terom@15: #define _GNU_SOURCE
terom@15: #include <stdlib.h>
terom@15: #include <string.h>
terom@13: 
terom@13: #include "url.h"
terom@15: #include "lex.h"
terom@15: #include "error.h"
terom@15: #include "misc.h"
terom@13: 
terom@14: enum url_token {
terom@13:     URL_INVALID,
terom@13:     
terom@14:     URL_BEGIN,
terom@14: 
terom@14:     // kludge to resolve ambiguous URL_SCHEME/URL_USERNAME+URL_PASSWORD/URL_HOSTNAME+URL_SERVICE at the beginning
terom@14:     URL_BEGIN_ALNUM,
terom@14:     URL_BEGIN_COLON,
terom@14: 
terom@13:     URL_SCHEME,
terom@13:     URL_SCHEME_SEP,
terom@13:     URL_SCHEME_END_COL,
terom@13:     URL_SCHEME_END_SLASH1,
terom@13:     URL_SCHEME_END_SLASH2,
terom@14: 
terom@14:     // kludge to resolve ambiguous URL_USERNAME+URL_PASSWORD/URL_HOSTNAME+URL_SERVICE after a scheme 
terom@14:     URL_USERHOST_ALNUM,
terom@14:     URL_USERHOST_COLON,
terom@14:     URL_USERHOST_ALNUM2,
terom@13:     
terom@13:     URL_USERNAME,
terom@13:     URL_PASSWORD_SEP,
terom@13:     URL_PASSWORD,
terom@13:     URL_USERNAME_END,
terom@13: 
terom@13:     URL_HOSTNAME,
terom@13: 
terom@13:     URL_SERVICE_SEP,
terom@13:     URL_SERVICE,
terom@13: 
terom@13:     URL_PATH_START,
terom@13:     URL_PATH,
terom@13: 
terom@13:     URL_OPT_START,
terom@13:     URL_OPT_KEY,
terom@13:     URL_OPT_EQ,
terom@13:     URL_OPT_VAL,
terom@13:     URL_OPT_SEP,
terom@14:     
terom@13:     URL_MAX,
terom@13: };
terom@13: 
terom@14: /*
terom@14:  * Parser state
terom@14:  */
terom@14: struct url_state {
terom@15:     // the URL to parse into
terom@14:     struct url *url;
terom@15:     
terom@15:     // our lookahead-kludge
terom@15:     const char *alnum, *alnum2;
terom@15:     
terom@15: };
terom@13: 
terom@15: static int _url_append_scheme (struct url *url, const char *data) {
terom@15:     
terom@15: }
terom@13: 
terom@15: static int _url_append_opt_key (struct url *url, const char *key) {
terom@15: 
terom@15: }
terom@15: 
terom@15: static int _url_append_opt_val (struct url *url, const char *value) {
terom@15: 
terom@15: }
terom@13: 
terom@14: static int url_lex_token (int _this_token, char *token_data, int _next_token, int _prev_token, void *arg) {
terom@14:     enum url_token this_token = _this_token, next_token = _next_token, prev_token = _prev_token;
terom@14:     struct url_state *state = arg;
terom@15:     const char **copy_to = NULL;
terom@13: 
terom@15:     (void) prev_token;
terom@15:     
terom@15:     switch (this_token) {
terom@15:         case URL_BEGIN_ALNUM:
terom@15:             switch (next_token) {
terom@15:                 case URL_SCHEME_SEP:
terom@15:                     // store the scheme
terom@15:                     if (_url_append_scheme(state->url, token_data))
terom@15:                         goto error;
terom@15:                     
terom@15:                     break;
terom@15:                 
terom@15:                 case URL_USERNAME_END:
terom@15:                     // store the username
terom@15:                     copy_to = &state->url->username; break;
terom@15:                 
terom@15:                 case URL_PATH_START:
terom@15:                 case URL_OPT_START:
terom@15:                 case LEX_EOF:
terom@15:                     // store the hostname
terom@15:                     copy_to = &state->url->hostname; break;
terom@14: 
terom@15:                 case URL_BEGIN_COLON:
terom@15:                     // gah...
