Re: [Cppcms-users] cppdb::statement copy and assign

[CN <cn...@fa...>]

Thank you for the reply!

Having received your message, I went back and study frontend.* and
backend.* of cppdb source code trying to understand more about it.
According to my findings so far, instances of cppdb::statement in
backend indeed are reference counted, meaning to me that these objects
will not be destroyed as long as they are still referenced.

But I can only go that deep into cppdb source. Because my original
troublesome code actually uses uncached statement but I do not fully
understand cppdb's statement cache mechanism, I begin to suspect this
part of cppdb.

I know this is not professional attitude, but the simplified version of
my troublesome code follows:

cppdb::statement create(cppdb::session &se)
{
  	cppdb::statement st=se.create_prepared_uncached_statement(
  		"UPDATE t SET c2=? WHERE c1=? AND c2=?")
  		<< 'xx' << 2;
  	return st;
}

void use()
{
  	cppdb::session se("connection params");
  	cppdb::statement st=create(se);
  	cppdb::transaction tx(se);
  	st.exec(); //Weird error occurs here.
  //Do more SQL here like this:
  //se << "INSERT INTO ...";
  	tx.commit();
}

Note that the data type of t.c2 is INTEGER.

And the PostgreSQL error log follows:

2015-10-31 14:20:42 CST ERROR:  invalid input syntax for integer:
"cppdb_psqlstmt_17"
2015-10-31 14:20:42 CST STATEMENT:  UPDATE t SET c2=$1 WHERE c1=$2 AND
c2=$3
2015-10-31 19:46:26 CST ERROR:  invalid input syntax for integer: ""
2015-10-31 19:46:26 CST STATEMENT:  UPDATE t SET c2=$1 WHERE c1=$2 AND
c2=$3
2015-10-31 22:42:24 CST ERROR:  invalid input syntax for integer: ""
2015-10-31 22:42:24 CST STATEMENT:  UPDATE t SET c2=$1 WHERE c1=$2 AND
c2=$3
2015-10-31 22:55:28 CST ERROR:  invalid input syntax for integer: "!"
2015-10-31 22:55:28 CST STATEMENT:  UPDATE t SET c2=$1 WHERE c1=$2 AND
c2=$3

As one might notice: How could I bind "cppdb_psqlstmt_17" to
cppdb::statement's parameters?
PostgreSQL complained weird arguments passed to c2 when the same
cppdb::statement::exec() was called repeatedly.

Any idea? Please!

Best Regards,
CN


On Thu, Nov 12, 2015, at 12:47 AM, Artyom Beilis wrote:
> Actually you are mistaken.
> 
> cppdb::statement, cppdb::session and cppdb::result keep the reference to
> underlying
> backed object representing them.
> 
> See: 
> http://cppcms.com/sql/cppdb/classcppdb_1_1statement.html#adc5ddc1e8df9248b74692990cffe40cf
> 
> So copy is totally valid and you can use it this way without problems.
> 
> Actually statement would also keep a track of the session and result of
> statement and session as
> well.
> 
> So it is correct way to do things.
> 
> Of course it is bad idea to start using copies of came cppdb::statement
> from two
> different locations simultaneously.
> 
> So look for the problem in some other place.
> 
> Artyom
> 
> 
> > Hi!
> > 
> > I recently made a mistake by doing this:
> > 
> > cppdb::statement create(){
> >   cppdb::session se("connect parameters");
> >   cppdb::statement  st=se << "SELECT * FROM t";
> >   return st;
> > }
> > 
> > void use(){
> >   cppdb::statement  st=create();
> >   st.query();
> > }
> > 
> > Very weird errors happens to st.query().
> > It took me quite some time to pin point the bug. After that, I re-read
> > the comments in cppdb/cppdb/frontend.h and I believe they confirm my
> > conclusion:
> > 
> > The problem here, I believe, is that cppdb::statement object is already
> > destroyed before st.query() is executed.
> > 
> > I am writing this note here in order to remind myself of avoiding such
> > practice. Hopefully no one else will make the same mistake as I did!
> > 
> > Meanwhile, I suspect it will be better off
> > 
> > (a) either to remove the cppdb::statement copy and assignment methods
> > from cppdb source altogether
> > (b) or to make cppdb::statement non-copyable
> > 
> > I have a feeling that these two methods serve only as a trap.
> > Yes? No?
> > 
> > Best Regards,
> > CN
> > 
> > -- 
> > http://www.fastmail.com - Access all of your messages and folders
> >                           wherever you are
> > 
> > 
> > ------------------------------------------------------------------------------
> > _______________________________________________
> > Cppcms-users mailing list
> > Cpp...@li...
> > https://lists.sourceforge.net/lists/listinfo/cppcms-users
> > 
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> Cppcms-users mailing list
> Cpp...@li...
> https://lists.sourceforge.net/lists/listinfo/cppcms-users

-- 
http://www.fastmail.com - Access your email from home and the web