Exceptions and Errors

Python uses exceptions to indicate an error has happened. The SQLite library uses integer error codes. APSW maps between the two systems as needed. Exceptions raised in Python code called by SQLite will have that exception present when control returns to Python, and SQLite will understand that an error occurred.

Chaining

When an error is reported to SQLite, it may take further actions. For example errors in VFS can result in error recovery attempts, while an error in a window function step method will result in the final method being called to do clean up. Your code implementing those could also have additional exceptions.

When multiple exceptions occur in the same SQLite control flow then they will be chained. Python’s traceback printing code will show all the exceptions.

Unraisable

There are a few places where it is not possible for a Python exception to be reported to SQLite as an error, and Python C code does not allow destructors to report exceptions. These exceptions are reported via sys.unraisablehook(), and if that is not present then sys.excepthook().

sqlite3_log is also called so that you will have the context of when the exception happened relative to the errors SQLite is logging.

Exception Classes

exception apsw.Error

This is the base for APSW exceptions.

Error.result

For exceptions corresponding to SQLite error codes codes this attribute is the numeric error code.

Error.extendedresult

APSW runs with extended result codes turned on. This attribute includes the detailed code.

As an example, if SQLite issued a read request and the system returned less data than expected then result would have the value SQLITE_IOERR while extendedresult would have the value SQLITE_IOERR_SHORT_READ.

Error.error_offset

The location of the error in the SQL when encoded in UTF-8. The value is from sqlite3_error_offset, and will be -1 when a specific token in the input is not the cause.

APSW specific exceptions

The following exceptions happen when APSW detects various problems.

exception apsw.ThreadingViolationError

You have used an object concurrently in two threads. For example you may try to use the same cursor in two different threads at the same time, or tried to close the same connection in two threads at the same time.

You can also get this exception by using a cursor as an argument to itself (eg as the input data for Cursor.executemany()). Cursors can only be used for one thing at a time.

exception apsw.ForkingViolationError

See apsw.fork_checker().

exception apsw.IncompleteExecutionError

You have tried to start a new SQL execute call before executing all the previous ones. See the execution model for more details.

exception apsw.ConnectionNotClosedError

This exception is no longer generated. It was required in earlier releases due to constraints in threading usage with SQLite.

exception apsw.ConnectionClosedError

You have called Connection.close() and then continued to use the Connection or associated cursors.

exception apsw.CursorClosedError

You have called Cursor.close() and then tried to use the cursor.

exception apsw.BindingsError

There are several causes for this exception. When using tuples, an incorrect number of bindings where supplied:

cursor.execute("select ?,?,?", (1,2))     # too few bindings
cursor.execute("select ?,?,?", (1,2,3,4)) # too many bindings

You are using named bindings, but not all bindings are named. You should either use entirely the named style or entirely numeric (unnamed) style:

cursor.execute("select * from foo where x=:name and y=?")

exception apsw.ExecutionCompleteError

Execution of the statements is complete and cannot be run further.

exception apsw.ExecTraceAbort

The execution tracer returned False so execution was aborted.

exception apsw.ExtensionLoadingError

An error happened loading an extension.

exception apsw.VFSNotImplementedError

A call cannot be made to an inherited Virtual File System (VFS) method as the VFS does not implement the method.

exception apsw.VFSFileClosedError

The VFS file is closed so the operation cannot be performed.

exception apsw.NoFTS5Error

The FTS5 extension is not present in SQLite.

exception apsw.InvalidContextError

Context is no longer valid. Examples include using an IndexInfo outside of the VTTable.BestIndexObject() method, a registered FTS5Tokenizer when the underlying tokenizer has been deleted/replaced, or Connection.vtab_config() when not inside VTModule.Create().

SQLite Exceptions

The following lists which Exception classes correspond to which SQLite error codes.

General Errors

exception apsw.SQLError

SQLITE_ERROR. The standard error code, unless a more specific one is applicable.

exception apsw.MismatchError

SQLITE_MISMATCH. Data type mismatch. For example a rowid or integer primary key must be an integer.

exception apsw.NotFoundError

SQLITE_NOTFOUND. Returned when various internal items were not found such as requests for non-existent system calls or file controls.

Internal Errors

exception apsw.InternalError

SQLITE_INTERNAL. (No longer used) Internal logic error in SQLite.

exception apsw.ProtocolError

SQLITE_PROTOCOL. (No longer used) Database lock protocol error.

exception apsw.MisuseError

SQLITE_MISUSE. SQLite library used incorrectly - typically similar to ValueError in Python. Examples include not having enough flags when opening a connection (eg not including a READ or WRITE flag), or out of spec such as registering a function with more than 127 parameters.

exception apsw.RangeError

SQLITE_RANGE. (Cannot be generated using APSW). 2nd parameter to sqlite3_bind out of range

Permissions Etc

exception apsw.PermissionsError

SQLITE_PERM. Access permission denied by the operating system.

exception apsw.ReadOnlyError

SQLITE_READONLY. Attempt to write to a readonly database.

exception apsw.CantOpenError

SQLITE_CANTOPEN. Unable to open the database file.

exception apsw.AuthError

SQLITE_AUTH. Authorization denied.

