cdump: new module.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

Makefile-ccan

@@ -41,6 +41,7 @@ MODS_WITH_SRC := antithread \
 	btree \
 	bytestring \
 	ccan_tokenizer \
+	cdump \
 	charset \
 	ciniparser \
 	crc \

ccan/cdump/LICENSE

+../../licenses/BSD-MIT
\ No newline at end of file

ccan/cdump/_info

+#include "config.h"
+#include <stdio.h>
+#include <string.h>
+
+/**
+ * cdump - routines to parse simple C structures.
+ *
+ * This code is designed to produce data structures summarizing C code.
+ * It only operates on simple, well-formed C code (eg. specific headers
+ * which you want to autogenerate from), but it should be fairly easy to
+ * enhance if desired.
+ *
+ * Author: Rusty Russell <rusty@rustcorp.com.au>
+ * License: BSD-MIT
+ *
+ * Example:
+ * // Creates a simple print function for a structure.
+ * #include <ccan/cdump/cdump.h>
+ * #include <ccan/tal/grab_file/grab_file.h>
+ * #include <ccan/err/err.h>
+ *
+ * static void print_as(const char *fmt, const char *member_name)
+ * {
+ * 	printf("\tprintf(\"%%s:%s\\n\", \"%s\", s->%s);\n",
+ * 	       fmt, member_name, member_name);
+ * }
+ *
+ * int main(int argc, char *argv[])
+ * {
+ * 	char *code, *problems;
+ * 	struct cdump_definitions *defs;
+ * 	int i, j;
+ *
+ * 	// Read code from stdin.
+ * 	code = grab_file(NULL, NULL);
+ *
+ * 	defs = cdump_extract(NULL, code, &problems);
+ * 	if (!defs)
+ * 		errx(1, "Parsing stdin: %s", problems);
+ *
+ * 	for (i = 1; i < argc; i++) {
+ * 		struct cdump_type *t = strmap_get(&defs->structs, argv[i]);
+ * 		if (!t)
+ * 			errx(1, "Could not find struct %s", argv[i]);
+ *
+ * 		printf("void print_struct_%s(const struct %s *s)\n"
+ * 			"{\n", argv[i], argv[i]);
+ * 		for (j = 0; j < tal_count(t->u.members); j++) {
+ * 			const struct cdump_member *m = t->u.members + j;
+ * 			switch (m->type->kind) {
+ * 			case CDUMP_STRUCT:
+ * 			case CDUMP_UNION:
+ * 			case CDUMP_ARRAY:
+ * 				// Too hard for this simple example.
+ * 				printf("\tprintf(\"%%s:???\\n\", \"%s\");\n",
+ * 				       m->name);
+ * 				break;
+ * 			case CDUMP_ENUM:
+ * 				print_as("%i", m->name);
+ * 				break;
+ * 			case CDUMP_POINTER:
+ * 				print_as("%p", m->name);
+ * 				break;
+ * 			case CDUMP_UNKNOWN:
+ * 				if (!strcmp(m->type->name, "int"))
+ * 					print_as("%i", m->name);
+ * 				else if (!strcmp(m->type->name, "long int"))
+ * 					print_as("%li", m->name);
+ * 				else if (!strcmp(m->type->name, "unsigned int"))
+ * 					print_as("%u", m->name);
+ * 				// etc...
+ * 				break;
+ * 			}
+ * 		}
+ * 		printf("}\n");
+ * 	}
+ * 	return 0;
+ * }
+ */
+int main(int argc, char *argv[])
+{
+	/* Expect exactly one argument */
+	if (argc != 2)
+		return 1;
+
+	if (strcmp(argv[1], "depends") == 0) {
+		printf("ccan/tal\n");
+		printf("ccan/tal/str\n");
+		printf("ccan/strmap\n");
+		return 0;
+	}
+
+	return 1;
+}

