parts/django/docs/topics/db/sql.txt
changeset 307 c6bca38c1cbf
equal deleted inserted replaced
306:5ff1fc726848 307:c6bca38c1cbf
       
     1 ==========================
       
     2 Performing raw SQL queries
       
     3 ==========================
       
     4 
       
     5 .. currentmodule:: django.db.models
       
     6 
       
     7 When the :doc:`model query APIs </topics/db/queries>` don't go far enough, you
       
     8 can fall back to writing raw SQL. Django gives you two ways of performing raw
       
     9 SQL queries: you can use :meth:`Manager.raw()` to `perform raw queries and
       
    10 return model instances`__, or you can avoid the model layer entirely and
       
    11 `execute custom SQL directly`__.
       
    12 
       
    13 __ `performing raw queries`_
       
    14 __ `executing custom SQL directly`_
       
    15 
       
    16 Performing raw queries
       
    17 ======================
       
    18 
       
    19 .. versionadded:: 1.2
       
    20 
       
    21 The ``raw()`` manager method can be used to perform raw SQL queries that
       
    22 return model instances:
       
    23 
       
    24 .. method:: Manager.raw(raw_query, params=None, translations=None)
       
    25 
       
    26 This method method takes a raw SQL query, executes it, and returns a
       
    27 :class:`~django.db.models.query.RawQuerySet` instance. This
       
    28 :class:`~django.db.models.query.RawQuerySet` instance can be iterated
       
    29 over just like an normal QuerySet to provide object instances.
       
    30 
       
    31 This is best illustrated with an example. Suppose you've got the following model::
       
    32 
       
    33     class Person(models.Model):
       
    34         first_name = models.CharField(...)
       
    35         last_name = models.CharField(...)
       
    36         birth_date = models.DateField(...)
       
    37 
       
    38 You could then execute custom SQL like so::
       
    39 
       
    40     >>> for p in Person.objects.raw('SELECT * FROM myapp_person'):
       
    41     ...     print p
       
    42     John Smith
       
    43     Jane Jones
       
    44 
       
    45 .. admonition:: Model table names
       
    46 
       
    47     Where'd the name of the ``Person`` table come from in that example?
       
    48 
       
    49     By default, Django figures out a database table name by joining the
       
    50     model's "app label" -- the name you used in ``manage.py startapp`` -- to
       
    51     the model's class name, with an underscore between them. In the example
       
    52     we've assumed that the ``Person`` model lives in an app named ``myapp``,
       
    53     so its table would be ``myapp_person``.
       
    54 
       
    55     For more details check out the documentation for the
       
    56     :attr:`~Options.db_table` option, which also lets you manually set the
       
    57     database table name.
       
    58 
       
    59 Of course, this example isn't very exciting -- it's exactly the same as
       
    60 running ``Person.objects.all()``. However, ``raw()`` has a bunch of other
       
    61 options that make it very powerful.
       
    62 
       
    63 Mapping query fields to model fields
       
    64 ------------------------------------
       
    65 
       
    66 ``raw()`` automatically maps fields in the query to fields on the model.
       
    67 
       
    68 The order of fields in your query doesn't matter. In other words, both
       
    69 of the following queries work identically::
       
    70 
       
    71     >>> Person.objects.raw('SELECT id, first_name, last_name, birth_date FROM myapp_person')
       
    72     ...
       
    73     >>> Person.objects.raw('SELECT last_name, birth_date, first_name, id FROM myapp_person')
       
    74     ...
       
    75 
       
    76 Matching is done by name. This means that you can use SQL's ``AS`` clauses to
       
    77 map fields in the query to model fields. So if you had some other table that
       
    78 had ``Person`` data in it, you could easily map it into ``Person`` instances::
       
    79 
       
    80     >>> Person.objects.raw('''SELECT first AS first_name,
       
    81     ...                              last AS last_name,
       
    82     ...                              bd AS birth_date,
       
    83     ...                              pk as id,
       
    84     ...                       FROM some_other_table''')
       
    85 
       
    86 As long as the names match, the model instances will be created correctly.
       
    87 
       
    88 Alternatively, you can map fields in the query to model fields using the
       
    89 ``translations`` argument to ``raw()``. This is a dictionary mapping names of
       
    90 fields in the query to names of fields on the model. For example, the above
       
    91 query could also be written::
       
    92 
       
    93     >>> name_map = {'first': 'first_name', 'last': 'last_name', 'bd': 'birth_date', 'pk': 'id'}
       
