[nginX-1.7.2] ngx_conf_read_token函数

该函数获取配置文件nginx.conf中的配置行或者配置块起始处的token.将这些token放在cf->args动态数组中, 并返回这些token所在配置文件中的位置.
例如一个配置行的话. 就返回NGX_OK. 配置行以分号结尾. 如果是一个配置块的起始处, 即以'{'结尾, 则返回NGX_CONF_BLOCK_START.
对于token的定义为: 连续非空白字符. 单双引号包围的字符. 单双引号内允许转义序列如\t, \r, \n.
'#'之后为注释. 其后行内的所有字符全部忽略.

理解该函数的要点只需理解几个标识的具体含义即可.
found: 为1表明发现一个token. 然后将该token压入cf->args数组中.
need_space: token被单双引号包围时, 最后一个分号后面必须是空白字符、分号、坐花括号或者右括号(左括号匹配在代码中没看见？). 否则出错.
last_space: 为1表明前面扫描的字符均为空白字符. 即还未发现token起始位置. 为0表明已找到token的起始位置.
sharp_comment: 扫描到'#'字符后置1, 其后所有字符都会忽略直至一行结束.
quoted: 转义序列, '\'之后的字符被跳过.然后quoted置0.
s_quoted, d_quoted: token以单或双引号起始, 必须以单或双引号结束.

例如:
worker_processes  1;
解析后cf->args数组里有两个ngx_str_t结构. 两个字符串分别是"worker_processes"和"1", 返回值为NGX_OK。
event {
     worker_connections     1024;
}
解析时cf_args数组里有一个ngx_str_t结构.字符串为"event". 返回值为NGX_CONF_BLOCK_START.

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 ( ;; ) {
//已处理完缓冲区中的字符.
        if (b->pos >= b->last) {

            if (cf->conf_file->file.offset >= file_size) {
//文件已处理完. 但是已解析出的token数大于0或者已发现一个token的
//起始位置. 配置行不完整. 配置语法出错。
                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;
                }
//处理完配置文件.
                return NGX_CONF_FILE_DONE;
            }
//未处理的缓冲区字符长度.
            len = b->pos - start;
//若长度等于缓冲区长度. 则配置出错.
//token长度不能大于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;
            }
//将未处理完的字符拷贝到缓冲区开始出. 这些字符是一个token的一部分
            if (len) {
                ngx_memmove(b->start, start, len);
            }
// 配置文件未处理的长度.
            size = (ssize_t) (file_size - cf->conf_file->file.offset);
//若超过可用缓冲区长度则一次读取可用缓冲区长度的字符.
            if (size > b->end - (b->start + len)) {
                size = b->end - (b->start + len);
            }
//从上次读取结束的地方开始读取size个字符继续处理.
            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) 之间是需要处理的字符.
            b->pos = b->start + len;
            b->last = b->pos + n;
            start = b->start;
        }

        ch = *b->pos++;
//当前字符是换行符. 行计数器增1, 若之前扫描到该行首是注释符, 则注释标识清
//0。
        if (ch == LF) {
            cf->conf_file->line++;

            if (sharp_comment) {
                sharp_comment = 0;
            }
        }
//当前行是注释行, 吃掉该行后续所有字符.
        if (sharp_comment) {
            continue;
        }
//跳过\后面的字符. 即转义序列.
        if (quoted) {
            quoted = 0;
            continue;
        }
//当token被单或双引号包围是need_space会被置1.
        if (need_space) {

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


//配置行以分号结尾.
            if (ch == ';') {
                return NGX_OK;
            }
//配置块以{开始
            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为1表明未发现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;
            }
//非空白字符是一个token的起始位置.
            start = b->pos - 1;
            start_line = cf->conf_file->line;

            switch (ch) {

            case ';':
            case '{':
//在分号和左括号之前必定会出现至少一个token. 否则是配置语法错误
                if (cf->args->nelts == 0) {
                    ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                       "unexpected \"%c\"", ch);
                    return NGX_ERROR;
                }
//配置块起始位置.
                if (ch == '{') {
                    return NGX_CONF_BLOCK_START;
                }
//配置行结束.
                return NGX_OK;
//配置块结束标识右括号独占一行.
            case '}':
                if (cf->args->nelts != 0) {
                    ngx_conf_log_error(NGX_LOG_EMERG, cf, 0,
                                       "unexpected \"}\"");
                    return NGX_ERROR;
                }

                return NGX_CONF_BLOCK_DONE;
//注释符.
            case '#':
                sharp_comment = 1;
                continue;
//转义序列.其后面的一个字符将被跳过.
            case '\\':
                quoted = 1;
                last_space = 0;
                continue;
//引号开头的token.
            case '"':
                start++;
                d_quoted = 1;
                last_space = 0;
                continue;

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

            default:
                last_space = 0;
            }
//到这里开始处理token字符.
        } else {
            if (ch == '{' && variable) {
                continue;
            }

            variable = 0;
//转义序列.其后面的一个字符将被跳过
            if (ch == '\\') {
                quoted = 1;
                continue;
            }
//？？？？？？这里是什么？？？
            if (ch == '$') {
                variable = 1;
                continue;
            }
//单双引号包围的字符是一个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、制表符、换行符、分号或是左花括号表示一个token的结束.
//也即发现一个token
            } else if (ch == ' ' || ch == '\t' || ch == CR || ch == LF
                       || ch == ';' || ch == '{')
            {
                last_space = 1;
                found = 1;
            }

            if (found) {
//将该token放入cf->args动态数组中.
                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;
                }
//将token中的转义序列转义后复制给word->data, 其他的按原样复制.
                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以分号结束则处理完一个配置行.
                if (ch == ';') {
                    return NGX_OK;
                }
//若token以左花括号结束则发现配置块起始位置.
                if (ch == '{') {
                    return NGX_CONF_BLOCK_START;
                }
//发现token标识清0
                found = 0;
            }
        }
    }
}