1
- fix to force the finalisation of queries in progress for transactions causes
2
errors; queries are now finalised twice, the second from basic_statement's
3
dtor, which causes a segfault. We could:
4
- keep a set of force-finalised sqlite3_stmt pointers in the database which
5
we use to check queries against before finalising them to make sure we
6
don't finalise them a second time
7
- an efficient implementation, but not very OO
8
- keep a map of active queries in the database (using the sqlite3_stmt
9
pointer as the key), so that we can obtain the query and tell it to
11
- this seems like a messy and complicated implementation
12
- would be improved by having a wrapper around the query handle
4
- rename _bind_index and _column_index to _next_*
14
6
- turn on extended errcodes in open() and handle them in sqlite_error
16
- make basic_statement and database keep a shared pointer to the database handle
17
so the classes can be made copyable. The wrappers around the handle
18
(implemented in sqlite::detail) can clean them up after use. This will also
19
make the implementation of rows (to get round the forced non-dependency of
20
rows on querys) a little easier to swallow.
21
- A similar wrapper should be created for statement handles, making
22
basic_statements, querys and commands copyable. Could weak_ptrs to these
23
also be used in the database's list active querys?
8
- query::prepare() isn't being called during construction (form
9
basic_statement's constructor)
25
11
- add columns() to row that returns a boost::tuple of various types so multple
26
12
columns can be fetched at once (look in to using BOOST_PP_ITERATE macro)
28
14
- use sqlite3_db_mutex() to provide extended error information during
29
sqlite_error construction. The genreeal procedure would be to lock the db
15
sqlite_error construction. The general procedure would be to lock the db
30
16
mutex, perform some sqlite3 command, check the error code, throw an
31
17
sqlite_error (whilst obtaining extended error info) and then unlock the db
32
18
mutex. Two options:
34
20
- a templated safe-calling object (passing the comman's arg types as
35
21
template params) may be overkill
26
- make basic_statement and database keep a shared pointer to the database handle
27
so the classes can be made copyable. The wrappers around the handle
28
(implemented in sqlite::detail) can clean them up after use. This will also
29
make the implementation of rows (to get round the forced non-dependency of
30
rows on querys) a little easier to swallow.
31
- A similar wrapper should be created for statement handles, making
32
basic_statements, querys and commands copyable. Could weak_ptrs to these
33
also be used in the database's list active querys?
37
35
- expand sqlite_error - perhaps use boost::system_error (see
38
36
boost/asio/error.hpp for an example of extending system_error)
46
44
this matter? they can't access database._handle anyway!
47
45
potential incompatibility when linking to libraries that also link
50
- query::prepare() isn't being called during construction (form
51
basic_statement's constructor)
added
removed
TODO
