def map():
yield dt.spaceless_string("map")
yield dt.LPAREN
key_type = yield ty
yield dt.COMMA
value_type = yield ty
yield dt.RPAREN
return dt.Map(key_type, value_type, nullable=False)
def map():
yield spaceless_string("map")
yield LANGLE
key_type = yield primitive
yield COMMA
value_type = yield ty
yield RANGLE
return Map(key_type, value_type)
def struct():
yield spaceless_string("struct")
yield LPAREN
field_names_types = yield (
p.seq(field, ty)
.combine(lambda field, ty: (field, ty))
.sep_by(COMMA)
)
yield RPAREN
return Struct.from_tuples(field_names_types)
def nested():
yield dt.spaceless_string("nested")
yield dt.LPAREN
field_names_types = yield (p.seq(
dt.SPACES.then(dt.FIELD.skip(at_least_one_space)),
ty).combine(lambda field, ty:
(field, dt.Array(ty, nullable=False))).sep_by(
dt.COMMA))
yield dt.RPAREN
return dt.Struct.from_tuples(field_names_types, nullable=False)
def decimal():
yield spaceless_string("decimal", "numeric")
prec_scale = (
yield LPAREN.then(
p.seq(PRECISION.skip(COMMA), SCALE).combine(
lambda prec, scale: (prec, scale)
)
)
.skip(RPAREN)
.optional()
) or default_decimal_parameters
return Decimal(*prec_scale)
def struct():
yield dt.spaceless_string("tuple")
yield dt.LPAREN
field_names_types = yield (p.seq(
dt.SPACES.then(dt.FIELD.skip(at_least_one_space).optional()),
ty,
).combine(lambda field, ty: (field, ty)).sep_by(dt.COMMA))
yield dt.RPAREN
return dt.Struct.from_tuples(
[(field_name if field_name is not None else f"f{i:d}", typ)
for i, (field_name, typ) in enumerate(field_names_types)],
nullable=False,
)
def parse(text: str, default_decimal_parameters=(18, 3)) -> DataType:
"""Parse a DuckDB type into an ibis data type."""
primitive = (
spaceless_string("interval").result(Interval())
| spaceless_string("bigint", "int8", "long").result(int64)
| spaceless_string("boolean", "bool", "logical").result(boolean)
| spaceless_string(
"blob",
"bytea",
"binary",
"varbinary",
).result(binary)
| spaceless_string("double", "float8").result(float64)
| spaceless_string("real", "float4", "float").result(float32)
| spaceless_string("smallint", "int2", "short").result(int16)
| spaceless_string("timestamp", "datetime").result(
Timestamp(timezone="UTC")
)
| spaceless_string("date").result(date)
| spaceless_string("time").result(time)
| spaceless_string("tinyint", "int1").result(int8)
| spaceless_string("integer", "int4", "int", "signed").result(int32)
| spaceless_string("ubigint").result(uint64)
| spaceless_string("usmallint").result(uint16)
| spaceless_string("uinteger").result(uint32)
| spaceless_string("utinyint").result(uint8)
| spaceless_string("uuid").result(uuid)
| spaceless_string(
"varchar",
"char",
"bpchar",
"text",
"string",
).result(string)
)
@p.generate
def decimal():
yield spaceless_string("decimal", "numeric")
prec_scale = (
yield LPAREN.then(
p.seq(PRECISION.skip(COMMA), SCALE).combine(
lambda prec, scale: (prec, scale)
)
)
.skip(RPAREN)
.optional()
) or default_decimal_parameters
return Decimal(*prec_scale)
@p.generate
def angle_type():
yield LANGLE
value_type = yield ty
yield RANGLE
return value_type
@p.generate
def list_array():
yield spaceless_string("list")
value_type = yield angle_type
return Array(value_type)
@p.generate
def pg_array():
value_type = yield non_pg_array_type
yield LBRACKET
yield RBRACKET
return Array(value_type)
@p.generate
def map():
yield spaceless_string("map")
yield LANGLE
key_type = yield primitive
yield COMMA
value_type = yield ty
yield RANGLE
return Map(key_type, value_type)
field = spaceless(FIELD)
@p.generate
def struct():
yield spaceless_string("struct")
yield LPAREN
field_names_types = yield (
p.seq(field, ty)
.combine(lambda field, ty: (field, ty))
.sep_by(COMMA)
)
yield RPAREN
return Struct.from_tuples(field_names_types)
non_pg_array_type = primitive | decimal | list_array | map | struct
ty = pg_array | non_pg_array_type
return ty.parse(text)
def list_array():
yield spaceless_string("list")
value_type = yield angle_type
return Array(value_type)
def array():
yield dt.spaceless_string("array")
value_type = yield paren_type
return dt.Array(value_type, nullable=False)
def decimal():
yield dt.spaceless_string("decimal", "numeric")
precision, scale = yield dt.LPAREN.then(
p.seq(dt.PRECISION.skip(dt.COMMA), dt.SCALE)).skip(dt.RPAREN)
return dt.Decimal(precision, scale, nullable=False)
def fixed_string():
yield dt.spaceless_string("fixedstring")
yield dt.LPAREN
yield dt.NUMBER
yield dt.RPAREN
return dt.String(nullable=False)
def nullable():
yield dt.spaceless_string("nullable")
yield dt.LPAREN
parsed_ty = yield ty
yield dt.RPAREN
return parsed_ty(nullable=True)
def datetime():
yield dt.spaceless_string("datetime64", "datetime")
timezone = yield parened_string.optional()
return dt.Timestamp(timezone=timezone, nullable=False)
def parse(text: str) -> DataType:
@p.generate
def datetime():
yield dt.spaceless_string("datetime64", "datetime")
timezone = yield parened_string.optional()
return dt.Timestamp(timezone=timezone, nullable=False)
primitive = (
datetime
| dt.spaceless_string("null", "nothing").result(dt.null)
| dt.spaceless_string("bigint", "int64").result(
dt.Int64(nullable=False))
| dt.spaceless_string("double", "float64").result(
dt.Float64(nullable=False))
| dt.spaceless_string("float32", "float").result(
dt.Float32(nullable=False))
| dt.spaceless_string("smallint", "int16", "int2").result(
dt.Int16(nullable=False))
| dt.spaceless_string("date32", "date").result(dt.Date(nullable=False))
| dt.spaceless_string("time").result(dt.Time(nullable=False))
| dt.spaceless_string("tinyint", "int8", "int1", "boolean",
"bool").result(dt.Int8(nullable=False))
| dt.spaceless_string("integer", "int32", "int4", "int").result(
dt.Int32(nullable=False))
| dt.spaceless_string("uint64").result(dt.UInt64(nullable=False))
| dt.spaceless_string("uint32").result(dt.UInt32(nullable=False))
| dt.spaceless_string("uint16").result(dt.UInt16(nullable=False))
| dt.spaceless_string("uint8").result(dt.UInt8(nullable=False))
| dt.spaceless_string("uuid").result(dt.UUID(nullable=False))
| dt.spaceless_string(
"longtext",
"mediumtext",
"tinytext",
"text",
"longblob",
"mediumblob",
"tinyblob",
"blob",
"varchar",
"char",
"string",
).result(dt.String(nullable=False)))
@p.generate
def parened_string():
yield dt.LPAREN
s = yield dt.RAW_STRING
yield dt.RPAREN
return s
@p.generate
def nullable():
yield dt.spaceless_string("nullable")
yield dt.LPAREN
parsed_ty = yield ty
yield dt.RPAREN
return parsed_ty(nullable=True)
@p.generate
def fixed_string():
yield dt.spaceless_string("fixedstring")
yield dt.LPAREN
yield dt.NUMBER
yield dt.RPAREN
return dt.String(nullable=False)
@p.generate
def decimal():
yield dt.spaceless_string("decimal", "numeric")
precision, scale = yield dt.LPAREN.then(
p.seq(dt.PRECISION.skip(dt.COMMA), dt.SCALE)).skip(dt.RPAREN)
return dt.Decimal(precision, scale, nullable=False)
@p.generate
def paren_type():
yield dt.LPAREN
value_type = yield ty
yield dt.RPAREN
return value_type
@p.generate
def array():
yield dt.spaceless_string("array")
value_type = yield paren_type
return dt.Array(value_type, nullable=False)
@p.generate
def map():
yield dt.spaceless_string("map")
yield dt.LPAREN
key_type = yield ty
yield dt.COMMA
value_type = yield ty
yield dt.RPAREN
return dt.Map(key_type, value_type, nullable=False)
at_least_one_space = p.regex(r"\s+")
@p.generate
def nested():
yield dt.spaceless_string("nested")
yield dt.LPAREN
field_names_types = yield (p.seq(
dt.SPACES.then(dt.FIELD.skip(at_least_one_space)),
ty).combine(lambda field, ty:
(field, dt.Array(ty, nullable=False))).sep_by(
dt.COMMA))
yield dt.RPAREN
return dt.Struct.from_tuples(field_names_types, nullable=False)
@p.generate
def struct():
yield dt.spaceless_string("tuple")
yield dt.LPAREN
field_names_types = yield (p.seq(
dt.SPACES.then(dt.FIELD.skip(at_least_one_space).optional()),
ty,
).combine(lambda field, ty: (field, ty)).sep_by(dt.COMMA))
yield dt.RPAREN
return dt.Struct.from_tuples(
[(field_name if field_name is not None else f"f{i:d}", typ)
for i, (field_name, typ) in enumerate(field_names_types)],
nullable=False,
)
ty = (nullable
| nested
| primitive
| fixed_string
| decimal
| array
| map
| struct)
return ty.parse(text)
