static guint
json_parse_statement (JsonParser *parser,
JsonScanner *scanner)
{
JsonParserPrivate *priv = parser->priv;
guint token;
token = json_scanner_peek_next_token (scanner);
switch (token)
{
case G_TOKEN_LEFT_CURLY:
JSON_NOTE (PARSER, "Statement is object declaration");
return json_parse_object (parser, scanner, &priv->root);
case G_TOKEN_LEFT_BRACE:
JSON_NOTE (PARSER, "Statement is array declaration");
return json_parse_array (parser, scanner, &priv->root);
/* some web APIs are not only passing the data structures: they are
* also passing an assigment, which makes parsing horribly complicated
* only because web developers are lazy, and writing "var foo = " is
* evidently too much to request from them.
*/
case JSON_TOKEN_VAR:
{
guint next_token;
gchar *name;
JSON_NOTE (PARSER, "Statement is an assignment");
/* swallow the 'var' token... */
token = json_scanner_get_next_token (scanner);
/* ... swallow the variable name... */
next_token = json_scanner_get_next_token (scanner);
if (next_token != G_TOKEN_IDENTIFIER)
{
priv->error_code = JSON_PARSER_ERROR_INVALID_BAREWORD;
return G_TOKEN_IDENTIFIER;
}
name = g_strdup (scanner->value.v_identifier);
/* ... and finally swallow the '=' */
next_token = json_scanner_get_next_token (scanner);
if (next_token != '=')
{
priv->error_code = JSON_PARSER_ERROR_INVALID_BAREWORD;
g_free (name);
return '=';
}
priv->has_assignment = TRUE;
priv->variable_name = name;
token = json_parse_statement (parser, scanner);
/* remove the trailing semi-colon */
next_token = json_scanner_peek_next_token (scanner);
if (next_token == ';')
{
token = json_scanner_get_next_token (scanner);
return G_TOKEN_NONE;
}
return token;
}
break;
case JSON_TOKEN_NULL:
case JSON_TOKEN_TRUE:
case JSON_TOKEN_FALSE:
case '-':
case G_TOKEN_INT:
case G_TOKEN_FLOAT:
case G_TOKEN_STRING:
case G_TOKEN_IDENTIFIER:
JSON_NOTE (PARSER, "Statement is a value");
token = json_scanner_get_next_token (scanner);
return json_parse_value (parser, scanner, token, &priv->root);
default:
JSON_NOTE (PARSER, "Unknown statement");
json_scanner_get_next_token (scanner);
priv->error_code = JSON_PARSER_ERROR_INVALID_BAREWORD;
return G_TOKEN_SYMBOL;
}
}
static guint
json_parse_array (JsonParser *parser,
JsonScanner *scanner,
JsonNode **node)
{
JsonParserPrivate *priv = parser->priv;
JsonNode *old_current;
JsonArray *array;
guint token;
gint idx;
old_current = priv->current_node;
priv->current_node = json_node_init_array (json_node_alloc (), NULL);
array = json_array_new ();
token = json_scanner_get_next_token (scanner);
g_assert (token == G_TOKEN_LEFT_BRACE);
g_signal_emit (parser, parser_signals[ARRAY_START], 0);
idx = 0;
while (token != G_TOKEN_RIGHT_BRACE)
{
guint next_token = json_scanner_peek_next_token (scanner);
JsonNode *element = NULL;
/* parse the element */
switch (next_token)
{
case G_TOKEN_LEFT_BRACE:
JSON_NOTE (PARSER, "Nested array at index %d", idx);
token = json_parse_array (parser, scanner, &element);
break;
case G_TOKEN_LEFT_CURLY:
JSON_NOTE (PARSER, "Nested object at index %d", idx);
token = json_parse_object (parser, scanner, &element);
break;
case G_TOKEN_RIGHT_BRACE:
goto array_done;
default:
token = json_scanner_get_next_token (scanner);
token = json_parse_value (parser, scanner, token, &element);
break;
}
if (token != G_TOKEN_NONE || element == NULL)
{
/* the json_parse_* functions will have set the error code */
json_array_unref (array);
json_node_free (priv->current_node);
priv->current_node = old_current;
return token;
}
next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_COMMA)
{
token = json_scanner_get_next_token (scanner);
next_token = json_scanner_peek_next_token (scanner);
