Types
Types are the fundamental building block of visions, representing “logical” abstractions of a sequence. This logical abstraction is distinct is often distinct from the underlying in-memory implementation.
To make this distinction more concrete we can imagine the sequence ['Apple', 'Orange', 'Pear'].
At a logical level these are of the type Fruit while under the hood each element is represented by the machine as String.
Types have two basic capabilities:
Type membership testing
Type relations
Membership
Type membership is an important component of every visions type, membership checks answer the question “is my sequence of a type?” In practice this looks like
>>> import visions as v
>>> test_series = pd.Series(['Apple', 'Orange', 'Pear'])
>>> test_series in v.String
True
>>> test_series in v.Integer
False
See also
Relations
Types also have special knowledge about their relations with other types. A relation represents a mapping between types. Let’s take the example of an integer
>>> import visions as v
>>> v.Integer.relations
[IdentityRelation(Generic -> Integer),
InferenceRelation(Float -> Integer),
InferenceRelation(String -> Integer)]
As we can see, there are three relations defined on the integer type. Each represents a mapping from another type to integer. This characteristic is generally true for all visions types; given two types, A, and B, relations defined on B represent mappings from A to B.
By defining relations in this way types are effectively decoupled from each other.
See also