Executing SQL statements is the primary way in which a Python application
communicates with Oracle Database. Statements are executed using the methods
Cursor.executemany(). Statements include
queries, Data Manipulation Language (DML), and Data Definition Language (DDL).
A few other specialty statements can also be executed.
PL/SQL statements are discussed in PL/SQL Execution. Other chapters contain information on specific data types and features. See Batch Statement Execution and Bulk Loading, Using CLOB and BLOB Data, Working with the JSON Data Type, and Working with XMLTYPE.
cx_Oracle can be used to execute individual statements, one at a time. It does
not read SQL*Plus “.sql” files. To read SQL files, use a technique like the one
RunSqlScript() in samples/SampleEnv.py
SQL statements should not contain a trailing semicolon (“;”) or forward slash (“/”). This will fail:
cur.execute("select * from MyTable;")
This is correct:
cur.execute("select * from MyTable")
Queries (statements beginning with SELECT or WITH) can only be executed using
Cursor.execute(). Rows can then be iterated over, or can be
fetched using one of the methods
Cursor.fetchall(). There is a
default type mapping to Python types that can be
Interpolating or concatenating user data with SQL statements, for example
cur.execute("SELECT * FROM mytab WHERE mycol = '" + myvar + "'"), is a security risk
and impacts performance. Use bind variables instead. For
cur.execute("SELECT * FROM mytab WHERE mycol = :mybv", mybv=myvar).
Cursor.execute(), the cursor is returned as a convenience. This
allows code to iterate over rows like:
cur = connection.cursor() for row in cur.execute("select * from MyTable"): print(row)
Rows can also be fetched one at a time using the method
cur = connection.cursor() cur.execute("select * from MyTable") while True: row = cur.fetchone() if row is None: break print(row)
cur = connection.cursor() cur.execute("select * from MyTable") numRows = 10 while True: rows = cur.fetchmany(numRows) if not rows: break for row in rows: print(row)
If all of the rows need to be fetched, and can be contained in memory, the
Cursor.fetchall() can be used.
cur = connection.cursor() cur.execute("select * from MyTable") rows = cur.fetchall() for row in rows: print(row)
A cursor may be used to execute multiple statements. Once it is no longer
needed, it should be closed by calling
close() in order to
reclaim resources in the database. It will be closed automatically when the
variable referencing it goes out of scope (and no further references are
retained). One other way to control the lifetime of a cursor is to use a “with”
block, which ensures that a cursor is closed once the block is completed. For
with connection.cursor() as cursor: for row in cursor.execute("select * from MyTable"): print(row)
This code ensures that, once the block is completed, the cursor is closed and
resources have been reclaimed by the database. In addition, any attempt to use
cursor outside of the block will simply fail.
Tuning Fetch Performance¶
For best performance, the cx_Oracle
Cursor.arraysize value should be set
Cursor.execute(). The default value is 100. For queries
that return a large number of rows, increasing
arraysize can improve
performance because it reduces the number of round-trips to the database.
However increasing this value increases the amount of memory required. The best
value for your system depends on factors like your network speed, the query row
size, and available memory. An appropriate value can be found by experimenting
with your application.
Regardless of which fetch method is used to get rows, internally all rows are
fetched in batches corresponding to the value of
arraysize. The size does
not affect how, or when, rows are returned to your application (other than being
used as the default size for
Cursor.fetchmany()). It does not limit the
minimum or maximum number of rows returned by a query.
Along with tuning
arraysize, make sure your SQL statements are optimal and avoid
selecting columns that are not required by the application. For queries that do
not need to fetch all data, use a row limiting clause to
reduce the number of rows processed by the database.
An example of setting
cur = connection.cursor() cur.arraysize = 500 for row in cur.execute("select * from MyTable"): print(row)
One place where increasing
arraysize is particularly useful is in copying
data from one database to another:
# setup cursors sourceCursor = sourceConnection.cursor() sourceCursor.arraysize = 1000 targetCursor = targetConnection.cursor() targetCursor.arraysize = 1000 # perform fetch and bulk insertion sourceCursor.execute("select * from MyTable") while True: rows = sourceCursor.fetchmany() if not rows: break targetCursor.executemany("insert into MyTable values (:1, :2)", rows) targetConnection.commit()
If you know that a query returns a small number of rows then you should reduce
the value of
arraysize. For example if you are fetching only one row, then
arraysize to 1:
cur = connection.cursor() cur.arraysize = 1 cur.execute("select * from MyTable where id = 1"): row = cur.fetchone() print(row)
In cx_Oracle, the
arraysize value is only examined when a statement is
executed the first time. To change the
arraysize for a repeated statement,
create a new cursor:
array_sizes = (10, 100, 1000) for size in array_sizes: cursor = connection.cursor() cursor.arraysize = size start = time.time() cursor.execute(sql).fetchall() elapsed = time.time() - start print("Time for", size, elapsed, "seconds")
Query Column Metadata¶
After executing a query, the column metadata such as column names and data types
can be obtained using
cur = connection.cursor() cur.execute("select * from MyTable") for column in cur.description: print(column)
This could result in metadata like:
('ID', <class 'cx_Oracle.NUMBER'>, 39, None, 38, 0, 0) ('NAME', <class 'cx_Oracle.STRING'>, 20, 20, None, None, 1)
Fetch Data Types¶
The following table provides a list of all of the data types that cx_Oracle knows how to fetch. The middle column gives the type that is returned in the query metadata. The last column gives the type of Python object that is returned by default. Python types can be changed with Output Type Handlers.
