From: Matt Whitlock <ccan@mattwhitlock.name>
Date: Mon, 14 Jul 2025 10:02:58 +0000 (-0400)
Subject: json_escape: add json_escape_unescape_len()
X-Git-Url: https://git.ozlabs.org/?a=commitdiff_plain;h=e43c61c4c760aed7fd4ab6b716b019fc170f280d;p=ccan

json_escape: add json_escape_unescape_len()

Lacking a variant of json_escape_unescape() that takes a sized span,
the common case of unescaping a JSON string that is embedded inside a
larger buffer is forced to make a copy of the escaped string so as to
terminate it with a NUL character. This extra copy can be avoided if
we can pass an explicit length to the unescaping function.

Also, a drive-by fix: shrink the returned unescaped string to fit its
contents, which can save the caller a subsequent call to strlen().
---

diff --git a/ccan/json_escape/json_escape.c b/ccan/json_escape/json_escape.c
index daa14abe..4bbb0032 100644
--- a/ccan/json_escape/json_escape.c
+++ b/ccan/json_escape/json_escape.c
@@ -1,6 +1,7 @@
 /* MIT (BSD) license - see LICENSE file for details */
 #include <ccan/json_escape/json_escape.h>
 #include <stdio.h>
+#include <ccan/tal/str/str.h>
 
 struct json_escape *json_escape_string_(const tal_t *ctx,
 					const void *bytes, size_t len)
@@ -137,19 +138,24 @@ struct json_escape *json_escape_len(const tal_t *ctx, const char *str TAKES,
 }
 
 /* By policy, we don't handle \u.  Use UTF-8. */
-const char *json_escape_unescape(const tal_t *ctx, const struct json_escape *esc)
+static const char *unescape(const tal_t *ctx, const char *esc TAKES, size_t len)
 {
-	char *unesc = tal_arr(ctx, char, strlen(esc->s) + 1);
+	/* Fast path: can steal, and nothing to unescape. */
+	if (is_taken(esc) && !memchr(esc, '\\', len))
+		return tal_strndup(ctx, esc, len);
+
+	char *unesc = tal_arr(ctx, char, len + 1);
 	size_t i, n;
 
-	for (i = n = 0; esc->s[i]; i++, n++) {
-		if (esc->s[i] != '\\') {
-			unesc[n] = esc->s[i];
+	for (i = n = 0; i < len; i++, n++) {
+		if (esc[i] != '\\') {
+			unesc[n] = esc[i];
 			continue;
 		}
 
-		i++;
-		switch (esc->s[i]) {
+		if (++i == len)
+			goto error;
+		switch (esc[i]) {
 		case 'n':
 			unesc[n] = '\n';
 			break;
@@ -168,13 +174,31 @@ const char *json_escape_unescape(const tal_t *ctx, const struct json_escape *esc
 		case '/':
 		case '\\':
 		case '"':
-			unesc[n] = esc->s[i];
+			unesc[n] = esc[i];
 			break;
 		default:
+		error:
+			if (taken(esc))
+				tal_free(esc);
 			return tal_free(unesc);
 		}
 	}
 
 	unesc[n] = '\0';
+	if (!tal_resize(&unesc, n + 1))
+		goto error;
+	if (taken(esc))
+		tal_free(esc);
 	return unesc;
 }
+
+const char *json_escape_unescape(const tal_t *ctx, const struct json_escape *esc)
+{
+	return unescape(ctx, esc->s, strlen(esc->s));
+}
+
+const char *json_escape_unescape_len(const tal_t *ctx,
+				     const char *esc TAKES, size_t len)
+{
+	return unescape(ctx, esc, len);
+}
diff --git a/ccan/json_escape/json_escape.h b/ccan/json_escape/json_escape.h
index 5b33432f..88d1eefc 100644
--- a/ccan/json_escape/json_escape.h
+++ b/ccan/json_escape/json_escape.h
@@ -41,4 +41,8 @@ struct json_escape *json_escape_string_(const tal_t *ctx,
 /* Be very careful here!  Can fail!  Doesn't handle \u: use UTF-8 please. */
 const char *json_escape_unescape(const tal_t *ctx,
 				 const struct json_escape *esc);
+
+/* Be very careful here!  Can fail!  Doesn't handle \u: use UTF-8 please. */
+const char *json_escape_unescape_len(const tal_t *ctx,
+				     const char *esc TAKES, size_t len);
 #endif /* CCAN_JSON_ESCAPE_H */