return ret;
}
+static int cmd_declare(const char *cmd, const char *arg)
+{
+ struct cdecl *decl;
+ const char *str;
+ int ret = -1;
+
+ /* The name of the command is significant here. */
+ decl = cdecl_parse_english(cmd);
+ if (!decl)
+ goto out;
+
+ /*
+ * English parses have at most one full declarator, so no loop is
+ * needed here.
+ */
+ str = do_format(cdecl_declare, decl);
+ if (!str)
+ goto out;
+
+ printf("%s\n", str);
+ ret = 1;
+out:
+ cdecl_free(decl);
+ return ret;
+}
+
static int cmd_quit(const char *cmd, const char *arg)
{
return 0;
} commands[] = {
{ "explain", cmd_explain, "Explain a C declaration." },
{ "simplify", cmd_simplify, "Simplify a C declaration." },
+ { "declare", cmd_declare, "Construct a C declaration." },
+ { "type", cmd_declare, "Construct a C type name." },
{ "help", cmd_help, "Print this list of commands." },
{ "quit", cmd_quit, "Quit the program." },
{ "exit", cmd_quit, NULL }
%destructor { free_declarator($$); } <declarator>
%destructor { free_decl($$); } <decl>
+/* Magic tokens */
%token T_LEX_ERROR
+%token T_ENGLISH
%token <strval> T_IDENT "identifier"
%token <uintval> T_UINT "integer constant"
%token T_UNION "union"
%token T_ENUM "enum"
+/*
+ * English keywords.
+ */
+%token T_TYPE "type"
+%token T_DECLARE "declare"
+%token T_POINTER "pointer"
+%token T_FUNCTION "function"
+%token T_RETURNING "returning"
+%token T_ARRAY "array"
+%token T_TO "to"
+%token T_OF "of"
+%token T_AS "as"
+%token T_VLA "variable-length"
+
%type <strval> vla_ident
%type <boolval> varargs
%type <uintval> declspec_simple typespec_simple qualifier_simple
%type <decl> declaration declarators declarator_wrap
%type <decl> parameter parameters
+%type <strval> english_vla
+%type <declspec> storage_func_specs post_specs
+%type <declspec> type_qual_spec type_qual_specs typedef_name_qual
+%type <declarator> english_declarator english_array english_function
+%type <declarator> english_parameter_list null_decl
+%type <decl> english_parameter english_parameters
+%type <decl> english english_declaration
+
+/*
+ * Harmless shift/reduce conflicts in english_parameter. See comments below
+ * for more details.
+ */
+%expect 2
+
%%
-input: declaration {
+input: T_ENGLISH english {
+ *out = $2;
+} | declaration {
*out = $1;
};
$$->child = $1;
}
+english: T_DECLARE T_IDENT T_AS english_declaration {
+ $$ = $4;
+ for (struct cdecl_declarator *d = $$->declarators; d; d = d->child) {
+ if (d->type == CDECL_DECL_NULL) {
+ d->type = CDECL_DECL_IDENT;
+ d->u.ident = $2;
+ }
+ }
+} | T_TYPE english_declaration {
+ $$ = $2;
+}
+
+storage_func_specs: { $$ = NULL; } | declspec_simple storage_func_specs {
+ ALLOC_STRUCT($$, struct cdecl_declspec,
+ .type = $1,
+ .next = $2);
+}
+
+type_qual_spec: typespec_noid | qualifier
+
+type_qual_specs: { $$ = NULL; } | type_qual_spec type_qual_specs {
+ $$ = $1;
+ $$->next = $2;
+}
+
+/*
+ * The "qualifiers" nonterminal needs to be used here to avoid shift/reduce
+ * conflicts with pointer declarators. So we end up needing to stitch
+ * together three different specifiers lists.