Abort/Busy Etc

exception apsw.AbortError

SQLITE_ABORT. Callback routine requested an abort.

exception apsw.BusyError

SQLITE_BUSY. The database file is locked. Use Connection.set_busy_timeout() to change how long SQLite waits for the database to be unlocked or Connection.set_busy_handler() to use your own handler.

exception apsw.LockedError

SQLITE_LOCKED. Shared cache lock.

exception apsw.InterruptError

SQLITE_INTERRUPT. Operation terminated by sqlite3_interrupt - use Connection.interrupt().

exception apsw.SchemaChangeError

SQLITE_SCHEMA. The database schema changed. A prepared statement becomes invalid if the database schema was changed. Behind the scenes SQLite reprepares the statement. Another or the same Connection may change the schema again before the statement runs. SQLite will retry before giving up and returning this error.

exception apsw.ConstraintError

SQLITE_CONSTRAINT. Abort due to constraint violation.

Memory/Disk

exception apsw.NoMemError

SQLITE_NOMEM. A memory

allocation failed.

exception apsw.IOError

SQLITE_IOERR. A disk I/O error occurred. The extended error code will give more detail.

exception apsw.CorruptError

SQLITE_CORRUPT. The database disk image appears to be a SQLite database but the values inside are inconsistent.

exception apsw.FullError

SQLITE_FULL. The disk appears to be full.

exception apsw.TooBigError

SQLITE_TOOBIG. String or BLOB exceeds size limit. You can change the limits using Connection.limit().

exception apsw.NoLFSError

SQLITE_NOLFS. SQLite has attempted to use a feature not supported by the operating system such as large file support.

exception apsw.EmptyError

SQLITE_EMPTY. Not currently used.

exception apsw.FormatError

SQLITE_FORMAT. (No longer used) Auxiliary database format error.

exception apsw.NotADBError

SQLITE_NOTADB. File opened that is not a database file. SQLite has a header on database files to verify they are indeed SQLite databases.

Augmented stack traces

When an exception occurs, Python does not include frames from non-Python code (ie the C code called from Python). This can make it more difficult to work out what was going on when an exception occurred for example when there are callbacks to collations, functions or virtual tables, triggers firing etc.

This is an example showing the difference between the tracebacks you would have got with earlier versions of apsw and the augmented traceback:

import apsw

def myfunc(x):
  1/0

con=apsw.Connection(":memory:")
con.create_scalar_function("foo", myfunc)
con.create_scalar_function("fam", myfunc)
cursor=con.cursor()
cursor.execute("create table bar(x,y,z);insert into bar values(1,2,3)")
cursor.execute("select foo(1) from bar")

Original Traceback
Traceback (most recent call last): File "t.py", line 11, in <module> cursor.execute("select foo(1) from bar") File "t.py", line 4, in myfunc 1/0 ZeroDivisionError: integer division or modulo by zero

Augmented Traceback
Traceback (most recent call last): File "t.py", line 11, in <module> cursor.execute("select foo(1) from bar") File "apsw.c", line 3412, in resetcursor File "apsw.c", line 1597, in user-defined-scalar-foo File "t.py", line 4, in myfunc 1/0 ZeroDivisionError: integer division or modulo by zero

In the original traceback you can’t even see that code in apsw was involved. The augmented traceback shows that there were indeed two function calls within apsw and gives you line numbers should you need to examine the code. Also note how you are told that the call was in user-defined-scalar-foo (ie you can tell which function was called.)

But wait, there is more!!! In order to further aid troubleshooting, the augmented stack traces make additional information available. Each frame in the traceback has local variables defined with more information. You can use apsw.ext.print_augmented_traceback() to print an exception with the local variables.

Here is a far more complex example from some virtual tables code I was writing. The BestIndex method in my code had returned an incorrect value. The augmented traceback includes local variables. I can see what was passed in to my method, what I returned and which item was erroneous. The original traceback is almost completely useless!

Original traceback:

Traceback (most recent call last):
  File "tests.py", line 1387, in testVtables
    cursor.execute(allconstraints)
TypeError: Bad constraint (#2) - it should be one of None, an integer or a tuple of an integer and a boolean

Augmented traceback with local variables:

Traceback (most recent call last):
  File "tests.py", line 1387, in testVtables
    cursor.execute(allconstraints)
                VTable =  __main__.VTable
                   cur =  <apsw.Cursor object at 0x988f30>
                     i =  10
                  self =  testVtables (__main__.APSW)
        allconstraints =  select rowid,* from foo where rowid>-1000 ....

  File "apsw.c", line 4050, in Cursor_execute.sqlite3_prepare
            Connection =  <apsw.Connection object at 0x978800>
             statement =  select rowid,* from foo where rowid>-1000 ....

  File "apsw.c", line 2681, in VirtualTable.xBestIndex
                  self =  <__main__.VTable instance at 0x98d8c0>
                  args =  (((-1, 4), (0, 32), (1, 8), (2, 4), (3, 64)), ((2, False),))
                result =  ([4, (3,), [2, False], [1], [0]], 997, u'\xea', False)

  File "apsw.c", line 2559, in VirtualTable.xBestIndex.result_constraint
               indices =  [4, (3,), [2, False], [1], [0]]
                  self =  <__main__.VTable instance at 0x98d8c0>
                result =  ([4, (3,), [2, False], [1], [0]], 997, u'\xea', False)
            constraint =  (3,)

TypeError: Bad constraint (#2) - it should be one of None, an integer or a tuple of an integer and a boolean