    94     >>> Person.objects.raw('SELECT * FROM some_other_table', translations=name_map)
       
    95 
       
    96 Index lookups
       
    97 -------------
       
    98 
       
    99 ``raw()`` supports indexing, so if you need only the first result you can
       
   100 write::
       
   101 
       
   102     >>> first_person = Person.objects.raw('SELECT * from myapp_person')[0]
       
   103 
       
   104 However, the indexing and slicing are not performed at the database level. If
       
   105 you have a big amount of ``Person`` objects in your database, it is more
       
   106 efficient to limit the query at the SQL level::
       
   107 
       
   108     >>> first_person = Person.objects.raw('SELECT * from myapp_person LIMIT 1')[0]
       
   109 
       
   110 Deferring model fields
       
   111 ----------------------
       
   112 
       
   113 Fields may also be left out::
       
   114 
       
   115     >>> people = Person.objects.raw('SELECT id, first_name FROM myapp_person')
       
   116 
       
   117 The ``Person`` objects returned by this query will be deferred model instances
       
   118 (see :meth:`~django.db.models.QuerySet.defer()`). This means that the fields
       
   119 that are omitted from the query will be loaded on demand. For example::
       
   120 
       
   121     >>> for p in Person.objects.raw('SELECT id, first_name FROM myapp_person'):
       
   122     ...     print p.first_name, # This will be retrieved by the original query
       
   123     ...     print p.last_name # This will be retrieved on demand
       
   124     ...
       
   125     John Smith
       
   126     Jane Jones
       
   127 
       
   128 From outward appearances, this looks like the query has retrieved both
       
   129 the first name and last name. However, this example actually issued 3
       
   130 queries. Only the first names were retrieved by the raw() query -- the
       
   131 last names were both retrieved on demand when they were printed.
       
   132 
       
   133 There is only one field that you can't leave out - the primary key
       
   134 field. Django uses the primary key to identify model instances, so it
       
   135 must always be included in a raw query. An ``InvalidQuery`` exception
       
   136 will be raised if you forget to include the primary key.
       
   137 
       
   138 Adding annotations
       
   139 ------------------
       
   140 
       
   141 You can also execute queries containing fields that aren't defined on the
       
   142 model. For example, we could use `PostgreSQL's age() function`__ to get a list
       
   143 of people with their ages calculated by the database::
       
   144 
       
   145     >>> people = Person.objects.raw('SELECT *, age(birth_date) AS age FROM myapp_person')
       
   146     >>> for p in people:
       
   147     ...     print "%s is %s." % (p.first_name, p.age)
       
   148     John is 37.
       
   149     Jane is 42.
       
   150     ...
       
   151 
       
   152 __ http://www.postgresql.org/docs/8.4/static/functions-datetime.html
       
   153 
       
   154 Passing parameters into ``raw()``
       
   155 ---------------------------------
       
   156 
       
   157 If you need to perform parameterized queries, you can use the ``params``
       
   158 argument to ``raw()``::
       
   159 
       
   160     >>> lname = 'Doe'
       
   161     >>> Person.objects.raw('SELECT * FROM myapp_person WHERE last_name = %s', [lname])
       
   162 
       
   163 ``params`` is a list of parameters. You'll use ``%s`` placeholders in the
       
   164 query string (regardless of your database engine); they'll be replaced with
       
   165 parameters from the ``params`` list.
       
   166 
       
   167 .. warning::
       
   168 
       
   169     **Do not use string formatting on raw queries!**
       
   170 
       
   171     It's tempting to write the above query as::
       
   172 
       
   173         >>> query = 'SELECT * FROM myapp_person WHERE last_name = %s' % lname
       
   174         >>> Person.objects.raw(query)
       
   175 
       
   176     **Don't.**
       
   177 
       
   178     Using the ``params`` list completely protects you from `SQL injection
       
   179     attacks`__, a common exploit where attackers inject arbitrary SQL into
       
   180     your database. If you use string interpolation, sooner or later you'll
       
   181     fall victim to SQL injection. As long as you remember to always use the
       
   182     ``params`` list you'll be protected.
       
   183 
       
   184 __ http://en.wikipedia.org/wiki/SQL_injection
       
   185 
       
   186 Executing custom SQL directly
       
   187 =============================
       
   188 
       
   189 Sometimes even :meth:`Manager.raw` isn't quite enough: you might need to
       
   190 perform queries that don't map cleanly to models, or directly execute
       
   191 ``UPDATE``, ``INSERT``, or ``DELETE`` queries.
       
