tal/str: accept take() for arguments.

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

ccan/tal/str/_info

@@ -45,6 +45,7 @@ int main(int argc, char *argv[])
 	if (strcmp(argv[1], "depends") == 0) {
 		printf("ccan/str\n");
 		printf("ccan/tal\n");
+		printf("ccan/take\n");
 		return 0;
 	}
 

ccan/tal/str/str.c

@@ -9,98 +9,149 @@
 #include <regex.h>
 #include <stdarg.h>
 #include <unistd.h>
-#include <ccan/tal/tal.h>
 #include <ccan/str/str.h>
+#include <ccan/tal/tal.h>
+#include <ccan/take/take.h>
 
-char **strsplit(const void *ctx, const char *string, const char *delims,
-		enum strsplit flags)
+char **strsplit(const tal_t *ctx,
+		const char *string, const char *delims, enum strsplit flags)
 {
-	char **lines = NULL;
+	char **parts, *str;
 	size_t max = 64, num = 0;
 
-	lines = tal_arr(ctx, char *, max+1);
+	parts = tal_arr(ctx, char *, max + 1);
+	if (unlikely(!parts)) {
+		if (taken(string))
+			tal_free(string);
+		if (taken(delims))
+			tal_free(delims);
+		return NULL;
+	}
+	str = tal_strdup(parts, string);
+	if (unlikely(!str))
+		goto fail;
+	if (unlikely(!delims) && is_taken(delims))
+		goto fail;
 
 	if (flags == STR_NO_EMPTY)
-		string += strspn(string, delims);
+		str += strspn(str, delims);
 
-	while (*string != '\0') {
-		size_t len = strcspn(string, delims), dlen;
+	while (*str != '\0') {
+		size_t len = strcspn(str, delims), dlen;
 
-		lines[num] = tal_arr(lines, char, len + 1);
-		memcpy(lines[num], string, len);
-		lines[num][len] = '\0';
-		string += len;
-		dlen = strspn(string, delims);
+		parts[num] = str;
+		dlen = strspn(str + len, delims);
+		parts[num][len] = '\0';
 		if (flags == STR_EMPTY_OK && dlen)
 			dlen = 1;
-		string += dlen;
-		if (++num == max)
-			tal_resize(&lines, max*=2 + 1);
+		str += len + dlen;
+		if (++num == max && !tal_resize(&parts, max*=2 + 1))
+			goto fail;
 	}
-	lines[num] = NULL;
-	return lines;
+	parts[num] = NULL;
+	if (taken(delims))
+		tal_free(delims);
+	return parts;
+
+fail:
+	tal_free(parts);
+	if (taken(delims))
+		tal_free(delims);
+	return NULL;
 }
 
-char *strjoin(const void *ctx, char *strings[], const char *delim,
-	      enum strjoin flags)
+char *strjoin(const tal_t *ctx,
+	      char *strings[], const char *delim, enum strjoin flags)
 {
 	unsigned int i;
-	char *ret = tal_strdup(ctx, "");
-	size_t totlen = 0, dlen = strlen(delim);
+	char *ret = NULL;
+	size_t totlen = 0, dlen;
 
+	if (unlikely(!strings) && is_taken(strings))
+		goto fail;
+
+	if (unlikely(!delim) && is_taken(delim))
+		goto fail;
+
+	dlen = strlen(delim);
+	ret = tal_arr(ctx, char, dlen*2+1);
+	if (!ret)
+		goto fail;
+
+	ret[0] = '\0';
 	for (i = 0; strings[i]; i++) {
 		size_t len = strlen(strings[i]);
+
 		if (flags == STR_NO_TRAIL && !strings[i+1])
 			dlen = 0;
-		tal_resize(&ret, totlen + len + dlen + 1);
+		if (!tal_resize(&ret, totlen + len + dlen + 1))
+			goto fail;
 		memcpy(ret + totlen, strings[i], len);
 		totlen += len;
 		memcpy(ret + totlen, delim, dlen);
 		totlen += dlen;
 	}
 	ret[totlen] = '\0';
+out:
+	if (taken(strings))
+		tal_free(strings);
+	if (taken(delim))
+		tal_free(delim);
 	return ret;
+fail:
+	ret = tal_free(ret);
+	goto out;
 }
 