terom@15:                     copy_to = &state->alnum; break;
terom@15:                 
terom@15: 
terom@15:                 default:
terom@15:                     FATAL("weird next token");
terom@15:             }
terom@15:             
terom@15:             break;
terom@15: 
terom@15:         case URL_BEGIN_COLON:
terom@15:             switch (next_token) {
terom@15:                 case URL_SCHEME_END_SLASH1:
terom@15:                     // store the schema
terom@15:                     if (_url_append_scheme(state->url, token_data))
terom@15:                         goto error;
terom@15: 
terom@15:                     break;
terom@15:                 
terom@15:                 case URL_USERHOST_ALNUM2:
terom@15:                     // gah..
terom@15:                     break;
terom@15: 
terom@15:                 default:
terom@15:                     FATAL("weird next token");
terom@15:             }
terom@15: 
terom@15:             break;
terom@15: 
terom@15:         case URL_SCHEME:
terom@15:             // store the scheme
terom@15:             if (_url_append_scheme(state->url, token_data))
terom@15:                 goto error;
terom@15: 
terom@15:             break;
terom@15:     
terom@15:         case URL_SCHEME_SEP:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_SCHEME_END_COL:
terom@15:         case URL_SCHEME_END_SLASH1:
terom@15:         case URL_SCHEME_END_SLASH2:
terom@15:             // ignore
terom@15:             break;
terom@15:         
terom@15:         case URL_USERHOST_ALNUM:
terom@15:             switch (next_token) {
terom@15:                 case URL_USERNAME_END:
terom@15:                     // store the username
terom@15:                     copy_to = &state->url->username; break;
terom@15:                 
terom@15:                 case URL_PATH_START:
terom@15:                 case URL_OPT_START:
terom@15:                 case LEX_EOF:
terom@15:                     // store the hostname
terom@15:                     copy_to = &state->url->hostname; break;
terom@15: 
terom@15:                 case URL_USERHOST_COLON:
terom@15:                     // gah...
terom@15:                     copy_to = &state->alnum; break;
terom@15: 
terom@15:                 default:
terom@15:                     FATAL("weird next token");
terom@15:             }
terom@15:             
terom@15:             break;
terom@15: 
terom@15:         case URL_USERHOST_COLON:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_USERHOST_ALNUM2:
terom@15:             switch (next_token) {
terom@15:                 case URL_USERNAME_END:
terom@15:                     // store the username and password
terom@15:                     state->url->username = state->alnum; state->alnum = NULL;
terom@15:                     copy_to = &state->url->password;
terom@15: 
terom@15:                     break;
terom@15: 
terom@15:                 case URL_PATH_START:
terom@15:                 case URL_OPT_START:
terom@15:                 case LEX_EOF:
terom@15:                     // store the service
terom@15:                     copy_to = &state->url->service; break;
terom@15: 
terom@15:                 default:
terom@15:                     FATAL("weird next token");
terom@15:             }
terom@15: 
terom@15:             break;
terom@15: 
terom@15:         case URL_USERNAME:
terom@15:         case URL_PASSWORD_SEP:
terom@15:         case URL_PASSWORD:
terom@15:             FATAL("these should be overshadowed");
terom@15:         
terom@15:         case URL_USERNAME_END:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_HOSTNAME:
terom@15:             // store
terom@15:             copy_to = &state->url->hostname; break;
terom@15: 
terom@15:         case URL_SERVICE_SEP:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_SERVICE:
terom@15:             // store
terom@15:             copy_to = &state->url->service; break;
terom@15:         
terom@15:         case URL_PATH_START:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_PATH:
terom@15:             // store
terom@15:             copy_to = &state->url->path; break;
terom@15: 
terom@15:         case URL_OPT_START:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_OPT_KEY:
terom@15:             // store
terom@15:             if (_url_append_opt_key(state->url, token_data))
terom@15:                 goto error;
terom@15: 
terom@15:             break;
terom@15: 
terom@15:         case URL_OPT_EQ:
terom@15:             // ignore
terom@15:             break;
terom@15: 
terom@15:         case URL_OPT_VAL:
terom@15:             // store
terom@15:             if (_url_append_opt_val(state->url, token_data))
terom@15:                 goto error;
terom@15: 
terom@15:             break;
terom@15:         
terom@15:         case URL_OPT_SEP:
terom@15:             // ignore
terom@15:             break;
terom@15:         
terom@15:         default:
terom@15:             FATAL("invalid token");
terom@15:     }
terom@15:     
terom@15:     if (copy_to) {
terom@15:         // copy the token data
terom@15:         if ((*copy_to = strdup(token_data)) == NULL)
terom@15:             ERROR("strdup");
terom@15:     }
terom@15: 
terom@15:     // good
terom@15:     return 0;
terom@15: 
terom@15: error:
terom@15:     // XXX: error codes?