|Oracle Database Type||cx_Oracle Type||Default Python type|
|INTERVAL DAY TO SECOND||
||float or int |
|TIMESTAMP WITH LOCAL TIME ZONE||
|TIMESTAMP WITH TIME ZONE||
|||(1, 2) In Python 2 these are fetched as unicode objects.|
|||If the precision and scale obtained from query column metadata indicate that the value can be expressed as an integer, the value will be returned as an int. If the column is unconstrained (no precision and scale specified), the value will be returned as a float or an int depending on whether the value itself is an integer. In all other cases the value is returned as a float. Note that in Python 2, values returned as integers will be int or long depending on the size of the integer.|
|||(1, 2) The timestamps returned are naive timestamps without any time zone information present.|
|||(1, 2) In Python 2 these are identical to str objects since Python 2 doesn’t have a native bytes object.|
|||These include all user-defined types such as VARRAY, NESTED TABLE, etc.|
Changing Fetched Data Types with Output Type Handlers¶
Sometimes the default conversion from an Oracle Database type to a Python type
must be changed in order to prevent data loss or to fit the purposes of the
Python application. In such cases, an output type handler can be specified for
queries. Output type handlers do not affect values returned from
Output type handlers can be specified on the
connection or on the
cursor. If specified on the cursor, fetch type handling is
only changed on that particular cursor. If specified on the connection, all
cursors created by that connection will have their fetch type handling changed.
The output type handler is expected to be a function with the following signature:
handler(cursor, name, defaultType, size, precision, scale)
The parameters are the same information as the query column metadata found in
Cursor.description. The function is called once for each column that is
going to be fetched. The function is expected to return a
variable object (generally by a call to
or the value
None. The value
None indicates that the default type
should be used.
Fetched Number Precision¶
One reason for using an output type handler is to ensure that numeric precision is not lost when fetching certain numbers. Oracle Database uses decimal numbers and these cannot be converted seamlessly to binary number representations like Python floats. In addition, the range of Oracle numbers exceeds that of floating point numbers. Python has decimal objects which do not have these limitations and cx_Oracle knows how to perform the conversion between Oracle numbers and Python decimal values if directed to do so.
The following code sample demonstrates the issue:
cur = connection.cursor() cur.execute("create table test_float (X number(5, 3))") cur.execute("insert into test_float values (7.1)") connection.commit() cur.execute("select * from test_float") val, = cur.fetchone() print(val, "* 3 =", val * 3)
7.1 * 3 = 21.299999999999997
Using Python decimal objects, however, there is no loss of precision:
import decimal def NumberToDecimal(cursor, name, defaultType, size, precision, scale): if defaultType == cx_Oracle.NUMBER: return cursor.var(decimal.Decimal, arraysize=cursor.arraysize) cur = connection.cursor() cur.outputtypehandler = NumberToDecimal cur.execute("select * from test_float") val, = cur.fetchone() print(val, "* 3 =", val * 3)
7.1 * 3 = 21.3
decimal.Decimal converter gets called with the string
representation of the Oracle number. The output from
returned in the output tuple.
Changing Query Results with Outconverters¶
cx_Oracle “outconverters” can be used with output type handlers to change returned data.
For example, to make queries return empty strings instead of NULLs:
def OutConverter(value): if value is None: return '' return value def OutputTypeHandler(cursor, name, defaultType, size, precision, scale): if defaultType in (cx_Oracle.STRING, cx_Oracle.FIXED_CHAR): return cursor.var(str, size, cur.arraysize, outconverter=OutConverter) connection.outputtypehandler = OutputTypeHandler
Scrollable cursors enable applications to move backwards, forwards, to skip rows, and to move to a particular row in a query result set. The result set is cached on the database server until the cursor is closed. In contrast, regular cursors are restricted to moving forward.