-bool strreg(const void *ctx, const char *string, const char *regex, ...)
+bool strreg(const tal_t *ctx, const char *string, const char *regex, ...)
 {
 	size_t nmatch = 1 + strcount(regex, "(");
-	regmatch_t *matches = tal_arr(ctx, regmatch_t, nmatch);
+	regmatch_t matches[nmatch];
 	regex_t r;
-	bool ret;
-
-	if (!matches || regcomp(&r, regex, REG_EXTENDED) != 0)
-		return false;
-
-	if (regexec(&r, string, nmatch, matches, 0) == 0) {
-		unsigned int i;
-		va_list ap;
-
-		ret = true;
-		va_start(ap, regex);
-		for (i = 1; i < nmatch; i++) {
-			char **arg;
-			arg = va_arg(ap, char **);
-			if (arg) {
-				/* eg. ([a-z])? can give "no match". */
-				if (matches[i].rm_so == -1)
-					*arg = NULL;
-				else {
-					*arg = tal_strndup(ctx,
-						      string + matches[i].rm_so,
-						      matches[i].rm_eo
-						      - matches[i].rm_so);
-					if (!*arg) {
-						ret = false;
-						break;
-					}
+	bool ret = false;
+	unsigned int i;
+	va_list ap;
+
+	if (unlikely(!regex) && is_taken(regex))
+		goto fail_no_re;
+
+	if (regcomp(&r, regex, REG_EXTENDED) != 0)
+		goto fail_no_re;
+
+	if (unlikely(!string) && is_taken(string))
+		goto fail;
+
+	if (regexec(&r, string, nmatch, matches, 0) != 0)
+		goto fail;
+
+	ret = true;
+	va_start(ap, regex);
+	for (i = 1; i < nmatch; i++) {
+		char **arg = va_arg(ap, char **);
+		if (arg) {
+			/* eg. ([a-z])? can give "no match". */
+			if (matches[i].rm_so == -1)
+				*arg = NULL;
+			else {
+				*arg = tal_strndup(ctx,
+						   string + matches[i].rm_so,
+						   matches[i].rm_eo
+						   - matches[i].rm_so);
+				/* FIXME: If we fail, we set some and leak! */
+				if (!*arg) {
+					ret = false;
+					break;
 				}
 			}
 		}
-		va_end(ap);
-	} else
-		ret = false;
-	tal_free(matches);
+	}
+	va_end(ap);
+fail:
 	regfree(&r);
+fail_no_re:
+	if (taken(regex))
+		tal_free(regex);
+	if (taken(string))
+		tal_free(string);
 	return ret;
 }

ccan/tal/str/str.h

@@ -13,16 +13,18 @@ enum strsplit {
 
 /**
  * strsplit - Split string into an array of substrings
- * @ctx: the parent to tal from (often NULL)
- * @string: the string to split
- * @delims: delimiters where lines should be split.
+ * @ctx: the context to tal from (often NULL).
+ * @string: the string to split (can be take()).
+ * @delims: delimiters where lines should be split (can be take()).
  * @flags: whether to include empty substrings.
  *
- * This function splits a single string into multiple strings.  The
- * original string is untouched: an array is allocated (using tal)
- * pointing to copies of each substring.  Multiple delimiters result
- * in empty substrings.  By definition, no delimiters will appear in
- * the substrings.
+ * This function splits a single string into multiple strings.
+ *
+ * If @string is take(), the returned array will point into the
+ * mangled @string.
+ *
+ * Multiple delimiters result in empty substrings.  By definition, no
+ * delimiters will appear in the substrings.
  *
  * The final char * in the array will be NULL.
  *
@@ -43,8 +45,8 @@ enum strsplit {
  *		return long_lines;
  *	}
  */
-char **strsplit(const void *ctx, const char *string, const char *delims,
-		enum strsplit flags);
+char **strsplit(const tal_t *ctx,
+		const char *string, const char *delims, enum strsplit flags);
 
