Add connection_parameters and ODBC_OPTION_DRIVER_COMPLETE option. #98
Conversation
Right now this is just a trivial wrapper for the backend and connection string and no extra functionality but it will be extended with connection options support in the subsequent commits. No changes in behaviour, this is a pure refactoring. Signed-off-by: Vadim Zeitlin <vz-soci@zeitlins.org>
No real changes, but just use "class" for consistency. Signed-off-by: Vadim Zeitlin <vz-soci@zeitlins.org>
…string. There are still no real changes as only the connect string part of the connection parameters is currently used, but this will make it possible to use other information stored in connection_parameters from the backend code in the future. Also replace session::lastFactory_ and lastConnectString_ with a single lastConnectParameters_ making the code slightly shorter and simpler.
It should be to connections.html, not the (non existent) basics.html.
Document the new session ctor taking connection_parameters and the class itself in the references page.
Allow passing arbitrary options via connection_parameters and add support for an option allowing to specify the "driver completion" mode, i.e. the degree of interactivity of the driver, in the ODBC backend. By default now do ask the user if the DSN doesn't include the user name and/or password for the connection but the old behaviour is still available if ODBC_OPTION_DRIVER_COMPLETE option is set to SQL_DRIVER_NOPROMPT before opening the session.
ghost
assigned 
|
I checked your submission on Windows 8 building SOCI with VS2012 and tested ODBC (MySQL, PostgreSQL, SQL Server). The DB2 builds fine and the three test cases run fine as well, so DB2 is not affected. I assigned you to this pull request, so if no objections are raised, feel free to merge it. I'm looking at the code now. |
|
@vadz These are the only comments I have. I did have a few thoughts where we could improve it, but that would touch the more advanced solutions we have discussed on the list. There is no point doing it now, no time. So, it all looks very good to me. Let's wait for votes from others. |
| : once(this), prepare(this), query_transformation_(NULL), logStream_(NULL), | ||
| lastConnectParameters_(parameters), | ||
| uppercaseColumnNames_(false), | ||
| isFromPool_(false), pool_(NULL), backEnd_(NULL) |
There was a problem hiding this comment.
Just looked at what unkle clang says in the build 215.2:
/home/travis/build/SOCI/soci/src/core/session.cpp:46:27: warning: field 'pool_'
will be initialized after field 'backEnd_' [-Wreorder]
isFromPool_(false), pool_(NULL), backEnd_(NULL)
^
There was a problem hiding this comment.
I'll fix this, thanks for noticing. I wonder why didn't g++ warn about it, normally it gives these warnings too. Unfortunately I have no idea how to specify the compilation flags with cmake...
BTW, it would be nice to get rid of many other warnings currently given by g++...
There was a problem hiding this comment.
You may take a look at src/cmake/SociConfig.cmake, feel free to update the flags.
I thoroughly agree, we should aim clean compilation, so let's not forget it #100
These commits add
connection_parametersclass which is a simple container for the backend, connection string and possibly some options, i.e. simply name/value pairs. It changes the factory classes to acceptconnection_parametersinstead of a simple string for future extensibility and updates the ODBC backend to honour the value ofODBC_OPTION_DRIVER_COMPLETEoption to allow specifying the "driver completion" parameter ofSQLDriverConnect().