/* look for trailing commas */
if (next_token == G_TOKEN_RIGHT_BRACE)
{
priv->error_code = JSON_PARSER_ERROR_TRAILING_COMMA;
json_array_unref (array);
json_node_free (priv->current_node);
json_node_free (element);
priv->current_node = old_current;
return G_TOKEN_RIGHT_BRACE;
}
}
JSON_NOTE (PARSER, "Array element %d completed", idx + 1);
json_node_set_parent (element, priv->current_node);
json_array_add_element (array, element);
g_signal_emit (parser, parser_signals[ARRAY_ELEMENT], 0,
array,
idx);
token = next_token;
}
array_done:
json_scanner_get_next_token (scanner);
json_node_take_array (priv->current_node, array);
json_node_set_parent (priv->current_node, old_current);
g_signal_emit (parser, parser_signals[ARRAY_END], 0, array);
if (node != NULL && *node == NULL)
*node = priv->current_node;
priv->current_node = old_current;
return G_TOKEN_NONE;
}
static guint
json_parse_object (JsonParser *parser,
JsonScanner *scanner,
JsonNode **node)
{
JsonParserPrivate *priv = parser->priv;
JsonObject *object;
JsonNode *old_current;
guint token;
old_current = priv->current_node;
priv->current_node = json_node_init_object (json_node_alloc (), NULL);
object = json_object_new ();
token = json_scanner_get_next_token (scanner);
g_assert (token == G_TOKEN_LEFT_CURLY);
g_signal_emit (parser, parser_signals[OBJECT_START], 0);
while (token != G_TOKEN_RIGHT_CURLY)
{
guint next_token = json_scanner_peek_next_token (scanner);
JsonNode *member = NULL;
gchar *name;
/* we need to abort here because empty objects do not
* have member names
*/
if (next_token == G_TOKEN_RIGHT_CURLY)
break;
/* parse the member's name */
if (next_token != G_TOKEN_STRING)
{
JSON_NOTE (PARSER, "Missing object member name");
priv->error_code = JSON_PARSER_ERROR_INVALID_BAREWORD;
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_STRING;
}
/* member name */
token = json_scanner_get_next_token (scanner);
name = g_strdup (scanner->value.v_string);
if (name == NULL || *name == '\0')
{
JSON_NOTE (PARSER, "Empty object member name");
priv->error_code = JSON_PARSER_ERROR_EMPTY_MEMBER_NAME;
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_STRING;
}
JSON_NOTE (PARSER, "Object member '%s'", name);
/* a colon separates names from values */
next_token = json_scanner_peek_next_token (scanner);
if (next_token != ':')
{
JSON_NOTE (PARSER, "Missing object member name separator");
priv->error_code = JSON_PARSER_ERROR_MISSING_COLON;
g_free (name);
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return ':';
}
/* we swallow the ':' */
token = json_scanner_get_next_token (scanner);
g_assert (token == ':');
next_token = json_scanner_peek_next_token (scanner);
/* parse the member's value */
switch (next_token)
{
case G_TOKEN_LEFT_BRACE:
JSON_NOTE (PARSER, "Nested array at member %s", name);
token = json_parse_array (parser, scanner, &member);
break;
case G_TOKEN_LEFT_CURLY:
JSON_NOTE (PARSER, "Nested object at member %s", name);
token = json_parse_object (parser, scanner, &member);
break;
default:
/* once a member name is defined we need a value */
token = json_scanner_get_next_token (scanner);
token = json_parse_value (parser, scanner, token, &member);
break;
}
if (token != G_TOKEN_NONE || member == NULL)
{
/* the json_parse_* functions will have set the error code */
g_free (name);
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return token;
}
next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_COMMA)
{
token = json_scanner_get_next_token (scanner);
next_token = json_scanner_peek_next_token (scanner);
/* look for trailing commas */
if (next_token == G_TOKEN_RIGHT_CURLY)
{
priv->error_code = JSON_PARSER_ERROR_TRAILING_COMMA;
json_object_unref (object);
json_node_free (member);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_RIGHT_BRACE;
}
}
else if (next_token == G_TOKEN_STRING)
{
priv->error_code = JSON_PARSER_ERROR_MISSING_COMMA;
json_object_unref (object);
json_node_free (member);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_COMMA;
}
JSON_NOTE (PARSER, "Object member '%s' completed", name);
json_node_set_parent (member, priv->current_node);
json_object_set_member (object, name, member);
g_signal_emit (parser, parser_signals[OBJECT_MEMBER], 0,
object,
name);
g_free (name);
token = next_token;
}
json_scanner_get_next_token (scanner);
json_node_take_object (priv->current_node, object);
json_node_set_parent (priv->current_node, old_current);
g_signal_emit (parser, parser_signals[OBJECT_END], 0, object);
if (node != NULL && *node == NULL)
*node = priv->current_node;
priv->current_node = old_current;
return G_TOKEN_NONE;
}
static guint
json_parse_value (JsonParser *parser,
JsonScanner *scanner,
guint token,
JsonNode **node)
{
JsonParserPrivate *priv = parser->priv;
JsonNode *current_node = priv->current_node;
gboolean is_negative = FALSE;
if (token == '-')
{
guint next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_INT ||
next_token == G_TOKEN_FLOAT)
{
is_negative = TRUE;
token = json_scanner_get_next_token (scanner);
}
else
return G_TOKEN_INT;
}
switch (token)
{
case G_TOKEN_INT:
JSON_NOTE (PARSER, "abs(node): %" G_GINT64_FORMAT " (sign: %s)",
scanner->value.v_int64,
is_negative ? "negative" : "positive");
*node = json_node_init_int (json_node_alloc (),
is_negative ? scanner->value.v_int64 * -1
: scanner->value.v_int64);
break;
case G_TOKEN_FLOAT:
JSON_NOTE (PARSER, "abs(node): %.6f (sign: %s)",
scanner->value.v_float,
is_negative ? "negative" : "positive");
*node = json_node_init_double (json_node_alloc (),
is_negative ? scanner->value.v_float * -1.0
: scanner->value.v_float);
break;
case G_TOKEN_STRING:
JSON_NOTE (PARSER, "node: '%s'",
scanner->value.v_string);
*node = json_node_init_string (json_node_alloc (), scanner->value.v_string);
break;
case JSON_TOKEN_TRUE:
case JSON_TOKEN_FALSE:
JSON_NOTE (PARSER, "node: '%s'",
JSON_TOKEN_TRUE ? "<true>" : "<false>");
*node = json_node_init_boolean (json_node_alloc (), token == JSON_TOKEN_TRUE ? TRUE : FALSE);
break;
case JSON_TOKEN_NULL:
JSON_NOTE (PARSER, "node: <null>");
*node = json_node_init_null (json_node_alloc ());
break;
case G_TOKEN_IDENTIFIER:
JSON_NOTE (PARSER, "node: identifier '%s'", scanner->value.v_identifier);
priv->error_code = JSON_PARSER_ERROR_INVALID_BAREWORD;
*node = NULL;
return G_TOKEN_SYMBOL;
default:
{
JsonNodeType cur_type;
*node = NULL;
JSON_NOTE (PARSER, "node: invalid token");
cur_type = json_node_get_node_type (current_node);
if (cur_type == JSON_NODE_ARRAY)
{
priv->error_code = JSON_PARSER_ERROR_PARSE;
return G_TOKEN_RIGHT_BRACE;
}
else if (cur_type == JSON_NODE_OBJECT)
{
priv->error_code = JSON_PARSER_ERROR_PARSE;
return G_TOKEN_RIGHT_CURLY;
}
else
{
priv->error_code = JSON_PARSER_ERROR_INVALID_BAREWORD;
return G_TOKEN_SYMBOL;
}
}
break;
}
return G_TOKEN_NONE;
}
static guint
json_parse_statement (JsonParser *parser,
JsonScanner *scanner)
{
JsonParserPrivate *priv = parser->priv;
guint token;
token = json_scanner_peek_next_token (scanner);
switch (token)
{
case G_TOKEN_LEFT_CURLY:
return json_parse_object (parser, scanner, &priv->root);
case G_TOKEN_LEFT_BRACE:
return json_parse_array (parser, scanner, &priv->root);
/* some web APIs are not only passing the data structures: they are
* also passing an assigment, which makes parsing horribly complicated
* only because web developers are lazy, and writing "var foo = " is
* evidently too much to request from them.
*/
case JSON_TOKEN_VAR:
{
guint next_token;
gchar *name;
/* swallow the 'var' token... */
token = json_scanner_get_next_token (scanner);
/* ... swallow the variable name... */
next_token = json_scanner_get_next_token (scanner);
if (next_token != G_TOKEN_IDENTIFIER)
return G_TOKEN_IDENTIFIER;
name = g_strdup (scanner->value.v_identifier);
/* ... and finally swallow the '=' */
next_token = json_scanner_get_next_token (scanner);
if (next_token != '=')
return '=';
priv->has_assignment = TRUE;
priv->variable_name = name;
token = json_parse_statement (parser, scanner);
/* remove the trailing semi-colon */
next_token = json_scanner_peek_next_token (scanner);
if (next_token == ';')
{
token = json_scanner_get_next_token (scanner);
return G_TOKEN_NONE;
}
return token;
}
break;
case JSON_TOKEN_NULL:
priv->root = priv->current_node = json_node_new (JSON_NODE_NULL);
json_scanner_get_next_token (scanner);
return G_TOKEN_NONE;
case JSON_TOKEN_TRUE:
case JSON_TOKEN_FALSE:
priv->root = priv->current_node = json_node_new (JSON_NODE_VALUE);
json_node_set_boolean (priv->current_node,
token == JSON_TOKEN_TRUE ? TRUE : FALSE);
json_scanner_get_next_token (scanner);
return G_TOKEN_NONE;
case '-':
{
guint next_token;
token = json_scanner_get_next_token (scanner);
next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_INT || next_token == G_TOKEN_FLOAT)
{
priv->root = priv->current_node = json_node_new (JSON_NODE_VALUE);
token = json_scanner_get_next_token (scanner);
switch (token)
{
case G_TOKEN_INT:
json_node_set_int (priv->current_node,
scanner->value.v_int64 * -1);
break;
case G_TOKEN_FLOAT:
json_node_set_double (priv->current_node,
scanner->value.v_float * -1.0);
break;
default:
return G_TOKEN_INT;
}
json_scanner_get_next_token (scanner);
return G_TOKEN_NONE;
}
else
return G_TOKEN_INT;
}
break;
case G_TOKEN_INT:
case G_TOKEN_FLOAT:
case G_TOKEN_STRING:
token = json_scanner_get_next_token (scanner);
return json_parse_value (parser, scanner, token, &priv->root);
default:
json_scanner_get_next_token (scanner);
return G_TOKEN_SYMBOL;
}
}
static guint
json_parse_object (JsonParser *parser,
JsonScanner *scanner,
JsonNode **node)
{
JsonParserPrivate *priv = parser->priv;
JsonObject *object;
JsonNode *old_current;
guint token;
old_current = priv->current_node;
priv->current_node = json_node_new (JSON_NODE_OBJECT);
object = json_object_new ();
token = json_scanner_get_next_token (scanner);
g_assert (token == G_TOKEN_LEFT_CURLY);
g_signal_emit (parser, parser_signals[OBJECT_START], 0);
while (token != G_TOKEN_RIGHT_CURLY)
{
guint next_token = json_scanner_peek_next_token (scanner);
JsonNode *member = NULL;
gchar *name;
/* we need to abort here because empty objects do not
* have member names
*/
if (next_token == G_TOKEN_RIGHT_CURLY)
break;
/* parse the member's name */
if (next_token != G_TOKEN_STRING)
{
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_STRING;
}
/* member name */
token = json_scanner_get_next_token (scanner);
name = g_strdup (scanner->value.v_string);
/* a colon separates names from values */
next_token = json_scanner_peek_next_token (scanner);
if (next_token != ':')
{
g_free (name);
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return ':';
}
/* we swallow the ':' */
token = json_scanner_get_next_token (scanner);
g_assert (token == ':');
next_token = json_scanner_peek_next_token (scanner);
/* parse the member's value */
switch (next_token)
{
case G_TOKEN_LEFT_BRACE:
token = json_parse_array (parser, scanner, &member);
break;
case G_TOKEN_LEFT_CURLY:
token = json_parse_object (parser, scanner, &member);
break;
case G_TOKEN_INT:
case G_TOKEN_FLOAT:
case G_TOKEN_STRING:
case '-':
case JSON_TOKEN_TRUE:
case JSON_TOKEN_FALSE:
case JSON_TOKEN_NULL:
token = json_scanner_get_next_token (scanner);
token = json_parse_value (parser, scanner, token, &member);
break;
default:
/* once a member name is defined we need a value */
token = G_TOKEN_SYMBOL;
break;
}
if (token != G_TOKEN_NONE || member == NULL)
{
/* the json_parse_* functions will have set the error code */
g_free (name);
json_object_unref (object);
json_node_free (priv->current_node);
priv->current_node = old_current;
return token;
}
next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_COMMA)
{
token = json_scanner_get_next_token (scanner);
next_token = json_scanner_peek_next_token (scanner);
/* look for trailing commas */
if (next_token == G_TOKEN_RIGHT_CURLY)
{
json_object_unref (object);
json_node_free (member);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_RIGHT_BRACE;
}
}
else if (next_token == G_TOKEN_STRING)
{
json_object_unref (object);
json_node_free (member);
json_node_free (priv->current_node);
priv->current_node = old_current;
return G_TOKEN_COMMA;
}
json_node_set_parent (member, priv->current_node);
json_object_set_member (object, name, member);
g_signal_emit (parser, parser_signals[OBJECT_MEMBER], 0,
object,
name);
g_free (name);
token = next_token;
}
json_scanner_get_next_token (scanner);
json_node_take_object (priv->current_node, object);
json_node_set_parent (priv->current_node, old_current);
g_signal_emit (parser, parser_signals[OBJECT_END], 0, object);
if (node != NULL && *node == NULL)
*node = priv->current_node;
priv->current_node = old_current;
return G_TOKEN_NONE;
}
static guint
json_parse_array (JsonParser *parser,
JsonScanner *scanner,
JsonNode **node)
{
JsonParserPrivate *priv = parser->priv;
JsonNode *old_current;
JsonArray *array;
guint token;
gint idx;
old_current = priv->current_node;
priv->current_node = json_node_new (JSON_NODE_ARRAY);
array = json_array_new ();
token = json_scanner_get_next_token (scanner);
g_assert (token == G_TOKEN_LEFT_BRACE);
g_signal_emit (parser, parser_signals[ARRAY_START], 0);
idx = 0;
while (token != G_TOKEN_RIGHT_BRACE)
{
guint next_token = json_scanner_peek_next_token (scanner);
JsonNode *element = NULL;
/* parse the element */
switch (next_token)
{
case G_TOKEN_LEFT_BRACE:
token = json_parse_array (parser, scanner, &element);
break;
case G_TOKEN_LEFT_CURLY:
token = json_parse_object (parser, scanner, &element);
break;
case G_TOKEN_INT:
case G_TOKEN_FLOAT:
case G_TOKEN_STRING:
case '-':
case JSON_TOKEN_TRUE:
case JSON_TOKEN_FALSE:
case JSON_TOKEN_NULL:
token = json_scanner_get_next_token (scanner);
token = json_parse_value (parser, scanner, token, &element);
break;
case G_TOKEN_RIGHT_BRACE:
goto array_done;
default:
if (next_token != G_TOKEN_RIGHT_BRACE)
token = G_TOKEN_RIGHT_BRACE;
break;
}
if (token != G_TOKEN_NONE || element == NULL)
{
/* the json_parse_* functions will have set the error code */
json_array_unref (array);
json_node_free (priv->current_node);
priv->current_node = old_current;
return token;
}
next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_COMMA)
{
token = json_scanner_get_next_token (scanner);
next_token = json_scanner_peek_next_token (scanner);
/* look for trailing commas */
if (next_token == G_TOKEN_RIGHT_BRACE)
{
json_array_unref (array);
json_node_free (priv->current_node);
json_node_free (element);
priv->current_node = old_current;
return G_TOKEN_RIGHT_BRACE;
}
}
json_node_set_parent (element, priv->current_node);
json_array_add_element (array, element);
g_signal_emit (parser, parser_signals[ARRAY_ELEMENT], 0,
array,
idx);
token = next_token;
}
array_done:
json_scanner_get_next_token (scanner);
json_node_take_array (priv->current_node, array);
json_node_set_parent (priv->current_node, old_current);
g_signal_emit (parser, parser_signals[ARRAY_END], 0, array);
if (node != NULL && *node == NULL)
*node = priv->current_node;
priv->current_node = old_current;
return G_TOKEN_NONE;
}
static guint
json_parse_value (JsonParser *parser,
JsonScanner *scanner,
guint token,
JsonNode **node)
{
JsonNode *current_node = parser->priv->current_node;
gboolean is_negative = FALSE;
if (token == '-')
{
guint next_token = json_scanner_peek_next_token (scanner);
if (next_token == G_TOKEN_INT ||
next_token == G_TOKEN_FLOAT)
{
is_negative = TRUE;
token = json_scanner_get_next_token (scanner);
}
else
return G_TOKEN_INT;
}
switch (token)
{
case G_TOKEN_INT:
*node = json_node_new (JSON_NODE_VALUE);
json_node_set_int (*node, is_negative ? scanner->value.v_int64 * -1
: scanner->value.v_int64);
break;
case G_TOKEN_FLOAT:
*node = json_node_new (JSON_NODE_VALUE);
json_node_set_double (*node, is_negative ? scanner->value.v_float * -1.0
: scanner->value.v_float);
break;
case G_TOKEN_STRING:
*node = json_node_new (JSON_NODE_VALUE);
json_node_set_string (*node, scanner->value.v_string);
break;
case JSON_TOKEN_TRUE:
case JSON_TOKEN_FALSE:
*node = json_node_new (JSON_NODE_VALUE);
json_node_set_boolean (*node, token == JSON_TOKEN_TRUE ? TRUE : FALSE);
break;
case JSON_TOKEN_NULL:
*node = json_node_new (JSON_NODE_NULL);
break;
default:
{
JsonNodeType cur_type;
*node = NULL;
cur_type = json_node_get_node_type (current_node);
if (cur_type == JSON_NODE_ARRAY)
return G_TOKEN_RIGHT_BRACE;
else if (cur_type == JSON_NODE_OBJECT)
return G_TOKEN_RIGHT_CURLY;
else
return G_TOKEN_SYMBOL;
}
}
return G_TOKEN_NONE;
}