terom@15:     return -1;
terom@14: }
terom@14: 
terom@14: static struct lex url_lex = {
terom@15:     .token_fn = url_lex_token,
terom@15:     .char_fn = NULL,
terom@15:     .end_fn = NULL,
terom@15: 
terom@14:     .state_count = URL_MAX,
terom@14:     .state_list = {
terom@14:         LEX_STATE ( URL_BEGIN ) {
terom@14:             LEX_ALNUM       (           URL_BEGIN_ALNUM         ),
terom@14:             LEX_CHAR        (   ':',    URL_SERVICE_SEP         ),
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_END
terom@14:         },
terom@14:         
terom@14:         // this can be URL_SCHEME, URL_USERNAME or URL_HOSTNAME
terom@14:         LEX_STATE_END ( URL_BEGIN_ALNUM ) {
terom@14:             LEX_ALNUM       (           URL_BEGIN_ALNUM         ),
terom@14:             LEX_CHAR        (   '+',    URL_SCHEME_SEP          ),  // it was URL_SCHEME
terom@14:             LEX_CHAR        (   ':',    URL_BEGIN_COLON         ), 
terom@14:             LEX_CHAR        (   '@',    URL_USERNAME_END        ),  // it was URL_USERNAME
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),  // it was URL_HOSTNAME
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),  // it was URL_HOSTNAME
terom@14:             LEX_END
terom@14:         },
terom@14:         
terom@14:         // this can be URL_SCHEME_END_COL, URL_USERNAME_END or URL_SERVICE_SEP
terom@14:         LEX_STATE ( URL_BEGIN_COLON ) {
terom@14:             LEX_CHAR        (   '/',    URL_SCHEME_END_SLASH1   ),  // it was URL_SCHEME
terom@14:             LEX_ALNUM       (           URL_USERHOST_ALNUM2     ),
terom@14:             LEX_END
terom@14:         },
terom@14:        
terom@14: 
terom@14:         LEX_STATE ( URL_SCHEME ) { 
terom@14:             LEX_ALNUM       (           URL_SCHEME              ),
terom@14:             LEX_CHAR        (   '+',    URL_SCHEME_SEP          ),
terom@14:             LEX_CHAR        (   ':',    URL_SCHEME_END_COL      ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE ( URL_SCHEME_SEP ) {
terom@14:             LEX_ALNUM       (           URL_SCHEME              ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE ( URL_SCHEME_END_COL ) {
terom@14:             LEX_CHAR        (   '/',    URL_SCHEME_END_SLASH1   ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE ( URL_SCHEME_END_SLASH1 ) {
terom@14:             LEX_CHAR        (   '/',    URL_SCHEME_END_SLASH2   ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_SCHEME_END_SLASH2 ) {
terom@14:             LEX_ALNUM       (           URL_USERHOST_ALNUM      ),
terom@14:             LEX_CHAR        (   ':',    URL_SERVICE_SEP         ),
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_END
terom@14:         },
terom@14:         
terom@14:         // this can be URL_USERNAME or URL_HOSTNAME
terom@14:         LEX_STATE_END ( URL_USERHOST_ALNUM ) {
terom@14:             LEX_ALNUM       (           URL_USERHOST_ALNUM      ),
terom@14:             LEX_CHAR        (   ':',    URL_USERHOST_COLON      ), 
terom@14:             LEX_CHAR        (   '@',    URL_USERNAME_END        ),  // it was URL_USERNAME
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),  // it was URL_HOSTNAME
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),  // it was URL_HOSTNAME
terom@14:             LEX_END
terom@15:         },
terom@14:         
terom@14:         // this can be URL_USERNAME_END or URL_SERVICE_SEP
terom@14:         LEX_STATE ( URL_USERHOST_COLON ) {
terom@14:             LEX_ALNUM       (           URL_USERHOST_ALNUM2        ),
terom@14:             LEX_END
terom@14:         },
terom@14:         
terom@14:         // this can be URL_PASSWORD or URL_SERVICE
terom@14:         LEX_STATE_END ( URL_USERHOST_ALNUM2 ) {
terom@14:             LEX_ALNUM       (           URL_USERHOST_ALNUM      ),
terom@14:             LEX_CHAR        (   '@',    URL_USERNAME_END        ),  // it was URL_PASSSWORD
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),  // it was URL_SERVICE
