2 #define I3__FILE__ "commands_parser.c"
4 * vim:ts=4:sw=4:expandtab
6 * i3 - an improved dynamic tiling window manager
7 * © 2009-2012 Michael Stapelberg and contributors (see also: LICENSE)
9 * commands_parser.c: hand-written parser to parse commands (commands are what
10 * you bind on keys and what you can send to i3 using the IPC interface, like
11 * 'move left' or 'workspace 4').
13 * We use a hand-written parser instead of lex/yacc because our commands are
14 * easy for humans, not for computers. Thus, it’s quite hard to specify a
15 * context-free grammar for the commands. A PEG grammar would be easier, but
16 * there’s downsides to every PEG parser generator I have come accross so far.
18 * This parser is basically a state machine which looks for literals or strings
19 * and can push either on a stack. After identifying a literal or string, it
20 * will either transition to the current state, to a different state, or call a
21 * function (like cmd_move()).
23 * Special care has been taken that error messages are useful and the code is
24 * well testable (when compiled with -DTEST_PARSER it will output to stdout
25 * instead of actually calling any function).
37 // Macros to make the YAJL API a bit easier to use.
38 #define y(x, ...) yajl_gen_ ## x (command_output.json_gen, ##__VA_ARGS__)
39 #define ystr(str) yajl_gen_string(command_output.json_gen, (unsigned char*)str, strlen(str))
41 /*******************************************************************************
42 * The data structures used for parsing. Essentially the current state and a
43 * list of tokens for that state.
45 * The GENERATED_* files are generated by generate-commands-parser.pl with the
46 * input parser-specs/commands.spec.
47 ******************************************************************************/
49 #include "GENERATED_enums.h"
51 typedef struct token {
54 /* This might be __CALL */
55 cmdp_state next_state;
57 uint16_t call_identifier;
61 typedef struct tokenptr {
66 #include "GENERATED_tokens.h"
68 /*******************************************************************************
69 * The (small) stack where identified literals are stored during the parsing
70 * of a single command (like $workspace).
71 ******************************************************************************/
74 /* Just a pointer, not dynamically allocated. */
75 const char *identifier;
79 /* 10 entries should be enough for everybody. */
80 static struct stack_entry stack[10];
83 * Pushes a string (identified by 'identifier') on the stack. We simply use a
84 * single array, since the number of entries we have to store is very small.
87 static void push_string(const char *identifier, char *str) {
88 for (int c = 0; c < 10; c++) {
89 if (stack[c].identifier != NULL)
91 /* Found a free slot, let’s store it here. */
92 stack[c].identifier = identifier;
97 /* When we arrive here, the stack is full. This should not happen and
98 * means there’s either a bug in this parser or the specification
99 * contains a command with more than 10 identified tokens. */
100 fprintf(stderr, "BUG: commands_parser stack full. This means either a bug "
101 "in the code, or a new command which contains more than "
102 "10 identified tokens.\n");
106 // XXX: ideally, this would be const char. need to check if that works with all
108 static char *get_string(const char *identifier) {
109 for (int c = 0; c < 10; c++) {
110 if (stack[c].identifier == NULL)
112 if (strcmp(identifier, stack[c].identifier) == 0)
118 static void clear_stack(void) {
119 for (int c = 0; c < 10; c++) {
120 if (stack[c].str != NULL)
122 stack[c].identifier = NULL;
127 // TODO: remove this if it turns out we don’t need it for testing.
129 /*******************************************************************************
130 * A dynamically growing linked list which holds the criteria for the current
132 ******************************************************************************/
134 typedef struct criterion {
138 TAILQ_ENTRY(criterion) criteria;
141 static TAILQ_HEAD(criteria_head, criterion) criteria =
142 TAILQ_HEAD_INITIALIZER(criteria);
145 * Stores the given type/value in the list of criteria.
146 * Accepts a pointer as first argument, since it is 'call'ed by the parser.
149 static void push_criterion(void *unused_criteria, const char *type,
151 struct criterion *criterion = malloc(sizeof(struct criterion));
152 criterion->type = strdup(type);
153 criterion->value = strdup(value);
154 TAILQ_INSERT_TAIL(&criteria, criterion, criteria);
158 * Clears the criteria linked list.
159 * Accepts a pointer as first argument, since it is 'call'ed by the parser.
