관리-도구
편집 파일: indexable.cpython-38.pyc
U -?�f+ � @ sB d Z ddlmZ ddlmZ ddlmZ dgZG dd� de�ZdS )a Define attributes on ORM-mapped classes that have "index" attributes for columns with :class:`_types.Indexable` types. "index" means the attribute is associated with an element of an :class:`_types.Indexable` column with the predefined index to access it. The :class:`_types.Indexable` types include types such as :class:`_types.ARRAY`, :class:`_types.JSON` and :class:`_postgresql.HSTORE`. The :mod:`~sqlalchemy.ext.indexable` extension provides :class:`_schema.Column`-like interface for any element of an :class:`_types.Indexable` typed column. In simple cases, it can be treated as a :class:`_schema.Column` - mapped attribute. Synopsis ======== Given ``Person`` as a model with a primary key and JSON data field. While this field may have any number of elements encoded within it, we would like to refer to the element called ``name`` individually as a dedicated attribute which behaves like a standalone column:: from sqlalchemy import Column, JSON, Integer from sqlalchemy.ext.declarative import declarative_base from sqlalchemy.ext.indexable import index_property Base = declarative_base() class Person(Base): __tablename__ = 'person' id = Column(Integer, primary_key=True) data = Column(JSON) name = index_property('data', 'name') Above, the ``name`` attribute now behaves like a mapped column. We can compose a new ``Person`` and set the value of ``name``:: >>> person = Person(name='Alchemist') The value is now accessible:: >>> person.name 'Alchemist' Behind the scenes, the JSON field was initialized to a new blank dictionary and the field was set:: >>> person.data {"name": "Alchemist'} The field is mutable in place:: >>> person.name = 'Renamed' >>> person.name 'Renamed' >>> person.data {'name': 'Renamed'} When using :class:`.index_property`, the change that we make to the indexable structure is also automatically tracked as history; we no longer need to use :class:`~.mutable.MutableDict` in order to track this change for the unit of work. Deletions work normally as well:: >>> del person.name >>> person.data {} Above, deletion of ``person.name`` deletes the value from the dictionary, but not the dictionary itself. A missing key will produce ``AttributeError``:: >>> person = Person() >>> person.name ... AttributeError: 'name' Unless you set a default value:: >>> class Person(Base): >>> __tablename__ = 'person' >>> >>> id = Column(Integer, primary_key=True) >>> data = Column(JSON) >>> >>> name = index_property('data', 'name', default=None) # See default >>> person = Person() >>> print(person.name) None The attributes are also accessible at the class level. Below, we illustrate ``Person.name`` used to generate an indexed SQL criteria:: >>> from sqlalchemy.orm import Session >>> session = Session() >>> query = session.query(Person).filter(Person.name == 'Alchemist') The above query is equivalent to:: >>> query = session.query(Person).filter(Person.data['name'] == 'Alchemist') Multiple :class:`.index_property` objects can be chained to produce multiple levels of indexing:: from sqlalchemy import Column, JSON, Integer from sqlalchemy.ext.declarative import declarative_base from sqlalchemy.ext.indexable import index_property Base = declarative_base() class Person(Base): __tablename__ = 'person' id = Column(Integer, primary_key=True) data = Column(JSON) birthday = index_property('data', 'birthday') year = index_property('birthday', 'year') month = index_property('birthday', 'month') day = index_property('birthday', 'day') Above, a query such as:: q = session.query(Person).filter(Person.year == '1980') On a PostgreSQL backend, the above query will render as:: SELECT person.id, person.data FROM person WHERE person.data -> %(data_1)s -> %(param_1)s = %(param_2)s Default Values ============== :class:`.index_property` includes special behaviors for when the indexed data structure does not exist, and a set operation is called: * For an :class:`.index_property` that is given an integer index value, the default data structure will be a Python list of ``None`` values, at least as long as the index value; the value is then set at its place in the list. This means for an index value of zero, the list will be initialized to ``[None]`` before setting the given value, and for an index value of five, the list will be initialized to ``[None, None, None, None, None]`` before setting the fifth element to the given value. Note that an existing list is **not** extended in place to receive a value. * for an :class:`.index_property` that is given any other kind of index value (e.g. strings usually), a Python dictionary is used as the default data structure. * The default data structure can be set to any Python callable using the :paramref:`.index_property.datatype` parameter, overriding the previous rules. Subclassing =========== :class:`.index_property` can be subclassed, in particular for the common use case of providing coercion of values or SQL expressions as they are accessed. Below is a common recipe for use with a PostgreSQL JSON type, where we want to also include automatic casting plus ``astext()``:: class pg_json_property(index_property): def __init__(self, attr_name, index, cast_type): super(pg_json_property, self).__init__(attr_name, index) self.cast_type = cast_type def expr(self, model): expr = super(pg_json_property, self).expr(model) return expr.astext.cast(self.cast_type) The above subclass can be used with the PostgreSQL-specific version of :class:`_postgresql.JSON`:: from sqlalchemy import Column, Integer from sqlalchemy.ext.declarative import declarative_base from sqlalchemy.dialects.postgresql import JSON Base = declarative_base() class Person(Base): __tablename__ = 'person' id = Column(Integer, primary_key=True) data = Column(JSON) age = pg_json_property('data', 'age', Integer) The ``age`` attribute at the instance level works as before; however when rendering SQL, PostgreSQL's ``->>`` operator will be used for indexed access, instead of the usual index operator of ``->``:: >>> query = session.query(Person).filter(Person.age < 20) The above query will render:: SELECT person.id, person.data FROM person WHERE CAST(person.data ->> %(data_1)s AS INTEGER) < %(param_1)s � )�inspect)�hybrid_property)� flag_modified�index_propertyc sZ e Zd ZdZe� Zedddf� fdd� Zddd�Zdd � Zd d � Z d d � Z dd� Z � Z S )r z�A property generator. The generated property describes an object attribute that corresponds to an :class:`_types.Indexable` column. .. seealso:: :mod:`sqlalchemy.ext.indexable` NTc s� |r t � �| j| j| j| j� nt � �| jdd| j� || _� | _|| _t � t �}|oX|}|dk rj|| _ n|r~� fdd�| _ nt | _ || _ dS )a} Create a new :class:`.index_property`. :param attr_name: An attribute name of an `Indexable` typed column, or other attribute that returns an indexable structure. :param index: The index to be used for getting and setting this value. This should be the Python-side index value for integers. :param default: A value which will be returned instead of `AttributeError` when there is not a value at given index. :param datatype: default datatype to use when the field is empty. By default, this is derived from the type of index used; a Python list for an integer index, or a Python dictionary for any other style of index. For a list, the list will be initialized to a list of None values that is at least ``index`` elements long. :param mutable: if False, writes and deletes to the attribute will be disallowed. :param onebased: assume the SQL representation of this value is one-based; that is, the first index in SQL is 1, not zero. Nc s dd� t � d �D �S )Nc S s g | ]}d �qS �N� )�.0�xr r �H/opt/hc_python/lib64/python3.8/site-packages/sqlalchemy/ext/indexable.py� <listcomp>! s z=index_property.__init__.<locals>.<lambda>.<locals>.<listcomp>� )�ranger ��indexr r �<lambda>! � z)index_property.__init__.<locals>.<lambda>)�super�__init__�fget�fset�fdel�expr� attr_namer �default� isinstance�int�datatype�dict�onebased)�selfr r r r Zmutabler Z is_numeric�� __class__r r r � s zindex_property.__init__c C s$ | j | jkrt| j�|�n| j S d S r )r �_NO_DEFAULT_ARGUMENT�AttributeErrorr )r �errr r r � _fget_default& s z index_property._fget_defaultc C sl | j }t||�}|d kr | �� S z|| j }W n4 ttfk rb } z| �|� W Y �S d }~X Y nX |S d S r )r �getattrr% r �KeyError� IndexError)r �instancer � column_value�valuer$ r r r r , s zindex_property.fgetc C sb | j }t||d �}|d kr.| �� }t|||� ||| j< t|||� |t|�jjkr^t||� d S r ) r r& r �setattrr r Zmapper�attrsr )r r) r+ r r* r r r r 8 s zindex_property.fsetc C sx | j }t||�}|d kr"t| j ��z || j= W n. tk r\ } zt| j �|�W 5 d }~X Y nX t|||� t||� d S r )r r& r# r r' r, r )r r) r r* r$ r r r r C s zindex_property.fdelc C s( t || j�}| j}| jr |d7 }|| S )Nr )r& r r r )r �model�columnr r r r r P s zindex_property.expr)N) �__name__� __module__� __qualname__�__doc__�objectr" r r% r r r r � __classcell__r r r r r � s �3 N) r3 � r Z ext.hybridr Zorm.attributesr �__all__r r r r r �<module> s V