Method search

tests/comparison/discrepancy_searcher.py:675–812

Returns an instance of SearchResults, which is a summary report. This method oversees the generation, execution, and comparison of queries. number_of_test_queries should an integer indicating the maximum number of queries to generate and execute.

(self, number_of_test_queries, stop_on_result_mismatch, stop_on_crash,
             query_timeout_seconds)

    return new_table

  def search(self, number_of_test_queries, stop_on_result_mismatch, stop_on_crash,
             query_timeout_seconds):
    '''Returns an instance of SearchResults, which is a summary report. This method
       oversees the generation, execution, and comparison of queries.

      number_of_test_queries should an integer indicating the maximum number of queries
      to generate and execute.
    '''
    start_time = time()
    self.query_result_comparator = QueryResultComparator(
        self.query_profile, self.ref_conn, self.test_conn, query_timeout_seconds)
    query_count = 0
    queries_resulted_in_data_count = 0
    mismatch_count = 0
    query_timeout_count = 0
    known_error_count = 0
    test_crash_count = 0
    last_error = None
    repeat_error_count = 0
    count_effective_dml_statements = 0
    count_rows_affected_by_dml = 0

    while number_of_test_queries > query_count:
      statement_type = self.query_profile.choose_statement()
      statement_generator = get_generator(statement_type)(self.query_profile)
      dml_table = None
      if issubclass(statement_type, (InsertStatement,)):
        dml_choice_src_table = self.query_profile.choose_table(self.common_tables)
        # Copy the table we want to INSERT/UPSERT INTO. Do this for the following reasons:
        #
        # 1. If we don't copy, the tables will get larger and larger
        # 2. If we want to avoid tables getting larger and larger, we have to come up
        # with some threshold about when to cut and start over.
        # 3. If we keep INSERT/UPSERTing into tables and finally find a crash, we have to
        # replay all previous INSERT/UPSERTs again. Those INSERTs may not produce the
        # same rows as before. To maximize the chance of bug reproduction, run every
        # INSERT/UPSERT on a pristine table.
        dml_table = self._concurrently_copy_table(dml_choice_src_table)
      statement = statement_generator.generate_statement(
          self.common_tables, dml_table=dml_table)
      if isinstance(statement, Query):
        # we can re-write statement execution here to possibly be a CREATE TABLE AS SELECT
        # or CREATE VIEW AS SELECT
        statement.execution = self.query_profile.get_query_execution()
      query_count += 1
      LOG.info('Running query #%s', query_count)
      result = self.query_result_comparator.compare_query_results(statement)
      if result.query_resulted_in_data:
        queries_resulted_in_data_count += 1
      if result.modified_rows_count:
        count_effective_dml_statements += 1
        count_rows_affected_by_dml += abs(
            result.modified_rows_count - self._dml_table_size)
      if isinstance(result.exception, DataLimitExceeded) \
          or isinstance(result.exception, TypeOverflow):
        continue
      if result.error:
        # TODO: These first two come from psycopg2, the postgres driver. Maybe we should