ODBC driver changes

To Whom It May Concern:

Please consider making the following changes to the psqlODBC driver.

Items 1 through 3 (the VARBIT and JSON-related changes) should be

non-controversial – I just think someone forgot to put them in the code.

1. In pgtypes.h,   please add these definitions:

#define PG_TYPE_JSON   114

#define PG_TYPE_VARBIT 1562

You might also consider adding this if it is correct:

#define PG_TYPE_JSONB  3802

2. In info.c,    add this line immediately after the line containing the

only occurrence of PG_TYPE_BIT  and right before the “break” line in the file:

                case PG_TYPE_VARBIT:

3. In columninfo.c, add these lines immediately after the line containing the

only occurrence of PG_TYPE_TIME_WITH_TMZONE and right before the “break”

line in the file:

                case PG_TYPE_BIT:

                case PG_TYPE_VARBIT:

4. In info.c,  in the function PGAPI_ProcedureColumns()  at line 5121 (in the version

15.00 sources)  the following code is added to the query to restrict the answer

to executable items:

                appendPQExpBuffer(&proc_query,

                                                                " has_function_privilege(p.oid, 'EXECUTE')");

I suggest changing this to the following  In order to truly make the function return info about actual procedures:

                appendPQExpBuffer(&proc_query,

                                                                 " p.prokind = 'p' and"

                                                                " has_function_privilege(p.oid, 'EXECUTE')");

5. In configure, line 5016 (in the version 15.00 sources) has this:

                                CPPFLAGS="$CPPFLAGS -I$with_libpq/include -  I$with_libpq/include/postgresql/internal"

                I think that really ought to be this:

                                CPPFLAGS="$CPPFLAGS -I$with_libpq/include -I$with_libpq/include/internal"

6. In connection.c,  I think you should make this change.   The comment explains my reasoning.

Change these lines from:

                appendPQExpBufferStr(&query_buf, query);

                if (appendq)

                {

                                appendPQExpBuffer(&query_buf, ";%s", appendq);

                }

                if (query_rollback)

                {

                                appendPQExpBuffer(&query_buf, ";%s %s", rlscmd, per_query_svp);

                }

to:

                appendPQExpBufferStr(&query_buf, query);

/* prepend newline too, in case user SQL ends in a */

               /* dash-dash comment without a final newline character.  Otherwise, */

                /* the appended SQL statement gets treated as part of the comment.  */

                if (appendq)

                {

                                appendPQExpBuffer(&query_buf, "\n;%s", appendq);

                }

                if (query_rollback)

                {

                                appendPQExpBuffer(&query_buf, "\n;%s %s", rlscmd, per_query_svp);

                }

7. In odbcapi30.c, in  SQLEndTran(),   I think it should return SQL_ERROR if the connection is NOT in auto-commit mode.

Thus, in the SQL_HANDLE_DBC case inside that function, these two lines:

                  CC_clear_error((ConnectionClass *) Handle);

                  ret = PGAPI_Transact(SQL_NULL_HENV, Handle, CompletionType);

should be:

                if (((ConnectionClass *)Handle)->__error_number == CONNECTION_SERVER_REPORTED_SEVERITY_FATAL

                                                    &&     (CC_loves_visible_trans( (ConnectionClass *)Handle  )) )

                  ret = SQL_ERROR;

                                else

                  {

                    CC_clear_error((ConnectionClass *) Handle);

                    ret = PGAPI_Transact(SQL_NULL_HENV, Handle, CompletionType);

                  }

I would be interested in getting your opinion on my suggestions,  and hope the changes merit inclusion in the next release.

Sincerely,

Robert Willis