+ }j-Rt^RIHt^RIHt^RIHt^RIHt^RIHt^RIHt^RIH t ^RIH t ^R I H t ^R I Ht^R IHt]'d ^R IHt^R IHtR.t] !R4t!RR]],4tR#)a7Define 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: .. sourcecode:: sql 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: .. sourcecode:: sql SELECT person.id, person.data FROM person WHERE CAST(person.data ->> %(data_1)s AS INTEGER) < %(param_1)s ) annotations)Any)Callable)cast)Optional) TYPE_CHECKING)TypeVar)Union)inspect)hybrid_property) flag_modified)SQLColumnExpression)_HasClauseElementindex_property_Tca]tRt^tRt]!]]!44t]RRR3RV3Rlllt RRRllt RR lt R R lt R R lt RRltRtV;t#)rzA property generator. The generated property describes an object attribute that corresponds to an :class:`_types.Indexable` column. .. seealso:: :mod:`sqlalchemy.ext.indexable` NTc 0V^8dQhRRRRRRRRR R R R /#)  attr_namestrindexzUnion[int, str]defaultrdatatypezOptional[Callable[[], Any]]mutableboolonebased)formats"Y/home/luca/screen-checking/.venv/lib/python3.14/site-packages/sqlalchemy/ext/indexable.py __annotate__index_property.__annotate__sF1!1!1!1! 1! . 1!  1!1!c <aV'd<\SV`VPVPVPVP 4M&\SV`VPRRVP 4WnSVnW0n\S\4pT;'dTpVeW@n MV'd V3RlVn M \Vn W`n R#)aCreate 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. NcR<\S^,4Uu.uFpRNK up#uupi)N)range)xrs r)index_property.__init__..6s"uUQY7G(H7G!7G(H(Hs $)super__init__fgetfsetfdelexprrrr isinstanceintrdictr) selfrrrrrr is_numeric __class__s &&f&&&& rr*index_property.__init__s@  G TYY 499dii H G TYYdDII >"  s+ **(  $M H $  r!c V^8dQhRRRR/#)rerrzOptional[BaseException]returnrr)rs"rrr ;s  !8 B r!c |VPVP8Xd\VP4VhVP#N)r_NO_DEFAULT_ARGUMENTAttributeErrorr)r2r7s&&r _fget_defaultindex_property._fget_default;s/ <<444 4 0c 9<< r!c V^8dQhRRRR/#)r_index_property__instancerr8rr)rs"rrr As  s r r!c VPp\W4pVfVP4#W0P,pV# \\ 3dpTPT4uRp?#Rp?ii;ir:)rgetattrr=rKeyError IndexError)r2r@r column_valuevaluer7s&& rr+index_property.fgetAsjNN z5  %%' '  ,EL*% +%%c* * +sAA-A("A-(A-c$V^8dQhRRRRRR/#)rinstancerrFrr8Noner)rs"rrr Ms! / /S / / /r!c VPp\WR4pVfVP4p\WV4W$VP&\WV4V\ V4P P9d\W4R#R#r:) rrBrsetattrrr mapperattrsr )r2rIrFrrEs&&& rr,index_property.fsetMsnNN xD9  ==?L H 6#(TZZ \2 )0066 6 ( . 7r!c V^8dQhRRRR/#)rrIrr8rJr)rs"rrr Xs / /S /T /r!c VPp\W4pVf\VP4hW0P\ WV4\ W4R# \ dp\TP4ThRp?ii;ir:)rrBr<rrLr rC)r2rIrrEr7s&& rr-index_property.fdelXspNN x3   0 0 /ZZ( H 6 ( .  : 0c 9 :s A A="A88A=c V^8dQhRRRR/#)rmodelrr8z5Union[_HasClauseElement[_T], SQLColumnExpression[_T]]r)rs"rrr es >r!c \WP4pVPpVP'd V^, pW#,#)r$)rBrrr)r2rTcolumnrs&& rr.index_property.expres5/  === QJE}r!)rrrrrr:)__name__ __module__ __qualname____firstlineno____doc__rrobjectr;r*r=r+r,r-r.__static_attributes__ __classcell__)r4s@rrrsP FH- +041!1!f  / /r!N)r\ __future__rtypingrrrrrrr r ext.hybridr orm.attributesr sqlr sql._typingr__all__rrrr!rrhs^Yv# (*)/   T]q_R(qr!