ccan/cdump/cdump.h

+/* MIT (BSD) license - see LICENSE file for details */
+#ifndef CCAN_CDUMP_H
+#define CCAN_CDUMP_H
+#include <ccan/strmap/strmap.h>
+#include <ccan/tal/tal.h>
+
+enum cdump_type_kind {
+	CDUMP_STRUCT,
+	CDUMP_UNION,
+	CDUMP_ENUM,
+	CDUMP_ARRAY,
+	CDUMP_POINTER,
+	CDUMP_UNKNOWN
+};
+
+struct cdump_member {
+	const char *name;
+	/* const, volatile */
+	const char *qualifiers;
+	struct cdump_type *type;
+};
+
+struct cdump_enum_val {
+	const char *name;
+	/* Either NULL, or whatever follows '=' sign */
+	const char *value;
+};
+
+struct cdump_array {
+	const char *size;
+	struct cdump_type *type;
+};
+
+struct cdump_type {
+	enum cdump_type_kind kind;
+	const char *name;
+	union {
+		/* CDUMP_STRUCT / CDUMP_UNION: array */
+		struct cdump_member *members;
+		/* CDUMP_ENUM: array */
+		struct cdump_enum_val *enum_vals;
+		/* CDUMP_ARRAY */
+		struct cdump_array arr;
+		/* CDUMP_POINTER */
+		const struct cdump_type *ptr;
+	} u;
+};
+
+/* The map of typenames to definitions */
+struct cdump_map {
+	STRMAP_MEMBERS(struct cdump_type *);
+};
+
+struct cdump_definitions {
+	struct cdump_map enums;
+	struct cdump_map structs;
+	struct cdump_map unions;
+};
+
+/**
+ * cdump_extract - extract definitions from simple C code.
+ * @ctx: context to tal() the return and @problems from (or NULL)
+ * @code: a nul-terminated string of C definitions
+ * @problems: a pointer to a char * to report problems (or NULL)
+ *
+ * This function parses @code and extracts enum, struct and union definitions
+ * into the return.  If there is a parse error, it will return NULL and
+ * allocate a problem string for human consumption.
+ *
+ * Example:
+ *	// Returns name of first field of 'struct @name' in @code.
+ *	static const char *first_field_of_struct(const char *code,
+ *						 const char *name)
+ *	{
+ *		char *problems;
+ *		struct cdump_definitions *defs;
+ *		struct cdump_type *t;
+ *
+ *		defs = cdump_extract(NULL, code, &problems);
+ *		if (!defs) {
+ *			fprintf(stderr, "%s", problems);
+ *			tal_free(problems);
+ *			return NULL;
+ *		}
+ *		t = strmap_get(&defs->structs, name);
+ *		if (!t) {
+ *			fprintf(stderr, "Couldn't find struct %s", name);
+ *			return NULL;
+ *		}
+ *		assert(t->kind == CDUMP_STRUCT);
+ *		return t->u.members[0].name;
+ *	}
+ */
+struct cdump_definitions *cdump_extract(const tal_t *ctx, const char *code,
+					char **problems);
+#endif /* CCAN_CDUMP_H */

ccan/cdump/test/run-forward-decl.c

+#include <ccan/cdump/cdump.h>
+/* Include the C files directly. */
+#include <ccan/cdump/cdump.c>
+#include <ccan/tap/tap.h>
+
+int main(void)
+{
+	struct cdump_definitions *defs;
+	const struct cdump_type *t, *t2;
+	char *ctx = tal(NULL, char), *problems;
+
+	/* This is how many tests you plan to run */
+	plan_tests(16);
+
+	defs = cdump_extract(ctx, "struct foo { struct bar *bar; };\n"
+			     "struct bar { int x; };", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	t2 = strmap_get(&defs->structs, "bar");
+	ok1(t2);
+
+	ok1(t2->kind == CDUMP_STRUCT);
+	ok1(streq(t2->name, "bar"));
+	ok1(tal_count(t2->u.members) == 1);
+	ok1(t2->u.members[0].type->kind == CDUMP_UNKNOWN);
+	ok1(streq(t2->u.members[0].type->name, "int"));
+
+	ok1(t->kind == CDUMP_STRUCT);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.members) == 1);
+	ok1(streq(t->u.members[0].name, "bar"));
+	ok1(t->u.members[0].type->kind == CDUMP_POINTER);
+	ok1(t->u.members[0].type->u.ptr == t2);
+
+	tal_free(ctx);
+
+	/* This exits depending on whether all tests passed */
+	return exit_status();
+}

