op = expr.op()
arg, digits = op.args
arg_formatted = translator.translate(arg)
if digits is not None:
digits_formatted = translator.translate(digits)
return 'bround({}, {})'.format(arg_formatted, digits_formatted)
return 'bround({})'.format(arg_formatted)
_operation_registry = {**operation_registry}
_operation_registry.update({
ops.IsNan:
unary('isnan'),
ops.IfNull:
fixed_arity('ifnull', 2),
ops.StructField:
_struct_field,
ops.MapValueForKey:
_map_value_for_key,
ops.ArrayLength:
unary('size'),
ops.Round:
_round,
ops.HLLCardinality:
reduction('approx_count_distinct'),
ops.StrRight:
fixed_arity('right', 2),
ops.StringSplit:
def _extract_epoch_seconds(translator, expr):
return 'CAST({} AS BIGINT)'.format(
unary('unix_timestamp')(translator, expr))
ops.RegexSearch:
_regex_search,
ops.RegexExtract:
_regex_extract,
ops.RegexReplace:
_regex_replace,
ops.GroupConcat:
reduction('STRING_AGG'),
ops.IfNull:
fixed_arity('IFNULL', 2),
ops.Cast:
_cast,
ops.StructField:
_struct_field,
ops.ArrayCollect:
unary('ARRAY_AGG'),
ops.ArrayConcat:
_array_concat,
ops.ArrayIndex:
_array_index,
ops.ArrayLength:
unary('ARRAY_LENGTH'),
ops.HLLCardinality:
reduction('APPROX_COUNT_DISTINCT'),
ops.Log:
_log,
ops.Sign:
unary('SIGN'),
ops.Modulus:
fixed_arity('MOD', 2),
ops.Date:
op = expr.op()
arg, digits = op.args
arg_formatted = translator.translate(arg)
if digits is not None:
digits_formatted = translator.translate(digits)
return 'bround({}, {})'.format(arg_formatted, digits_formatted)
return 'bround({})'.format(arg_formatted)
_operation_registry = {**operation_registry}
_operation_registry.update(
{
ops.IsNan: unary('isnan'),
ops.IfNull: fixed_arity('ifnull', 2),
ops.StructField: _struct_field,
ops.MapValueForKey: _map_value_for_key,
ops.ArrayLength: unary('size'),
ops.Round: _round,
ops.HLLCardinality: reduction('approx_count_distinct'),
ops.StrRight: fixed_arity('right', 2),
ops.StringSplit: fixed_arity('SPLIT', 2),
ops.RegexSearch: fixed_arity('rlike', 2),
ops.StringConcat: _string_concat,
ops.ArrayConcat: fixed_arity('concat', 2),
ops.GroupConcat: _group_concat,
ops.Cast: _cast,
ops.ExtractYear: unary('year'),
ops.ExtractMonth: unary('month'),
ops.DenseRank: _subtract_one,
ops.MinRank: _subtract_one,
ops.NTile: _subtract_one,
}
def _string_like(translator, expr):
arg, pattern, _ = expr.op().args
return '{} LIKE {}'.format(translator.translate(arg),
translator.translate(pattern))
_operation_registry = {
# string operations
ops.StringLength:
unary('length'),
ops.StringAscii:
unary('ascii'),
ops.Lowercase:
unary('lower'),
ops.Uppercase:
unary('upper'),
ops.Reverse:
unary('reverse'),
ops.Strip:
unary('trim'),
ops.LStrip:
unary('ltrim'),
ops.RStrip:
unary('rtrim'),
ops.Capitalize:
ops.RegexSearch:
_regex_search,
ops.RegexExtract:
_regex_extract,
ops.RegexReplace:
_regex_replace,
ops.GroupConcat:
reduction("STRING_AGG"),
ops.IfNull:
fixed_arity("NVL", 2),
ops.Cast:
_cast,
ops.StructField:
_struct_field,
ops.ArrayCollect:
unary("ARRAY_AGG"),
ops.ArrayConcat:
_array_concat,
ops.ArrayIndex:
_array_index,
ops.ArrayLength:
unary("ARRAY_LENGTH"),
ops.HLLCardinality:
reduction("APPROX_COUNT_DISTINCT"),
ops.Log:
_log,
ops.Sign:
unary("SIGN"),
ops.Modulus:
fixed_arity("MOD", 2),
ops.Date:
_operation_registry.update({
# aggregate methods
ops.Count:
_reduction_count(sa.func.count),
ops.Max:
_reduction('max'),
ops.Min:
_reduction('min'),
ops.Sum:
_reduction('sum'),
ops.Mean:
_reduction('avg', 'float64'),
# string methods
ops.LStrip:
unary(sa.func.ltrim),
ops.Lowercase:
unary(sa.func.lower),
ops.RStrip:
unary(sa.func.rtrim),
ops.Repeat:
fixed_arity(sa.func.replicate, 2),
ops.Reverse:
unary(sa.func.reverse),
ops.StringFind:
_string_find,
ops.StringLength:
unary(sa.func.length),
ops.StringReplace:
fixed_arity(sa.func.replace, 3),
ops.Strip:
_operation_registry.update(
{
ops.Literal: _literal,
# We override this here to support time zones
ops.TableColumn: _table_column,
# types
ops.Cast: _cast,
ops.TypeOf: _typeof,
# Floating
ops.IsNan: _is_nan,
ops.IsInf: _is_inf,
# null handling
ops.IfNull: fixed_arity(sa.func.coalesce, 2),
# boolean reductions
ops.Any: unary(sa.func.bool_or),
ops.All: unary(sa.func.bool_and),
ops.NotAny: unary(lambda x: sa.not_(sa.func.bool_or(x))),
ops.NotAll: unary(lambda x: sa.not_(sa.func.bool_and(x))),
# strings
ops.Substring: _substr,
ops.StringFind: _string_find,
ops.GroupConcat: _string_agg,
ops.Capitalize: unary(sa.func.initcap),
ops.RegexSearch: _regex_search,
ops.RegexReplace: _regex_replace,
ops.RegexExtract: _regex_extract,
ops.Translate: fixed_arity('translate', 3),
ops.StringSplit: fixed_arity(sa.func.string_to_array, 2),
ops.StringJoin: _string_join,
ops.FindInSet: _find_in_set,
ops.StringReplace: fixed_arity('REPLACE', 3),
ops.StringSplit: fixed_arity('SPLIT', 2),
ops.StringConcat: _string_concat,
ops.StringJoin: _string_join,
ops.StringAscii: _string_ascii,
ops.StringFind: _string_find,
ops.StrRight: _string_right,
ops.Repeat: fixed_arity('REPEAT', 2),
ops.RegexSearch: _regex_search,
ops.RegexExtract: _regex_extract,
ops.RegexReplace: _regex_replace,
ops.GroupConcat: reduction('STRING_AGG'),
ops.IfNull: fixed_arity('IFNULL', 2),
ops.Cast: _cast,
ops.StructField: _struct_field,
ops.ArrayCollect: unary('ARRAY_AGG'),
ops.ArrayConcat: _array_concat,
ops.ArrayIndex: _array_index,
ops.ArrayLength: unary('ARRAY_LENGTH'),
ops.HLLCardinality: reduction('APPROX_COUNT_DISTINCT'),
ops.Log: _log,
ops.Sign: unary('SIGN'),
ops.Modulus: fixed_arity('MOD', 2),
ops.Date: unary('DATE'),
# BigQuery doesn't have these operations built in.
# ops.ArrayRepeat: _array_repeat,
# ops.ArraySlice: _array_slice,
ops.Literal: _literal,
ops.Arbitrary: _arbitrary,
ops.TimestampTruncate: _truncate('TIMESTAMP', _timestamp_units),
ops.DateTruncate: _truncate('DATE', _date_units),