mirror of
https://github.com/ElementsProject/lightning.git
synced 2024-11-19 09:54:16 +01:00
305795b01e
It's the only user of them, and it's going to get optimized. Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> gossip.pydiff --git a/common/test/run-json.c b/common/test/run-json.c index 956fdda35..db52d6b01 100644
163 lines
3.1 KiB
C
163 lines
3.1 KiB
C
#include <lightningd/json_escaped.h>
|
|
#include <stdio.h>
|
|
|
|
struct json_escaped *json_escaped_string_(const tal_t *ctx,
|
|
const void *bytes, size_t len)
|
|
{
|
|
struct json_escaped *esc;
|
|
|
|
esc = (void *)tal_arr_label(ctx, char, len + 1,
|
|
TAL_LABEL(struct json_escaped, ""));
|
|
memcpy(esc->s, bytes, len);
|
|
esc->s[len] = '\0';
|
|
return esc;
|
|
}
|
|
|
|
struct json_escaped *json_to_escaped_string(const tal_t *ctx,
|
|
const char *buffer,
|
|
const jsmntok_t *tok)
|
|
{
|
|
if (tok->type != JSMN_STRING)
|
|
return NULL;
|
|
/* jsmn always gives us ~ well-formed strings. */
|
|
return json_escaped_string_(ctx, buffer + tok->start,
|
|
tok->end - tok->start);
|
|
}
|
|
|
|
bool json_escaped_eq(const struct json_escaped *a,
|
|
const struct json_escaped *b)
|
|
{
|
|
return streq(a->s, b->s);
|
|
}
|
|
|
|
static struct json_escaped *escape(const tal_t *ctx,
|
|
const char *str TAKES,
|
|
bool partial)
|
|
{
|
|
struct json_escaped *esc;
|
|
size_t i, n;
|
|
|
|
/* Worst case: all \uXXXX */
|
|
esc = (struct json_escaped *)tal_arr(ctx, char, strlen(str) * 6 + 1);
|
|
|
|
for (i = n = 0; str[i]; i++, n++) {
|
|
char escape = 0;
|
|
switch (str[i]) {
|
|
case '\n':
|
|
escape = 'n';
|
|
break;
|
|
case '\b':
|
|
escape = 'b';
|
|
break;
|
|
case '\f':
|
|
escape = 'f';
|
|
break;
|
|
case '\t':
|
|
escape = 't';
|
|
break;
|
|
case '\r':
|
|
escape = 'r';
|
|
break;
|
|
case '\\':
|
|
if (partial) {
|
|
/* Don't double-escape standard escapes. */
|
|
if (str[i+1] == 'n'
|
|
|| str[i+1] == 'b'
|
|
|| str[i+1] == 'f'
|
|
|| str[i+1] == 't'
|
|
|| str[i+1] == 'r'
|
|
|| str[i+1] == '/'
|
|
|| str[i+1] == '\\'
|
|
|| str[i+1] == '"') {
|
|
escape = str[i+1];
|
|
i++;
|
|
break;
|
|
}
|
|
if (str[i+1] == 'u'
|
|
&& cisxdigit(str[i+2])
|
|
&& cisxdigit(str[i+3])
|
|
&& cisxdigit(str[i+4])
|
|
&& cisxdigit(str[i+5])) {
|
|
memcpy(esc->s + n, str + i, 6);
|
|
n += 5;
|
|
i += 5;
|
|
continue;
|
|
}
|
|
} /* fall thru */
|
|
case '"':
|
|
escape = str[i];
|
|
break;
|
|
default:
|
|
if ((unsigned)str[i] < ' ' || str[i] == 127) {
|
|
snprintf(esc->s + n, 7, "\\u%04X", str[i]);
|
|
n += 5;
|
|
continue;
|
|
}
|
|
}
|
|
if (escape) {
|
|
esc->s[n++] = '\\';
|
|
esc->s[n] = escape;
|
|
} else
|
|
esc->s[n] = str[i];
|
|
}
|
|
|
|
esc->s[n] = '\0';
|
|
if (taken(str))
|
|
tal_free(str);
|
|
return esc;
|
|
}
|
|
|
|
struct json_escaped *json_partial_escape(const tal_t *ctx, const char *str TAKES)
|
|
{
|
|
return escape(ctx, str, true);
|
|
}
|
|
|
|
struct json_escaped *json_escape(const tal_t *ctx, const char *str TAKES)
|
|
{
|
|
return escape(ctx, str, false);
|
|
}
|
|
|
|
/* By policy, we don't handle \u. Use UTF-8. */
|
|
const char *json_escaped_unescape(const tal_t *ctx,
|
|
const struct json_escaped *esc)
|
|
{
|
|
char *unesc = tal_arr(ctx, char, strlen(esc->s) + 1);
|
|
size_t i, n;
|
|
|
|
for (i = n = 0; esc->s[i]; i++, n++) {
|
|
if (esc->s[i] != '\\') {
|
|
unesc[n] = esc->s[i];
|
|
continue;
|
|
}
|
|
|
|
i++;
|
|
switch (esc->s[i]) {
|
|
case 'n':
|
|
unesc[n] = '\n';
|
|
break;
|
|
case 'b':
|
|
unesc[n] = '\b';
|
|
break;
|
|
case 'f':
|
|
unesc[n] = '\f';
|
|
break;
|
|
case 't':
|
|
unesc[n] = '\t';
|
|
break;
|
|
case 'r':
|
|
unesc[n] = '\r';
|
|
break;
|
|
case '/':
|
|
case '\\':
|
|
case '"':
|
|
unesc[n] = esc->s[i];
|
|
break;
|
|
default:
|
|
return tal_free(unesc);
|
|
}
|
|
}
|
|
|
|
unesc[n] = '\0';
|
|
return unesc;
|
|
}
|