psycopg2/doc/src/faq.rst

82 lines
4.3 KiB
ReStructuredText
Raw Normal View History

Frequently Asked Questions
==========================
.. sectionauthor:: Daniele Varrazzo <daniele.varrazzo@gmail.com>
Here are a few gotchas you may encounter using :mod:`psycopg2`. Feel free to
suggest new entries!
.. cssclass:: faq
Why does :mod:`!psycopg2` leave database sessions "idle in transaction"?
Psycopg normally starts a new transaction the first time a query is
executed, e.g. calling :meth:`cursor.execute`, even if the command is a
:sql:`SELECT`. The transaction is not closed until an explicit
:meth:`~connection.commit` or :meth:`~connection.rollback`.
If you are writing a long-living program, you should probably ensure to
call one of the transaction closing methods before leaving the connection
unused for a long time (which may also be a few seconds, depending on the
concurrency level in your database). Alternatively you can use a
connection in :ref:`autocommit <autocommit>` mode to avoid a new
transaction to be started at the first command.
Why does :meth:`!cursor.execute` raise the exception *can't adapt*?
Psycopg converts Python objects in a SQL string representation by looking
at the object class. The exception is raised when you are trying to pass
as query parameter an object for which there is no adapter registered for
its class. See :ref:`adapting-new-types` for informations.
I can't pass an integer or a float parameter to my query: it says *a number is required*, but *it is* a number!
In your query string, you always have to use ``%s`` placeholders,
event when passing a number. All Python objects are converted by Psycopg
in their SQL representation, so they get passed to the query as strings.
See :ref:`query-parameters`. ::
>>> cur.execute("INSERT INTO numbers VALUES (%d)", (42,)) # WRONG
>>> cur.execute("INSERT INTO numbers VALUES (%s)", (42,)) # correct
I try to execute a query but it fails with the error *not all arguments converted during string formatting* (or *object does not support indexing*). Why?
Psycopg always require positional arguments to be passed as a tuple, even
when the query takes a single parameter. And remember that to make a
single item tuple in Python you need a comma! See :ref:`query-parameters`.
::
>>> cur.execute("INSERT INTO foo VALUES (%s)", "bar") # WRONG
>>> cur.execute("INSERT INTO foo VALUES (%s)", ("bar")) # WRONG
>>> cur.execute("INSERT INTO foo VALUES (%s)", ("bar",)) # correct
I receive the error *current transaction is aborted, commands ignored until end of transaction block* and can't do anything else!
There was a problem *in the previous* command to the database, which
resulted in an error. The database will not recover automatically from
this condition: you must run a :meth:`~connection.rollback` before sending
new commands to the session (if this seems too harsh, remember that
PostgreSQL supports nested transactions using the |SAVEPOINT|_ command).
.. |SAVEPOINT| replace:: :sql:`SAVEPOINT`
.. _SAVEPOINT: http://www.postgresql.org/docs/8.4/static/sql-savepoint.html
Why do i get the error *current transaction is aborted, commands ignored until end of transaction block* when I use :mod:`!multiprocessing` (or any other forking system) and not when use :mod:`!threading`?
Psycopg's connections can't be shared across processes (but are thread
safe). If you are forking the Python process ensure to create a new
connection in each forked child.
My database is Unicode, but I receive all the strings as UTF-8 :class:`str`. Can I receive :class:`unicode` objects instead?
The following magic formula will do the trick::
psycopg2.extensions.register_type(psycopg2.extensions.UNICODE)
psycopg2.extensions.register_type(psycopg2.extensions.UNICODEARRAY)
See :ref:`unicode-handling` for the gory details.
I can't compile :mod:`!psycopg2`: the compiler says *error: Python.h: No such file or directory*. What am I missing?
You need to install a Python development package: it is usually called
``python-dev``.
I can't compile :mod:`!psycopg2`: the compiler says *error: libpq-fe.h: No such file or directory*. What am I missing?
You need to install the development version of the libpq: the package is
usually called ``libpq-dev``.