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),  // it was URL_SERVICE
terom@14:             LEX_END
terom@14:         },
terom@14:         
terom@14:         // dummy states, covered by URL_USERHOST_ALNUM/URL_USERHOST_COLON/URL_USERHOST_ALNUM2
terom@14:         LEX_STATE ( URL_USERNAME ) {
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE ( URL_PASSWORD_SEP ) {
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE ( URL_PASSWORD ) {
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14: 
terom@14:         LEX_STATE_END ( URL_USERNAME_END ) {
terom@14:             LEX_ALNUM       (           URL_HOSTNAME            ), 
terom@14:             LEX_CHAR        (   ':',    URL_SERVICE_SEP         ),
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14: 
terom@14:         LEX_STATE_END ( URL_HOSTNAME ) {
terom@14:             LEX_ALNUM       (           URL_HOSTNAME            ), 
terom@14:             LEX_CHAR        (   ':',    URL_SERVICE_SEP         ),
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14: 
terom@14:         LEX_STATE ( URL_SERVICE_SEP ) {
terom@14:             LEX_ALNUM       (           URL_SERVICE            ), 
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_SERVICE ) {
terom@14:             LEX_ALNUM       (           URL_SERVICE            ), 
terom@14:             LEX_CHAR        (   '/',    URL_PATH_START          ),
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_END
terom@14:         },
terom@14: 
terom@14: 
terom@14:         LEX_STATE_END ( URL_PATH_START ) {
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_DEFAULT     (           URL_PATH                ),
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_PATH ) {
terom@14:             LEX_CHAR        (   '?',    URL_OPT_START           ),
terom@14:             LEX_DEFAULT     (           URL_PATH                ),
terom@14:         },
terom@14: 
terom@14: 
terom@14:         LEX_STATE_END ( URL_OPT_START ) {
terom@14:             LEX_CHAR        (   '&',    URL_OPT_SEP             ),
terom@15:             LEX_INVALID     (   '='                             ),
terom@14:             LEX_DEFAULT     (           URL_OPT_KEY             ),
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_OPT_KEY ) {
terom@14:             LEX_CHAR        (   '&',    URL_OPT_SEP             ),
terom@14:             LEX_CHAR        (   '=',    URL_OPT_EQ              ),
terom@14:             LEX_DEFAULT     (           URL_OPT_KEY             ),
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_OPT_EQ ) {
terom@14:             LEX_CHAR        (   '&',    URL_OPT_SEP             ),
terom@15:             LEX_INVALID     (   '='                             ),
terom@14:             LEX_DEFAULT     (           URL_OPT_VAL             ),
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_OPT_VAL ) {
terom@14:             LEX_CHAR        (   '&',    URL_OPT_SEP             ),
terom@15:             LEX_INVALID     (   '='                             ),
terom@14:             LEX_DEFAULT     (           URL_OPT_VAL             ),
terom@14:         },
terom@14: 
terom@14:         LEX_STATE_END ( URL_OPT_SEP ) {
terom@14:             LEX_CHAR        (   '&',    URL_OPT_SEP             ),
terom@15:             LEX_INVALID     (   '='                             ),
terom@14:             LEX_DEFAULT     (           URL_OPT_KEY             ),
terom@14:         },
terom@14:         
terom@14:         LEX_STATE ( URL_ERROR ) {
terom@14:             LEX_END
terom@14:         },
terom@15:     }
terom@14: };
terom@14: 
terom@14: int url_parse (struct url *url, const char *text) {
terom@14:     struct url_state state; ZINIT(state);
terom@14:     int ret;
terom@14: 
terom@14:     // set up state
terom@14:     state.url = url;
terom@14:     
terom@14:     // parse it
terom@14:     if ((ret = lexer(&url_lex, text, &state)))
terom@14:         ERROR("invalid URL");
terom@14: 
terom@14:     // success
terom@14:     return 0;
terom@14: 
terom@14: error:
terom@14:     return -1;
terom@14: }
terom@14: