Geometry

Shapely geometry classes, such as shapely.Point, are the central data types in Shapely. Each geometry class extends the shapely.Geometry base class, which is a container of the underlying GEOS geometry object, to provide geometry type-specific attributes and behavior. The Geometry object keeps track of the underlying GEOS geometry and lets the python garbage collector free its memory when it is not used anymore.

Geometry objects are immutable. This means that after constructed, they cannot be changed in place. Every Shapely operation will result in a new object being returned.

Geometry types

Point(*args)	A geometry type that represents a single coordinate with x,y and possibly z values.
LineString([coordinates])	A geometry type composed of one or more line segments.
LinearRing([coordinates])	A geometry type composed of one or more line segments that forms a closed loop.
Polygon([shell, holes])	A geometry type representing an area that is enclosed by a linear ring.
MultiPoint([points])	A collection of one or more Points.
MultiLineString([lines])	A collection of one or more LineStrings.
MultiPolygon([polygons])	A collection of one or more Polygons.
GeometryCollection([geoms])	A collection of one or more geometries that may contain more than one type of geometry.

Construction

Geometries can be constructed directly using Shapely geometry classes:

>>> from shapely import Point, LineString
>>> Point(5.2, 52.1)
<POINT (5.2 52.1)>
>>> LineString([(0, 0), (1, 2)])
<LINESTRING (0 0, 1 2)>

Geometries can also be constructed from a WKT (Well-Known Text) or WKB (Well-Known Binary) representation:

>>> from shapely import from_wkb, from_wkt
>>> from_wkt("POINT (5.2 52.1)")
<POINT (5.2 52.1)>
>>> from_wkb(b"\x01\x01\x00\x00\x00\x00\x00\x00\x00\x00\x00\xf0?\x00\x00\x00\x00\x00\x00\xf0?")
<POINT (1 1)>

A more efficient way of constructing geometries is by making use of the (vectorized) functions described in shapely.creation.

Pickling

Geometries can be serialized using pickle:

>>> import pickle
>>> from shapely import Point
>>> pickled = pickle.dumps(Point(1, 1))
>>> pickle.loads(pickled)
<POINT (1 1)>

Warning

Pickling will convert linearrings to linestrings. See shapely.to_wkb() for a complete list of limitations.

Hashing

Geometries can be used as elements in sets or as keys in dictionaries. Python uses a technique called hashing for lookups in these datastructures. Shapely generates this hash from the WKB representation. Therefore, geometries are equal if and only if their WKB representations are equal.

>>> from shapely import Point
>>> point_1 = Point(5.2, 52.1)
>>> point_2 = Point(1, 1)
>>> point_3 = Point(5.2, 52.1)
>>> {point_1, point_2, point_3}
{<POINT (1 1)>, <POINT (5.2 52.1)>}

Warning

Due to limitations of WKB, linearrings will equal linestrings if they contain the exact same points. See shapely.to_wkb().

Comparing two geometries directly is also supported. This is the same as using shapely.predicates.equals_exact() with a tolerance value of zero.

>>> point_1 == point_2
False
>>> point_1 == point_3
True
>>> point_1 != point_2
True

Formatting

Geometries can be formatted to strings using properties, functions, or a Python format specification.

The most convenient is to use .wkb_hex and .wkt properties.

>>> from shapely import Point, to_wkb, to_wkt, to_geojson
>>> pt = Point(-169.910918, -18.997564)
>>> pt.wkb_hex
0101000000CF6A813D263D65C0BDAAB35A60FF32C0
>>> pt.wkt
POINT (-169.910918 -18.997564)

More output options can be found using using to_wkb(), to_wkt(), and to_geojson() functions.

>>> to_wkb(pt, hex=True, byte_order=0)
0000000001C0653D263D816ACFC032FF605AB3AABD
>>> to_wkt(pt, rounding_precision=3)
POINT (-169.911 -18.998)
>>> print(to_geojson(pt, indent=2))
{
  "type": "Point",
  "coordinates": [
    -169.910918,
    -18.997564
  ]
}

A format specification may also be used to control the format and precision.

>>> print(f"Cave near {pt:.3f}")
Cave near POINT (-169.911 -18.998)
>>> print(f"or hex-encoded as {pt:x}")
or hex-encoded as 0101000000cf6a813d263d65c0bdaab35a60ff32c0

Shapely has a format specification inspired from Python’s Format Specification Mini-Language described next.

Semantic for format specification

format_spec ::=  [0][.precision][type]
precision   ::=  digit+
type        ::=  "f" | "F" | "g" | "G" | "x" | "X"

Format types 'f' and 'F' are to use a fixed-point notation, which is activated by setting GEOS’ trim option off. The upper case variant converts nan to NAN and inf to INF.

Format types 'g' and 'G' are to use a “general format”, where unnecessary digits are trimmed. This notation is activated by setting GEOS’ trim option on. The upper case variant is similar to 'F', and may also display an upper-case "E" if scientific notation is required. Note that this representation may be different for GEOS 3.10.0 and later, which does not use scientific notation.

For numeric outputs 'f' and 'g', the precision is optional, and if not specified, rounding precision will be disabled showing full precision.

Format types 'x' and 'X' show a hex-encoded string representation of WKB or Well-Known Binary, with the case of the output matched the case of the format type character.