X-Git-Url: http://mj.ucw.cz/gitweb/?a=blobdiff_plain;f=ucw-json%2Fparse.c;h=df99ea405196efad68538da24d28ad15073170cc;hb=HEAD;hp=181c6afb5bcdff274600080d6233bb358df4ca24;hpb=5b7aeae4b33b686e52304127d1648f7b71c9b405;p=libucw.git

diff --git a/ucw-json/parse.c b/ucw-json/parse.c
index 181c6afb..df99ea40 100644
--- a/ucw-json/parse.c
+++ b/ucw-json/parse.c
@@ -7,11 +7,10 @@
  *	of the GNU Lesser General Public License.
  */
 
-#undef LOCAL_DEBUG
-
 #include <ucw/lib.h>
-#include <ucw/trans.h>
+#include <ucw/fastbuf.h>
 #include <ucw/ff-unicode.h>
+#include <ucw/trans.h>
 #include <ucw/unicode.h>
 #include <ucw-json/json.h>
 
@@ -22,17 +21,15 @@ void json_set_input(struct json_context *js, struct fastbuf *in)
 {
   js->in_fb = in;
   js->in_line = 1;
+  js->in_column = 0;
   js->next_char = -1;
   js->next_token = NULL;
   js->in_eof = 0;
-  if (!js->trivial_token)
-    js->trivial_token = json_new_node(js, JSON_INVALID);
 }
 
-// FIXME: Report column as well as line?
 static void NONRET json_parse_error(struct json_context *js, const char *msg)
 {
-  trans_throw("ucw.js.parse", js, "%s at line %u", msg, js->in_line);
+  trans_throw("ucw.json.parse", js, "%s at line %u:%u", msg, js->in_line, js->in_column);
 }
 
 static int json_get_char(struct json_context *js)
@@ -43,9 +40,9 @@ static int json_get_char(struct json_context *js)
       if (c == -2)
 	json_parse_error(js, "Malformed UTF-8 character");
       js->in_eof = 1;
-      // FIXME: Reject alternative sequences
       return c;
     }
+  js->in_column++;
   return c;
 }
 
@@ -96,6 +93,7 @@ static struct json_node *json_parse_number(struct json_context *js, int c)
   if (c == '.')
     {
       p = mp_append_char(js->pool, p, c);
+      c = json_get_char(js);
       if (!(c >= '0' && c <= '9'))
 	json_parse_error(js, "Malformed number: no digits after decimal point");
       while (c >= '0' && c <= '9')
@@ -138,28 +136,30 @@ static struct json_node *json_parse_number(struct json_context *js, int c)
 
 static struct json_node *json_parse_name(struct json_context *js, int c)
 {
-  mp_push(js->pool);
-  char *p = mp_start_noalign(js->pool, 0);
+  char name[16];
+  uint i = 0;
 
   while (c >= 'a' && c <= 'z')
     {
-      p = mp_append_char(js->pool, p, c);
+      if (i < sizeof(name) - 1)
+	name[i++] = c;
       c = json_get_char(js);
     }
+  if (i >= sizeof(name) - 1)
+    json_parse_error(js, "Invalid literal name");
+  name[i] = 0;
   json_unget_char(js, c);
 
-  p = mp_end_string(js->pool, p);
   struct json_node *n;
-  if (!strcmp(p, "null"))
+  if (!strcmp(name, "null"))
     n = json_new_null(js);
-  else if (!strcmp(p, "false"))
+  else if (!strcmp(name, "false"))
     n = json_new_bool(js, 0);
-  else if (!strcmp(p, "true"))
+  else if (!strcmp(name, "true"))
     n = json_new_bool(js, 1);
   else
     json_parse_error(js, "Invalid literal name");
 