162 static void clear_criteria(void *unused_criteria) {
163 struct criterion *criterion;
164 while (!TAILQ_EMPTY(&criteria)) {
165 criterion = TAILQ_FIRST(&criteria);
166 free(criterion->type);
167 free(criterion->value);
168 TAILQ_REMOVE(&criteria, criterion, criteria);
174 /*******************************************************************************
176 ******************************************************************************/
178 static cmdp_state state;
180 static Match current_match;
182 static struct CommandResult subcommand_output;
183 static struct CommandResult command_output;
185 #include "GENERATED_call.h"
188 static void next_state(const cmdp_token *token) {
189 if (token->next_state == __CALL) {
190 subcommand_output.json_gen = command_output.json_gen;
191 subcommand_output.needs_tree_render = false;
192 GENERATED_call(token->extra.call_identifier, &subcommand_output);
193 /* If any subcommand requires a tree_render(), we need to make the
194 * whole parser result request a tree_render(). */
195 if (subcommand_output.needs_tree_render)
196 command_output.needs_tree_render = true;
201 state = token->next_state;
202 if (state == INITIAL) {
207 /* TODO: Return parsing errors via JSON. */
208 struct CommandResult *parse_command(const char *input) {
209 DLOG("COMMAND: *%s*\n", input);
212 /* A YAJL JSON generator used for formatting replies. */
214 command_output.json_gen = yajl_gen_alloc(NULL);
216 command_output.json_gen = yajl_gen_alloc(NULL, NULL);
220 command_output.needs_tree_render = false;
222 const char *walk = input;
223 const size_t len = strlen(input);
225 const cmdp_token *token;
228 // TODO: make this testable
230 cmd_criteria_init(¤t_match, &subcommand_output);
233 /* The "<=" operator is intentional: We also handle the terminating 0-byte
234 * explicitly by looking for an 'end' token. */
235 while ((walk - input) <= len) {
236 /* skip whitespace and newlines before every token */
237 while ((*walk == ' ' || *walk == '\t' ||
238 *walk == '\r' || *walk == '\n') && *walk != '\0')
241 cmdp_token_ptr *ptr = &(tokens[state]);
242 token_handled = false;
243 for (c = 0; c < ptr->n; c++) {
244 token = &(ptr->array[c]);
247 if (token->name[0] == '\'') {
248 if (strncasecmp(walk, token->name + 1, strlen(token->name) - 1) == 0) {
249 if (token->identifier != NULL)
250 push_string(token->identifier, sstrdup(token->name + 1));
251 walk += strlen(token->name) - 1;
253 token_handled = true;
259 if (strcmp(token->name, "string") == 0 ||
260 strcmp(token->name, "word") == 0) {
261 const char *beginning = walk;
262 /* Handle quoted strings (or words). */
266 while (*walk != '\0' && (*walk != '"' || *(walk-1) == '\\'))
269 if (token->name[0] == 's') {
270 /* For a string (starting with 's'), the delimiters are
271 * comma (,) and semicolon (;) which introduce a new
272 * operation or command, respectively. Also, newlines
274 while (*walk != ';' && *walk != ',' &&
275 *walk != '\0' && *walk != '\r' &&
279 /* For a word, the delimiters are white space (' ' or
280 * '\t'), closing square bracket (]), comma (,) and
282 while (*walk != ' ' && *walk != '\t' &&
283 *walk != ']' && *walk != ',' &&
284 *walk != ';' && *walk != '\r' &&
285 *walk != '\n' && *walk != '\0')
289 if (walk != beginning) {
290 char *str = scalloc(walk-beginning + 1);
291 /* We copy manually to handle escaping of characters. */
293 for (inpos = 0, outpos = 0;
294 inpos < (walk-beginning);
296 /* We only handle escaped double quotes to not break
297 * backwards compatibility with people using \w in
298 * regular expressions etc. */
299 if (beginning[inpos] == '\\' && beginning[inpos+1] == '"')
301 str[outpos] = beginning[inpos];
303 if (token->identifier)
304 push_string(token->identifier, str);
305 /* If we are at the end of a quoted string, skip the ending
310 token_handled = true;
315 if (strcmp(token->name, "end") == 0) {
316 if (*walk == '\0' || *walk == ',' || *walk == ';') {
318 token_handled = true;
319 /* To make sure we start with an appropriate matching
320 * datastructure for commands which do *not* specify any
321 * criteria, we re-initialize the criteria system after
323 // TODO: make this testable
325 if (*walk == '\0' || *walk == ';')
326 cmd_criteria_init(¤t_match, &subcommand_output);
334 if (!token_handled) {
335 /* Figure out how much memory we will need to fill in the names of
336 * all tokens afterwards. */
338 for (c = 0; c < ptr->n; c++)
339 tokenlen += strlen(ptr->array[c].name) + strlen("'', ");
341 /* Build up a decent error message. We include the problem, the
342 * full input, and underline the position where the parser
345 char *possible_tokens = smalloc(tokenlen + 1);
346 char *tokenwalk = possible_tokens;
347 for (c = 0; c < ptr->n; c++) {
348 token = &(ptr->array[c]);
349 if (token->name[0] == '\'') {
350 /* A literal is copied to the error message enclosed with
353 strcpy(tokenwalk, token->name + 1);
354 tokenwalk += strlen(token->name + 1);
357 /* Any other token is copied to the error message enclosed
358 * with angle brackets. */
360 strcpy(tokenwalk, token->name);
361 tokenwalk += strlen(token->name);
364 if (c < (ptr->n - 1)) {
370 sasprintf(&errormessage, "Expected one of these tokens: %s",
372 free(possible_tokens);
374 /* Contains the same amount of characters as 'input' has, but with
375 * the unparseable part highlighted using ^ characters. */
376 char *position = smalloc(len + 1);
377 for (const char *copywalk = input; *copywalk != '\0'; copywalk++)
378 position[(copywalk - input)] = (copywalk >= walk ? '^' : ' ');
379 position[len] = '\0';
381 ELOG("%s\n", errormessage);
382 ELOG("Your command: %s\n", input);
383 ELOG(" %s\n", position);
385 /* Format this error message as a JSON reply. */
393 ystr("errorposition");
406 return &command_output;
409 /*******************************************************************************
410 * Code for building the stand-alone binary test.commands_parser which is used
411 * by t/187-commands-parser.t.
412 ******************************************************************************/
417 * Logs the given message to stdout while prefixing the current time to it,
418 * but only if debug logging was activated.
419 * This is to be called by DLOG() which includes filename/linenumber
422 void debuglog(char *fmt, ...) {
426 fprintf(stdout, "# ");
427 vfprintf(stdout, fmt, args);
431 void errorlog(char *fmt, ...) {
435 vfprintf(stderr, fmt, args);
439 int main(int argc, char *argv[]) {
441 fprintf(stderr, "Syntax: %s <command>\n", argv[0]);
444 parse_command(argv[1]);