 enum strjoin {
 	STR_TRAIL,
@@ -53,9 +55,9 @@ enum strjoin {
 
 /**
  * strjoin - Join an array of substrings into one long string
- * @ctx: the context to tal from (often NULL)
- * @strings: the NULL-terminated array of strings to join
- * @delim: the delimiter to insert between the strings
+ * @ctx: the context to tal from (often NULL).
+ * @strings: the NULL-terminated array of strings to join (can be take())
+ * @delim: the delimiter to insert between the strings (can be take())
  * @flags: whether to add a delimieter to the end
  *
  * This function joins an array of strings into a single string.  The
@@ -80,8 +82,8 @@ char *strjoin(const void *ctx, char *strings[], const char *delim,
 /**
  * strreg - match and extract from a string via (extended) regular expressions.
  * @ctx: the context to tal from (often NULL)
- * @string: the string to try to match.
- * @regex: the regular expression to match.
+ * @string: the string to try to match (can be take())
+ * @regex: the regular expression to match (can be take())
  * ...: pointers to strings to allocate for subexpressions.
  *
  * Returns true if we matched, in which case any parenthesized

ccan/tal/str/test/run-strreg.c

@@ -2,16 +2,6 @@
 #include <ccan/tal/str/str.c>
 #include <ccan/tap/tap.h>
 
-static unsigned int tal_total_blocks(tal_t *ctx)
-{
-	unsigned int num = 1;
-	tal_t *i;
-
-	for (i = tal_first(ctx); i; i = tal_next(ctx, i))
-		num++;
-	return num;
-}
-
 static bool find_parent(tal_t *child, tal_t *parent)
 {
 	tal_t *i;
@@ -26,12 +16,11 @@ static bool find_parent(tal_t *child, tal_t *parent)
 int main(int argc, char *argv[])
 {
 	void *ctx = tal_strdup(NULL, "toplevel");
-	unsigned int top_blocks = tal_total_blocks(ctx);
 	char *a, *b;
 	/* If it accesses this, it will crash. */
 	char **invalid = (char **)1L;
 
-	plan_tests(25);
+	plan_tests(40);
 	/* Simple matching. */
 	ok1(strreg(ctx, "hello world!", "hello") == true);
 	ok1(strreg(ctx, "hello world!", "hi") == false);
@@ -88,7 +77,42 @@ int main(int argc, char *argv[])
 	tal_free(a);
 
 	/* No leaks! */
-	ok1(tal_total_blocks(ctx) == top_blocks);
+	ok1(!tal_first(ctx));
+
+	/* NULL arg with take means always fail. */
+	ok1(strreg(ctx, take(NULL), "((hello|goodbye) world)",
+		   &b, NULL, invalid) == false);
+
+	/* Take string. */
+	a = tal_strdup(ctx, "hello world!");
+	ok1(strreg(ctx, take(a), "([a-z]+)", &b, invalid) == true);
+	ok1(streq(b, "hello"));
+	ok1(tal_parent(b) == ctx);
+	tal_free(b);
+	ok1(tal_first(ctx) == NULL);
+
+	/* Take regex. */
+	a = tal_strdup(ctx, "([a-z]+)");
+	ok1(strreg(ctx, "hello world!", take(a), &b, invalid) == true);
+	ok1(streq(b, "hello"));
+	ok1(tal_parent(b) == ctx);
+	tal_free(b);
+	ok1(tal_first(ctx) == NULL);
+
+	/* Take both. */
+	a = tal_strdup(ctx, "([a-z]+)");
+	ok1(strreg(ctx, take(tal_strdup(ctx, "hello world!")),
+		   take(a), &b, invalid) == true);
+	ok1(streq(b, "hello"));
+	ok1(tal_parent(b) == ctx);
+	tal_free(b);
+	ok1(tal_first(ctx) == NULL);
+
+	/* ... even if we fail to match. */
+	a = tal_strdup(ctx, "([a-z]+)");
+	ok1(strreg(ctx, take(tal_strdup(ctx, "HELLO WORLD!")),
+		   take(a), &b, invalid) == false);
+	ok1(tal_first(ctx) == NULL);
 	tal_free(ctx);
 
 	return exit_status();

ccan/tal/str/test/run.c

@@ -14,7 +14,7 @@ int main(int argc, char *argv[])
 	char **split, *str;
 	void *ctx;
 
-	plan_tests(24);
+	plan_tests(65);
 	split = strsplit(NULL, "hello  world", " ", STR_EMPTY_OK);
 	ok1(!strcmp(split[0], "hello"));
 	ok1(!strcmp(split[1], ""));
@@ -61,5 +61,93 @@ int main(int argc, char *argv[])
 	ok1(tal_parent(str) == ctx);
 	tal_free(ctx);
 