   192 
       
   193 In these cases, you can always access the database directly, routing around
       
   194 the model layer entirely.
       
   195 
       
   196 The object ``django.db.connection`` represents the
       
   197 default database connection, and ``django.db.transaction`` represents the
       
   198 default database transaction. To use the database connection, call
       
   199 ``connection.cursor()`` to get a cursor object. Then, call
       
   200 ``cursor.execute(sql, [params])`` to execute the SQL and ``cursor.fetchone()``
       
   201 or ``cursor.fetchall()`` to return the resulting rows. After performing a data
       
   202 changing operation, you should then call
       
   203 ``transaction.commit_unless_managed()`` to ensure your changes are committed
       
   204 to the database. If your query is purely a data retrieval operation, no commit
       
   205 is required. For example::
       
   206 
       
   207     def my_custom_sql():
       
   208         from django.db import connection, transaction
       
   209         cursor = connection.cursor()
       
   210 
       
   211         # Data modifying operation - commit required
       
   212         cursor.execute("UPDATE bar SET foo = 1 WHERE baz = %s", [self.baz])
       
   213         transaction.commit_unless_managed()
       
   214 
       
   215         # Data retrieval operation - no commit required
       
   216         cursor.execute("SELECT foo FROM bar WHERE baz = %s", [self.baz])
       
   217         row = cursor.fetchone()
       
   218 
       
   219         return row
       
   220 
       
   221 If you are using more than one database you can use
       
   222 ``django.db.connections`` to obtain the connection (and cursor) for a
       
   223 specific database. ``django.db.connections`` is a dictionary-like
       
   224 object that allows you to retrieve a specific connection using it's
       
   225 alias::
       
   226 
       
   227     from django.db import connections
       
   228     cursor = connections['my_db_alias'].cursor()
       
   229 
       
   230 .. _transactions-and-raw-sql:
       
   231 
       
   232 Transactions and raw SQL
       
   233 ------------------------
       
   234 If you are using transaction decorators (such as ``commit_on_success``) to
       
   235 wrap your views and provide transaction control, you don't have to make a
       
   236 manual call to ``transaction.commit_unless_managed()`` -- you can manually
       
   237 commit if you want to, but you aren't required to, since the decorator will
       
   238 commit for you. However, if you don't manually commit your changes, you will
       
   239 need to manually mark the transaction as dirty, using
       
   240 ``transaction.set_dirty()``::
       
   241 
       
   242     @commit_on_success
       
   243     def my_custom_sql_view(request, value):
       
   244         from django.db import connection, transaction
       
   245         cursor = connection.cursor()
       
   246 
       
   247         # Data modifying operation
       
   248         cursor.execute("UPDATE bar SET foo = 1 WHERE baz = %s", [value])
       
   249 
       
   250         # Since we modified data, mark the transaction as dirty
       
   251         transaction.set_dirty()
       
   252 
       
   253         # Data retrieval operation. This doesn't dirty the transaction,
       
   254         # so no call to set_dirty() is required.
       
   255         cursor.execute("SELECT foo FROM bar WHERE baz = %s", [value])
       
   256         row = cursor.fetchone()
       
   257 
       
   258         return render_to_response('template.html', {'row': row})
       
   259 
       
   260 The call to ``set_dirty()`` is made automatically when you use the Django ORM
       
   261 to make data modifying database calls. However, when you use raw SQL, Django
       
   262 has no way of knowing if your SQL modifies data or not. The manual call to
       
   263 ``set_dirty()`` ensures that Django knows that there are modifications that
       
   264 must be committed.
       
   265 
       
   266 Connections and cursors
       
   267 -----------------------
       
   268 
       
   269 ``connection`` and ``cursor`` mostly implement the standard `Python DB-API`_
       
   270 (except when it comes to :doc:`transaction handling </topics/db/transactions>`).
       
   271 If you're not familiar with the Python DB-API, note that the SQL statement in
       
   272 ``cursor.execute()`` uses placeholders, ``"%s"``, rather than adding parameters
       
   273 directly within the SQL. If you use this technique, the underlying database
       
   274 library will automatically add quotes and escaping to your parameter(s) as
       
   275 necessary. (Also note that Django expects the ``"%s"`` placeholder, *not* the
       
   276 ``"?"`` placeholder, which is used by the SQLite Python bindings. This is for
       
   277 the sake of consistency and sanity.)
       
   278 
       
   279 .. _Python DB-API: http://www.python.org/dev/peps/pep-0249/