ccan/cdump/test/run-qualifiers.c

+#include <ccan/cdump/cdump.h>
+/* Include the C files directly. */
+#include <ccan/cdump/cdump.c>
+#include <ccan/tap/tap.h>
+
+int main(void)
+{
+	struct cdump_definitions *defs;
+	const struct cdump_type *t, *p;
+	char *ctx = tal(NULL, char), *problems;
+
+	/* This is how many tests you plan to run */
+	plan_tests(63);
+
+	defs = cdump_extract(ctx,
+			     "struct foo {\n"
+			     "	long l;\n"
+			     "	long int li;\n"
+			     "	unsigned long *ulp;\n"
+			     "	unsigned long int *ulip;\n"
+			     "};", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->enums));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_STRUCT);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.members) == 4);
+
+	ok1(streq(t->u.members[0].name, "l"));
+	p = t->u.members[0].type;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "long"));
+
+	ok1(streq(t->u.members[1].name, "li"));
+	p = t->u.members[1].type;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "long int"));
+
+	ok1(streq(t->u.members[2].name, "ulp"));
+	p = t->u.members[2].type;
+	ok1(p->kind == CDUMP_POINTER);
+	p = p->u.ptr;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "unsigned long"));
+
+	ok1(streq(t->u.members[3].name, "ulip"));
+	p = t->u.members[3].type;
+	ok1(p->kind == CDUMP_POINTER);
+	p = p->u.ptr;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "unsigned long int"));
+
+	defs = cdump_extract(ctx,
+			     "struct foo {\n"
+			     "	volatile long vl;\n"
+			     "	const long cl;\n"
+			     "	volatile const long long int *vclli;\n"
+			     "};", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->enums));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_STRUCT);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.members) == 3);
+
+	ok1(streq(t->u.members[0].name, "vl"));
+	ok1(streq(t->u.members[0].qualifiers, "volatile"));
+	p = t->u.members[0].type;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "long"));
+
+	ok1(streq(t->u.members[1].name, "cl"));
+	ok1(streq(t->u.members[1].qualifiers, "const"));
+	p = t->u.members[1].type;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "long"));
+
+	ok1(streq(t->u.members[2].name, "vclli"));
+	ok1(streq(t->u.members[2].qualifiers, "volatile const"));
+	p = t->u.members[2].type;
+	ok1(p->kind == CDUMP_POINTER);
+	p = p->u.ptr;
+	ok1(p->kind == CDUMP_UNKNOWN);
+	ok1(streq(p->name, "long long int"));
+
+	defs = cdump_extract(ctx,
+			     "struct foo {\n"
+			     "	volatile struct bar *a, b;\n"
+			     "};", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->enums));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_STRUCT);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.members) == 2);
+
+	ok1(streq(t->u.members[0].name, "a"));
+	ok1(streq(t->u.members[0].qualifiers, "volatile"));
+	p = t->u.members[0].type;
+	ok1(p->kind == CDUMP_POINTER);
+	p = p->u.ptr;
+	ok1(p->kind == CDUMP_STRUCT);
+	ok1(streq(p->name, "bar"));
+
+	ok1(streq(t->u.members[1].name, "b"));
+	ok1(streq(t->u.members[1].qualifiers, "volatile"));
+	p = t->u.members[1].type;
+	ok1(p->kind == CDUMP_STRUCT);
+	ok1(streq(p->name, "bar"));
+
+	tal_free(ctx);
+
+	/* This exits depending on whether all tests passed */
+	return exit_status();
+}