-  mp_pop(js->pool);
   return n;
 }
 
@@ -203,7 +203,7 @@ static struct json_node *json_parse_string(struct json_context *js, int c)
 	  else
 	    json_parse_error(js, "Invalid private-use character in string");
 	}
-      if (unlikely(c > 0xf0000))
+      if (unlikely(c >= 0xf0000))
 	{
 	  if (c > 0x10ffff)
 	    json_parse_error(js, "Invalid non-Unicode character in string");
@@ -249,7 +249,7 @@ static struct json_node *json_parse_string(struct json_context *js, int c)
 			  y = json_parse_hex4(js);
 			if (!(y >= 0xdc00 && y < 0xe000))
 			  json_parse_error(js, "Escaped high surrogate codepoint must be followed by a low surrogate codepoint");
-			c = 0x10000 | ((x & 0x03ff) << 10) | (y & 0x03ff);
+			c = 0x10000 + ((x & 0x03ff) << 10) | (y & 0x03ff);
 			if (c > 0xf0000)
 			  json_parse_error(js, "Invalid escaped private-use character");
 		      }
@@ -261,6 +261,8 @@ static struct json_node *json_parse_string(struct json_context *js, int c)
 		    else
 		      json_parse_error(js, "Invalid escaped private-use character");
 		  }
+		else
+		  c = x;
 		break;
 	      }
 	    default:
@@ -275,7 +277,7 @@ static struct json_node *json_parse_string(struct json_context *js, int c)
   return json_new_string_ref(js, p);
 }
 
-struct json_node *json_peek_token(struct json_context *js)
+static struct json_node *json_read_token(struct json_context *js)
 {
   if (unlikely(js->in_eof))
     return json_triv_token(js, JSON_EOF);
@@ -289,7 +291,10 @@ struct json_node *json_peek_token(struct json_context *js)
   while (c == 0x20 || c == 0x09 || c == 0x0a || c == 0x0d)
     {
       if (c == 0x0a)
-	js->in_line++;
+	{
+	  js->in_line++;
+	  js->in_column = 0;
+	}
       c = json_get_char(js);
     }
   if (c < 0)
@@ -318,21 +323,34 @@ struct json_node *json_peek_token(struct json_context *js)
       return json_triv_token(js, JSON_NAME_SEP);
     case ',':
       return json_triv_token(js, JSON_VALUE_SEP);
+    case '.':
+      json_parse_error(js, "Numbers must start with a digit");
+    case 0xfeff:
+      json_parse_error(js, "Misplaced byte-order mark, complain in Redmond");
     default:
       json_parse_error(js, "Invalid character");
     }
 }
 
-struct json_node *json_next_token(struct json_context *js)
+struct json_node *json_peek_token(struct json_context *js)
 {
   if (!js->next_token)
-    json_peek_token(js);
+    js->next_token = json_read_token(js);
+  return js->next_token;
+}
+
+struct json_node *json_next_token(struct json_context *js)
+{
   struct json_node *t = js->next_token;
-  js->next_token = NULL;
-  return t;
+  if (t)
+    {
+      js->next_token = NULL;
+      return t;
+    }
+  return json_read_token(js);
 }
 
-struct json_node *json_next_object(struct json_context *js)
+struct json_node *json_next_value(struct json_context *js)
 {
   struct json_node *t = json_next_token(js);
 
@@ -356,7 +374,7 @@ struct json_node *json_next_object(struct json_context *js)
 	  json_next_token(js);
 	else for (;;)
 	  {
-	    struct json_node *v = json_next_object(js);
+	    struct json_node *v = json_next_value(js);
 	    if (!v)
 	      json_parse_error(js, "Unterminated array");
 	    json_array_append(a, v);
@@ -365,7 +383,7 @@ struct json_node *json_next_object(struct json_context *js)
 	    if (t->type == JSON_END_ARRAY)
 	      break;
 	    if (t->type != JSON_VALUE_SEP)
-	      json_parse_error(js, "Comma expected");
+	      json_parse_error(js, "Comma or right bracket expected");
 	  }
 	return a;
       }
@@ -378,7 +396,7 @@ struct json_node *json_next_object(struct json_context *js)
 	  json_next_token(js);
 	else for (;;)
 	  {
-	    struct json_node *k = json_next_object(js);
+	    struct json_node *k = json_next_value(js);
 	    if (!k)
 	      json_parse_error(js, "Unterminated object");
 	    if (k->type != JSON_STRING)
@@ -388,7 +406,7 @@ struct json_node *json_next_object(struct json_context *js)
 	    if (t->type != JSON_NAME_SEP)
 	      json_parse_error(js, "Colon expected");
 
-	    struct json_node *v = json_next_object(js);
+	    struct json_node *v = json_next_value(js);
 	    if (!v)
 	      json_parse_error(js, "Unterminated object");
 	    if (json_object_get(o, k->string))		// FIXME: Optimize
@@ -422,7 +440,7 @@ struct json_node *json_parse(struct json_context *js, struct fastbuf *fb)
 {
   json_set_input(js, fb);
 
-  struct json_node *n = json_next_object(js);
+  struct json_node *n = json_next_value(js);
   if (!n)
     json_parse_error(js, "Empty input");