A scrollable cursor is created by setting the parameter
when creating the cursor. The method
Cursor.scroll() is used to move to
different locations in the result set.
cursor = connection.cursor(scrollable=True) cursor.execute("select * from ChildTable order by ChildId") cursor.scroll(mode="last") print("LAST ROW:", cursor.fetchone()) cursor.scroll(mode="first") print("FIRST ROW:", cursor.fetchone()) cursor.scroll(8, mode="absolute") print("ROW 8:", cursor.fetchone()) cursor.scroll(6) print("SKIP 6 ROWS:", cursor.fetchone()) cursor.scroll(-4) print("SKIP BACK 4 ROWS:", cursor.fetchone())
Query data is commonly broken into one or more sets:
- To give an upper bound on the number of rows that a query has to process, which can help improve database scalability.
- To perform ‘Web pagination’ that allows moving from one set of rows to a next, or previous, set on demand.
- For fetching of all data in consecutive small sets for batch processing. This happens because the number of records is too large for Python to handle at one time.
The latter can be handled by calling
Cursor.fetchmany() with one
execution of the SQL query.
‘Web pagination’ and limiting the maximum number of rows are discussed in this section. For each ‘page’ of results, a SQL query is executed to get the appropriate set of rows from a table. Since the query may be executed more than once, make sure to use bind variables for row numbers and row limits.
Oracle Database 12c SQL introduced an
FETCH clause which is
similar to the
LIMIT keyword of MySQL. In Python you can fetch a set of
myoffset = 0 // do not skip any rows (start at row 1) mymaxnumrows = 20 // get 20 rows sql = """SELECT last_name FROM employees ORDER BY last_name OFFSET :offset ROWS FETCH NEXT :maxnumrows ROWS ONLY""" cur = connection.cursor() for row in cur.execute(sql, offset=myoffset, maxnumrows=mymaxnumrows): print(row)
In applications where the SQL query is not known in advance, this method
sometimes involves appending the
OFFSET clause to the ‘real’ user query. Be
very careful to avoid SQL injection security issues.
For Oracle Database 11g and earlier there are several alternative ways to limit the number of rows returned. The old, canonical paging query is:
SELECT * FROM (SELECT a.*, ROWNUM AS rnum FROM (YOUR_QUERY_GOES_HERE -- including the order by) a WHERE ROWNUM <= MAX_ROW) WHERE rnum >= MIN_ROW
MIN_ROW is the row number of first row and
MAX_ROW is the row
number of the last row to return. For example:
SELECT * FROM (SELECT a.*, ROWNUM AS rnum FROM (SELECT last_name FROM employees ORDER BY last_name) a WHERE ROWNUM <= 20) WHERE rnum >= 1
This always has an ‘extra’ column, here called RNUM.
An alternative and preferred query syntax for Oracle Database 11g uses the
ROW_NUMBER() function. For example to get the 1st to 20th names the
SELECT last_name FROM (SELECT last_name, ROW_NUMBER() OVER (ORDER BY last_name) AS myr FROM employees) WHERE myr BETWEEN 1 and 20
Make sure to use bind variables for the upper and lower limit values.
Querying Corrupt Data¶
If queries fail with the error “codec can’t decode byte” when you select data, then:
connection = cx_Oracle.connect("hr", userpwd, "dbhost.example.com/orclpdb1", encoding="UTF-8", nencoding="UTF-8")
Check for corrupt data in the database.
If data really is corrupt, you can pass options to the internal decode() used by
cx_Oracle to allow it to be selected and prevent the whole query failing. Do
this by creating an outputtypehandler and setting
encodingErrors. For example to replace corrupt characters in character
def OutputTypeHandler(cursor, name, defaultType, size, precision, scale): if defaultType == cx_Oracle.STRING: return cursor.var(defaultType, size, arraysize=cursor.arraysize, encodingErrors="replace") cursor.outputtypehandler = OutputTypeHandler cursor.execute("select column1, column2 from SomeTableWithBadData")
Other codec behaviors can be chosen for
encodingErrors, see Error Handlers.
INSERT and UPDATE Statements¶
SQL Data Manipulation Language statements (DML) such as INSERT and UPDATE can easily be executed with cx_Oracle. For example:
cur = connection.cursor() cur.execute("insert into MyTable values (:idbv, :nmbv)", [1, "Fredico"])
Do not concatenate or interpolate user data into SQL statements. See Using Bind Variables instead.
See Transaction Management for best practices on committing and rolling back data changes.
Oracle requires a type, even for null values. When you pass the value None, then cx_Oracle assumes the type is STRING. If this is not the desired type, you can explicitly set it. For example, to insert a null Oracle Spatial SDO_GEOMETRY object:
typeObj = connection.gettype("SDO_GEOMETRY") cur = connection.cursor() cur.setinputsizes(typeObj) cur.execute("insert into sometable values (:1)", [None])