wsme/doc/source/types.rst
Stephen Finucane 7092903ba1 Rework documentation build
Modern docs! This fixes a couple of issues introduced in the previous
patches and generally cleans up a lot of mess.

Change-Id: Ib964c16251bce12fe498b13455ed3515ef205916
Signed-off-by: Stephen Finucane <stephenfin@redhat.com>
2019-11-18 13:58:59 -08:00

6.8 KiB

Types

Three kinds of data types can be used as input or output by WSME.

Native types

The native types are a fixed set of standard Python types that different protocols map to their own basic types.

The native types are :

  • .. wsme:type:: bytes

    A pure-ASCII string (:pywsme.types.bytes which is :pystr in Python 2 and :pybytes in Python 3).

  • .. wsme:type:: text

    A unicode string (:pywsme.types.text which is :pyunicode in Python 2 and :pystr in Python 3).

  • .. wsme:type:: int

    An integer (:pyint)

  • .. wsme:type:: float

    A float (:pyfloat)

  • .. wsme:type:: bool

    A boolean (:pybool)

  • .. wsme:type:: Decimal

    A fixed-width decimal (:pydecimal.Decimal)

  • .. wsme:type:: date

    A date (:pydatetime.date)

  • .. wsme:type:: datetime

    A date and time (:pydatetime.datetime)

  • .. wsme:type:: time

    A time (:pydatetime.time)

  • Arrays

    This is a special case. When stating a list datatype, always state its content type as the unique element of a list. Example:

    class SomeWebService(object):
        @expose([str])
        def getlist(self):
            return ['a', 'b', 'c']
  • Dictionaries

    Statically typed mappings are allowed. When exposing a dictionary datatype, you can specify the key and value types, with a restriction on the key value that must be a 'pod' type. Example:

    class SomeType(object):
        amap = {str: SomeOthertype}

There are other types that are supported out of the box. See the pre-defined-user-types.

User types

User types allow you to define new, almost-native types.

The idea is that you may have Python data that should be transported as base types by the different protocols, but needs conversion to/from these base types, or needs to validate data integrity.

To define a user type, you just have to inherit from wsme.types.UserType and instantiate your new class. This instance will be your new type and can be used as @wsme.expose or @wsme.validate parameters.

Note that protocols can choose to specifically handle a user type or a base class of user types. This is case with the two pre-defined user types, wsme.types.Enum and wsme.types.binary.

Pre-defined user types

WSME provides some pre-defined user types:

  • binary <wsme.types.binary> -- for transporting binary data as base64 strings.
  • Enum <wsme.types.Enum> -- enforce that the values belongs to a pre-defined list of values.

These types are good examples of how to define user types. Have a look at their source code!

Here is a little example that combines binary <wsme.types.binary> and Enum <wsme.types.Enum>:

ImageKind = Enum(str, 'jpeg', 'gif')

class Image(object):
    name = unicode
    kind = ImageKind
    data = binary

wsme.types.binary

The wsme.types.BinaryType instance to use when you need to transfer base64 encoded data.

wsme.types.BinaryType

wsme.types.Enum

Complex types

Complex types are structured types. They are defined as simple Python classes and will be mapped to adequate structured types in the various protocols.

A base class for structured types is provided, wsme.types.Base, but is not mandatory. The only thing it adds is a default constructor.

The attributes that are set at the class level will be used by WSME to discover the structure. These attributes can be:

  • A datatype -- Any native, user or complex type.
  • A wsattr <wsme.wsattr> -- This allows you to add more information about the attribute, for example if it is mandatory.
  • A wsproperty <wsme.wsproperty> -- A special typed property. Works like standard property with additional properties like wsattr <wsme.wsattr>.

Attributes having a leading '_' in their name will be ignored, as well as the attributes that are not in the above list. This means the type can have methods, they will not get in the way.

Example

Gender = wsme.types.Enum(str, 'male', 'female')
Title = wsme.types.Enum(str, 'M', 'Mrs')

class Person(wsme.types.Base):
    lastname = wsme.types.wsattr(unicode, mandatory=True)
    firstname = wsme.types.wsattr(unicode, mandatory=True)

    age = int
    gender = Gender
    title = Title

    hobbies = [unicode]

Rules

A few things you should know about complex types:

  • The class must have a default constructor --Since instances of the type will be created by the protocols when used as input types, they must be instantiable without any argument.

  • Complex types are registered automatically (and thus inspected) as soon a they are used in expose or validate, even if they are nested in another complex type.

    If for some reason you need to control when type is inspected, you can use wsme.types.register_type.

  • The datatype attributes will be replaced.

    When using the 'short' way of defining attributes, ie setting a simple data type, they will be replaced by a wsattr instance.

    So, when you write:

    class Person(object):
        name = unicode

    After type registration the class will actually be equivalent to:

    class Person(object):
        name = wsattr(unicode)

    You can still access the datatype by accessing the attribute on the class, along with the other wsattr properties:

    class Person(object):
        name = unicode
    
    register_type(Person)
    
    assert Person.name.datatype is unicode
    assert Person.name.key == "name"
    assert Person.name.mandatory is False
  • The default value of instance attributes is Unset <wsme.Unset>.

    class Person(object):
        name = wsattr(unicode)
    
    p = Person()
    assert p.name is Unset

    This allows the protocol to make a clear distinction between null values that will be transmitted, and unset values that will not be transmitted.

    For input values, it allows the code to know if the values were, or were not, sent by the caller.

  • When 2 complex types refer to each other, their names can be used as datatypes to avoid adding attributes afterwards:

    class A(object):
        b = wsattr('B')
    
    class B(object):
        a = wsattr(A)

Predefined Types

wsme

  • wsme.types.File