+ */
+post_specs: qualifiers typespec type_qual_specs {
+ $$ = $2;
+ $$->next = $1;
+ for (struct cdecl_declspec *s = $1; s; s = s->next) {
+ if (!s->next) {
+ s->next = $3;
+ break;
+ }
+ }
+}
+
+english_declaration: storage_func_specs english_declarator post_specs {
+ ALLOC_STRUCT($$, struct cdecl,
+ .specifiers = $3,
+ .declarators = $2);
+
+ for (struct cdecl_declspec *s = $$->specifiers; s; s = s->next) {
+ if (!s->next) {
+ s->next = $1;
+ break;
+ }
+ }
+}
+
+english_declarator: {
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_NULL);
+} | english_declarator qualifiers T_POINTER T_TO {
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_POINTER,
+ .child = $1,
+ .u.pointer.qualifiers = $2);
+} | english_declarator english_array {
+ $$ = $2;
+ $$->child = $1;
+} | english_declarator english_function {
+ $$ = $2;
+ $$->child = $1;
+}
+
+english_function: T_FUNCTION T_RETURNING {
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_FUNCTION,
+ .u.function.parameters = NULL);
+} | T_FUNCTION T_LPAREN english_parameter_list T_RPAREN T_RETURNING {
+ $$ = $3;
+}
+
+english_parameter_list: english_parameters varargs {
+ struct cdecl *p, *c, *n;
+
+ /* Parameters were accumulated in reverse order. */
+ for (p = NULL, c = $1; c; p = c, c = n) {
+ n = c->next;
+ c->next = p;
+ }
+
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_FUNCTION,
+ .u.function.parameters = p,
+ .u.function.variadic = $2);
+}
+
+english_parameters: english_parameters T_COMMA english_parameter {
+ $$ = $3;
+ $$->next = $1;
+} | english_parameter
+
+typedef_name_qual: T_IDENT qualifiers {
+ ALLOC_STRUCT($$, struct cdecl_declspec,
+ .type = CDECL_TYPE_IDENT,
+ .ident = $1,
+ .next = $2);
+}
+
+null_decl: {
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_NULL);
+}
+
+/*
+ * There is a small shift/reduce conflict here. An unadorned identifier
+ * as the first thing in the parameter might be a typedef name deep in the
+ * first english_declaration (thus empty storage_func_specs and empty
+ * english_declarator need to be reduced) or it might be the identifier
+ * before the "as" (thus the identifier should be shifted).
+ *
+ * The typedef name conflict is the only issue, so treating it as a special
+ * case makes the shift harmless.
+ */
+english_parameter: english_declaration | typedef_name_qual null_decl {
+ ALLOC_STRUCT($$, struct cdecl,
+ .specifiers = $1,
+ .declarators = $2);
+} | T_IDENT T_AS english_declaration {
+ $$ = $3;
+ for (struct cdecl_declarator *d = $$->declarators; d; d = d->child) {
+ if (d->type == CDECL_DECL_NULL) {
+ d->type = CDECL_DECL_IDENT;
+ d->u.ident = $1;
+ }
+ }
+}
+
+english_array: T_VLA T_ARRAY english_vla T_OF {
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_ARRAY,
+ .u.array.vla = $3);
+} | T_ARRAY T_UINT T_OF {
+ if ($2 == 0)
+ FAIL("array length must be positive");
+
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_ARRAY,
+ .u.array.length = $2);
+} | T_ARRAY T_OF {
+ ALLOC_STRUCT($$, struct cdecl_declarator,
+ .type = CDECL_DECL_ARRAY,
+ .u.array.length = 0);
+}
+
+english_vla: T_IDENT | {
+ ALLOC($$, sizeof "");
+ strcpy($$, "");
+}
+
%%
void
yyerror(YYLTYPE *loc, yyscan_t scanner, struct cdecl **out,
* along with this program. If not, see <http://www.gnu.org/licenses/>.
*/
- #include "parse.h"
+#include "parse.h"
}
%option noyywrap bison-locations reentrant
+%option extra-type="_Bool"
%{
#define lex_error(msg) do { \
yyerror(yylloc, NULL, NULL, (msg)); \
return T_LEX_ERROR; \
} while(0)
+
+#define dup_token() do { \
+ yylval->strval = malloc(yyleng+1); \
+ if (!yylval->strval) \
+ lex_error("failed to allocate memory"); \
+ strcpy(yylval->strval, yytext); \
+} while(0)
%}
+%s ENGLISH
+
IDENT [_[:alpha:]][_[:alnum:]]*
INTEGER 0x[[:xdigit:]]+|0[0-7]+|[[:digit:]]+
%%
+%{
+ if (yyextra) {
+ yyextra = 0;
+ BEGIN(ENGLISH);
+ return T_ENGLISH;
+ }
+%}
+
"..." return T_ELLIPSIS;
";" return T_SEMICOLON;
"*" return T_ASTERISK;
return T_UINT;
}
-{IDENT} {
- yylval->strval = malloc(yyleng+1);
- if (!yylval->strval)
- lex_error("failed to allocate memory");
-
- strcpy(yylval->strval, yytext);
- return T_IDENT;
+<ENGLISH>{
+ "variable-length" return T_VLA;
+ "type" return T_TYPE;
+ "declare" return T_DECLARE;
+ "pointer" return T_POINTER;
+ "function" return T_FUNCTION;
+ "returning" return T_RETURNING;
+ "array" return T_ARRAY;
+ "to" return T_TO;
+ "of" return T_OF;
+ "as" return T_AS;
}
+{IDENT} { dup_token(); return T_IDENT; }
+
[[:space:]]+
. {
char buf[] = "syntax error, unexpected #";