Better error recovery for unwritable nodes

Author: kddnewton

config.yml

@@ -104,6 +104,14 @@ errors:
  - EXPECT_STRING_CONTENT
  - EXPECT_WHEN_DELIMITER
  - EXPRESSION_BARE_HASH
+ - EXPRESSION_NOT_WRITABLE
+ - EXPRESSION_NOT_WRITABLE_ENCODING
+ - EXPRESSION_NOT_WRITABLE_FALSE
+ - EXPRESSION_NOT_WRITABLE_FILE
+ - EXPRESSION_NOT_WRITABLE_LINE
+ - EXPRESSION_NOT_WRITABLE_NIL
+ - EXPRESSION_NOT_WRITABLE_SELF
+ - EXPRESSION_NOT_WRITABLE_TRUE
  - FLOAT_PARSE
  - FOR_COLLECTION
  - FOR_IN

src/prism.c

@@ -12945,6 +12945,32 @@ parse_write(pm_parser_t *parser, pm_node_t *target, pm_token_t *operator, pm_nod
  }
 }
 
+/**
+ * Certain expressions are not writable, but in order to provide a better
+ * experience we give a specific error message. In order to maintain as much
+ * information in the tree as possible, we replace them with local variable
+ * writes.
+ */
+static pm_node_t *
+parse_unwriteable_write(pm_parser_t *parser, pm_node_t *target, const pm_token_t *equals, pm_node_t *value) {
+ switch (PM_NODE_TYPE(target)) {
+ case PM_SOURCE_ENCODING_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_ENCODING); break;
+ case PM_FALSE_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_FALSE); break;
+ case PM_SOURCE_FILE_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_FILE); break;
+ case PM_SOURCE_LINE_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_LINE); break;
+ case PM_NIL_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_NIL); break;
+ case PM_SELF_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_SELF); break;
+ case PM_TRUE_NODE: pm_parser_err_token(parser, equals, PM_ERR_EXPRESSION_NOT_WRITABLE_TRUE); break;
+ default: break;
+ }
+
+ pm_constant_id_t name = pm_parser_constant_id_location(parser, target->location.start, target->location.end);
+ pm_local_variable_write_node_t *result = pm_local_variable_write_node_create(parser, name, 0, value, &target->location, equals);
+
+ pm_node_destroy(parser, target);
+ return (pm_node_t *) result;
+}
+
 /**
  * Parse a list of targets for assignment. This is used in the case of a for
  * loop or a multi-assignment. For example, in the following code:
@@ -19357,13 +19383,25 @@ parse_expression_infix(pm_parser_t *parser, pm_node_t *node, pm_binding_power_t
  pm_node_t *value = parse_assignment_values(parser, previous_binding_power, PM_BINDING_POWER_MULTI_ASSIGNMENT + 1, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_EQUAL);
  return parse_write(parser, (pm_node_t *) multi_target, &token, value);
  }
+ case PM_SOURCE_ENCODING_NODE:
+ case PM_FALSE_NODE:
+ case PM_SOURCE_FILE_NODE:
+ case PM_SOURCE_LINE_NODE:
+ case PM_NIL_NODE:
+ case PM_SELF_NODE:
+ case PM_TRUE_NODE: {
+ // In these special cases, we have specific error messages
+ // and we will replace them with local variable writes.
+ parser_lex(parser);
+ pm_node_t *value = parse_assignment_values(parser, previous_binding_power, binding_power, accepts_command_call, PM_ERR_EXPECT_EXPRESSION_AFTER_EQUAL);
+ return parse_unwriteable_write(parser, node, &token, value);
+ }
  default:
+ // In this case we have an = sign, but we don't know what
+ // it's for. We need to treat it as an error. We'll mark it
+ // as an error and skip past it.
  parser_lex(parser);
-
- // In this case we have an = sign, but we don't know what it's for. We
- // need to treat it as an error. For now, we'll mark it as an error
- // and just skip right past it.
- pm_parser_err_token(parser, &token, PM_ERR_EXPECT_EXPRESSION_AFTER_EQUAL);
+ pm_parser_err_token(parser, &token, PM_ERR_EXPRESSION_NOT_WRITABLE);
  return node;
  }
  }

templates/src/diagnostic.c.erb

@@ -188,6 +188,14 @@ static const pm_diagnostic_data_t diagnostic_messages[PM_DIAGNOSTIC_ID_MAX] = {
  [PM_ERR_EXPECT_STRING_CONTENT] = { "expected string content after opening string delimiter", PM_ERROR_LEVEL_SYNTAX },
  [PM_ERR_EXPECT_WHEN_DELIMITER] = { "expected a delimiter after the predicates of a `when` clause", PM_ERROR_LEVEL_SYNTAX },
  [PM_ERR_EXPRESSION_BARE_HASH] = { "unexpected bare hash in expression", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE] = { "unexpected '='; target cannot be written", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_ENCODING] = { "Can't assign to __ENCODING__", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_FALSE] = { "Can't assign to false", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_FILE] = { "Can't assign to __FILE__", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_LINE] = { "Can't assign to __LINE__", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_NIL] = { "Can't assign to nil", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_SELF] = { "Can't change the value of self", PM_ERROR_LEVEL_SYNTAX },
+ [PM_ERR_EXPRESSION_NOT_WRITABLE_TRUE] = { "Can't assign to true", PM_ERROR_LEVEL_SYNTAX },
  [PM_ERR_FLOAT_PARSE] = { "could not parse the float '%.*s'", PM_ERROR_LEVEL_SYNTAX },
  [PM_ERR_FOR_COLLECTION] = { "expected a collection after the `in` in a `for` statement", PM_ERROR_LEVEL_SYNTAX },
  [PM_ERR_FOR_INDEX] = { "expected an index after `for`", PM_ERROR_LEVEL_SYNTAX },