//配置塊結束標識右括號獨佔一行.
            case '}':
                if (cf->args->nelts != 0) {
                    ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                       "unexpected \"}\"");
                    return NGX_ERROR;
                }

                return NGX_CONF_BLOCK_DONE;
//註釋符.
            case '#':
                sharp_comment = 1;
                continue;
//轉義序列.其後面的一個字符將被跳過.
            case '\\':
                quoted = 1;
                last_space = 0;
                continue;
//引號開頭的token.
            case '"':
                start++;
                d_quoted = 1;
                last_space = 0;
                continue;

            case '\'':
                start++;
                s_quoted = 1;
                last_space = 0;
                continue;

            default:
                last_space = 0;
            }
//到這裏開始處理token字符.
        } else {
            if (ch == '{' && variable) {
                continue;
            }

            variable = 0;
//轉義序列.其後面的一個字符將被跳過
            if (ch == '\\') {
                quoted = 1;
                continue;
            }
//？？？？？？這裏是什麼？？？
            if (ch == '$') {
                variable = 1;
                continue;
            }
//單雙引號包圍的字符是一個token.
            if (d_quoted) {
                if (ch == '"')
{
                    d_quoted = 0;
                    need_space = 1;
                    found = 1;
                }

            } else if (s_quoted) {
                if (ch == '\'') {
                    s_quoted = 0;
                    need_space = 1;
                    found = 1;
                }
//字符是空格、製表符、換行符、分號或是左花括號表示一個token的結束.
//也即發現一個token
            } else if (ch == ' ' || ch == '\t' || ch == CR || ch == LF
                       || ch == ';' || ch == '{')
            {
                last_space = 1;
                found = 1;
            }

            if (found) {
//將該token放入cf->args動態數組中.
                word = ngx_array_push(cf->args);
                if (word == NULL) {
                    return NGX_ERROR;
                }

                word->data = ngx_pnalloc(cf->pool, b->pos - start + 1);
                if (word->data == NULL) {
                    return NGX_ERROR;
                }
//將token中的轉義序列轉義後複製給word->data, 其他的按原樣複製.
                for (dst = word->data, src = start, len = 0;
                     src < b->pos - 1;
                     len++)
                {
                    if (*src == '\\') {
                        switch (src[1]) {
                        case '"':
                        case '\'':
                        case '\\':
                            src++;
                            break;

                        case 't':
                            *dst++ = '\t';
                            src += 2;
                            continue;

                        case 'r':
                            *dst++ = '\r';
                            src += 2;
                            continue;

                        case 'n':
                            *dst++ = '\n';
                            src += 2;
                            continue;
                        }

                    }
                    *dst++ = *src++;
                }
                *dst = '\0';
                word->len = len;
//若token以分號結束則處理完一個配置行.
                if (ch == ';') {
                    return NGX_OK;
                }
//若token以左花括號結束則發現配置塊起始位置.
                if (ch == '{') {
                    return NGX_CONF_BLOCK_START;
                }
//發現token標識清0
                found = 0;
            }
        }
    }
}