ccan/cdump/test/run.c

+#include <ccan/cdump/cdump.h>
+/* Include the C files directly. */
+#include <ccan/cdump/cdump.c>
+#include <ccan/tap/tap.h>
+
+int main(void)
+{
+	struct cdump_definitions *defs;
+	const struct cdump_type *t, *p;
+	char *ctx = tal(NULL, char), *problems;
+
+	/* This is how many tests you plan to run */
+	plan_tests(94);
+
+	defs = cdump_extract(ctx, "enum foo { BAR };", NULL);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+
+	ok1(strmap_empty(&defs->structs));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->enums, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_ENUM);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.enum_vals) == 1);
+	ok1(streq(t->u.enum_vals[0].name, "BAR"));
+	ok1(!t->u.enum_vals[0].value);
+
+	defs = cdump_extract(ctx, "enum foo { BAR = 7 };", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->structs));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->enums, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_ENUM);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.enum_vals) == 1);
+	ok1(streq(t->u.enum_vals[0].name, "BAR"));
+	ok1(streq(t->u.enum_vals[0].value, "7"));
+
+	defs = cdump_extract(ctx, "enum foo { BAR = 7, BAZ, FUZZ };", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->structs));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->enums, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_ENUM);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.enum_vals) == 3);
+	ok1(streq(t->u.enum_vals[0].name, "BAR"));
+	ok1(streq(t->u.enum_vals[0].value, "7"));
+	ok1(streq(t->u.enum_vals[1].name, "BAZ"));
+	ok1(!t->u.enum_vals[1].value);
+	ok1(streq(t->u.enum_vals[2].name, "FUZZ"));
+	ok1(!t->u.enum_vals[2].value);
+
+	defs = cdump_extract(ctx, "struct foo { int x; };", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->enums));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_STRUCT);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.members) == 1);
+	ok1(streq(t->u.members[0].name, "x"));
+	ok1(t->u.members[0].type->kind == CDUMP_UNKNOWN);
+	ok1(streq(t->u.members[0].type->name, "int"));
+
+	defs = cdump_extract(ctx, "struct foo { int x[5<< 1]; struct foo *next; struct unknown **ptrs[10]; };", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+
+	ok1(strmap_empty(&defs->enums));
+	ok1(strmap_empty(&defs->unions));
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	ok1(t->kind == CDUMP_STRUCT);
+	ok1(streq(t->name, "foo"));
+	ok1(tal_count(t->u.members) == 3);
+
+	ok1(streq(t->u.members[0].name, "x"));
+	ok1(t->u.members[0].type->kind == CDUMP_ARRAY);
+	ok1(streq(t->u.members[0].type->u.arr.size, "5<< 1"));
+	ok1(t->u.members[0].type->u.arr.type->kind == CDUMP_UNKNOWN);
+	ok1(streq(t->u.members[0].type->u.arr.type->name, "int"));
+
+	ok1(streq(t->u.members[1].name, "next"));
+	ok1(t->u.members[1].type->kind == CDUMP_POINTER);
+	ok1(t->u.members[1].type->u.ptr == t);
+
+	ok1(streq(t->u.members[2].name, "ptrs"));
+	p = t->u.members[2].type;
+	ok1(p->kind == CDUMP_ARRAY);
+	ok1(streq(p->u.arr.size, "10"));
+	p = p->u.arr.type;
+	ok1(p->kind == CDUMP_POINTER);
+	p = p->u.ptr;
+	ok1(p->kind == CDUMP_POINTER);
+	p = p->u.ptr;
+	ok1(p->kind == CDUMP_STRUCT);
+	ok1(streq(p->name, "unknown"));
+	ok1(p->u.members == NULL);
+
+	/* We don't put undefined structs into definition maps. */
+	ok1(!strmap_get(&defs->structs, "unknown"));
+
+	/* unions and comments. */
+	defs = cdump_extract(ctx, "#if 0\n"
+			     "/* Normal comment */\n"
+			     "struct foo { int x[5 * 7/* Comment */]; };\n"
+			     "// One-line comment\n"
+			     "union bar { enum sometype x; union yun// Comment\n"
+			     " y;};\n"
+			     "#endif", &problems);
+	ok1(defs);
+	ok1(tal_parent(defs) == ctx);
+	ok1(!problems);
+	t = strmap_get(&defs->structs, "foo");
+	ok1(t);
+	ok1(tal_count(t->u.members) == 1);
+	ok1(streq(t->u.members[0].name, "x"));
+	ok1(t->u.members[0].type->kind == CDUMP_ARRAY);
+	ok1(streq(t->u.members[0].type->u.arr.size, "5 * 7"));
+	ok1(t->u.members[0].type->u.arr.type->kind == CDUMP_UNKNOWN);
+	ok1(streq(t->u.members[0].type->u.arr.type->name, "int"));
+
+	t = strmap_get(&defs->unions, "bar");
+	ok1(t);
+	ok1(tal_count(t->u.members) == 2);
+	ok1(streq(t->u.members[0].name, "x"));
+	ok1(t->u.members[0].type->kind == CDUMP_ENUM);
+	ok1(streq(t->u.members[0].type->name, "sometype"));
+	ok1(!t->u.members[0].type->u.enum_vals);
+	ok1(streq(t->u.members[1].name, "y"));
+	ok1(t->u.members[1].type->kind == CDUMP_UNION);
+	ok1(streq(t->u.members[1].type->name, "yun"));
+	ok1(!t->u.members[1].type->u.members);
+
+	/* This exits depending on whether all tests passed */
+	return exit_status();
+}