Function build_insert_into_statement

testdata/bin/generate-schema-statements.py:583–624

(insert, db_name, db_suffix, table_name, file_format,
                                hdfs_path, for_impala=False, scale_factor="")

def build_insert_into_statement(insert, db_name, db_suffix, table_name, file_format,
                                hdfs_path, for_impala=False, scale_factor=""):
  can_hint = for_impala and (file_format == 'parquet' or is_iceberg_table(file_format))
  params = build_replacement_params(
    db_name,
    db_suffix,
    table_name,
    hdfs_location=hdfs_path,
    impala_home=IMPALA_HOME,
    hint="",
    part_predicate="",
    scale_factor=scale_factor)

  m = PARTITIONED_INSERT_RE.search(insert)
  if m:
    insert_statement = build_partitioned_load(insert, m, can_hint, params)
  else:
    if can_hint:
      params[VAR_HINT] = HINT_SHUFFLE
    insert_statement = insert.format(**params)

  # Kudu tables are managed and don't support OVERWRITE, so we replace OVERWRITE
  # with INTO to make this a regular INSERT.
  if file_format == 'kudu':
    insert_statement = insert_statement.replace("OVERWRITE", "INTO")

  if for_impala:
    return insert_statement

  statement = SET_PARTITION_MODE_NONSTRICT_STATEMENT + "\n"
  statement += SET_DYNAMIC_PARTITION_STATEMENT + "\n"
  statement += SET_MAX_DYNAMIC_PARTITIONS_STATEMENT + "\n"
  statement += "set hive.auto.convert.join=true;\n"

  # For some reason (hive bug?) we need to have the CombineHiveInputFormat set
  # for cases where we are compressing in bzip on certain tables that
  # have multiple files.
  if 'multi' in table_name and ('bzip' in db_suffix):
    statement += SET_HIVE_INPUT_FORMAT % "CombineHiveInputFormat"
  else:
    statement += SET_HIVE_INPUT_FORMAT % "HiveInputFormat"
  return statement + insert_statement


def build_hbase_insert(db_name, db_suffix, table_name):