+	ctx = tal_strdup(NULL, "context");
+	/* Pass through NULLs from take. */
+	ok1(strsplit(NULL, take(NULL), " ", STR_EMPTY_OK) == NULL);
+	ok1(strsplit(NULL, "foo", take(NULL), STR_EMPTY_OK) == NULL);
+
+	/* strsplit take string.  It reallocs it to same size, but
+	 * that sometimes causes a move, so we can't directly check
+	 * that split[0] == str. */
+	str = tal_strdup(ctx, "hello world");
+	ok1(tal_check(ctx, NULL));
+	ok1(tal_check(str, NULL));
+	split = strsplit(ctx, take(str), " ", STR_EMPTY_OK);
+	ok1(tal_parent(split) == ctx);
+	ok1(!strcmp(split[0], "hello"));
+	ok1(!strcmp(split[1], "world"));
+	ok1(split[2] == NULL);
+	ok1(tal_check(split, NULL));
+	ok1(tal_check(ctx, NULL));
+	tal_free(split);
+	/* Previous free should get rid of str */
+	ok1(!tal_first(ctx));
+
+	/* strsplit take delims */
+	str = tal_strdup(ctx, " ");
+	split = strsplit(ctx, "hello world", take(str), STR_EMPTY_OK);
+	ok1(tal_parent(split) == ctx);
+	ok1(!strcmp(split[0], "hello"));
+	ok1(!strcmp(split[1], "world"));
+	ok1(split[2] == NULL);
+	ok1(tal_check(split, NULL));
+	ok1(tal_check(ctx, NULL));
+	tal_free(split);
+	/* str is gone... */
+	ok1(!tal_first(ctx));
+
+	/* strsplit takes both. */
+	split = strsplit(ctx, take(tal_strdup(NULL, "hello world")),
+			 take(tal_strdup(NULL, " ")), STR_EMPTY_OK);
+	ok1(tal_parent(split) == ctx);
+	ok1(!strcmp(split[0], "hello"));
+	ok1(!strcmp(split[1], "world"));
+	ok1(split[2] == NULL);
+	ok1(tal_check(split, NULL));
+	ok1(tal_check(ctx, NULL));
+	tal_free(split);
+	/* temp allocs are gone... */
+	ok1(!tal_first(ctx));
+
+	/* strjoin passthrough taken NULLs OK. */
+	ok1(strjoin(ctx, take(NULL), "", STR_TRAIL) == NULL);
+	ok1(strjoin(ctx, take(NULL), "", STR_NO_TRAIL) == NULL);
+	ok1(strjoin(ctx, split, take(NULL), STR_TRAIL) == NULL);
+	ok1(strjoin(ctx, split, take(NULL), STR_NO_TRAIL) == NULL);
+
+	/* strjoin take strings[] */
+	split = strsplit(ctx, "hello world", " ", STR_EMPTY_OK);
+	str = strjoin(ctx, take(split), " there ", STR_NO_TRAIL);
+	ok1(!strcmp(str, "hello there world"));
+	ok1(tal_parent(str) == ctx);
+	/* split is gone... */
+	ok1(tal_first(ctx) == str && !tal_next(ctx, str));
+	tal_free(str);
+	ok1(!tal_first(ctx));
+
+	/* strjoin take delim */
+	split = strsplit(ctx, "hello world", " ", STR_EMPTY_OK);
+	str = strjoin(ctx, split, take(tal_strdup(ctx, " there ")),
+		      STR_NO_TRAIL);
+	ok1(!strcmp(str, "hello there world"));
+	ok1(tal_parent(str) == ctx);
+	tal_free(split);
+	/* tmp alloc is gone, str is only remainder. */
+	ok1(tal_first(ctx) == str && !tal_next(ctx, str));
+	tal_free(str);
+	ok1(!tal_first(ctx));
+
+	/* strjoin take both. */
+	str = strjoin(ctx, take(strsplit(ctx, "hello world", " ",
+					   STR_EMPTY_OK)),
+		      take(tal_strdup(ctx, " there ")), STR_NO_TRAIL);
+	ok1(!strcmp(str, "hello there world"));
+	ok1(tal_parent(str) == ctx);
+	/* tmp allocs are gone, str is only remainder. */
+	ok1(tal_first(ctx) == str && !tal_next(ctx, str));
+	tal_free(str);
+	ok1(!tal_first(ctx));
+	tal_free(ctx);
+
 	return exit_status();
 }