From boris at codesynthesis.com  Wed Aug  1 09:36:23 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Aug  1 09:24:29 2012
Subject: [odb-users] How to make a "lazy" QList?
In-Reply-To: <CA+AwhR0vC553XwwJd3tQMmNvqJf3C8a-R6wSA5hnkx8ftXiCoQ@mail.gmail.com>
References: <CA+AwhR0vC553XwwJd3tQMmNvqJf3C8a-R6wSA5hnkx8ftXiCoQ@mail.gmail.com>
Message-ID: <boris.20120801153044@codesynthesis.com>

Hi Rene,

Rene Jensen <rene@catatonic.dk> writes:

> Do you have any plans for making a similar way to postpone loading of
> many-to-many keys and arrays?

Yes, this is on our TODO list. There also doesn't seem to be any
architectural/technical issues in supporting something like this.
Container loading is already a separate statement execution so
doing that at a later time shouldn't be a problem. Probably the
biggest effort will be to implement lazy variants for all the
standard, Boost, and Qt containers. Just need to find the time.

Boris

From rene at catatonic.dk  Wed Aug  1 11:57:51 2012
From: rene at catatonic.dk (Rene Jensen)
Date: Wed Aug  1 11:57:59 2012
Subject: [odb-users] How to make a "lazy" QList?
In-Reply-To: <boris.20120801153044@codesynthesis.com>
References: <CA+AwhR0vC553XwwJd3tQMmNvqJf3C8a-R6wSA5hnkx8ftXiCoQ@mail.gmail.com>
	<boris.20120801153044@codesynthesis.com>
Message-ID: <CA+AwhR1m2MRC9x6Tw5S_L-b6JXdeUWYEyCW5mW4VTGmKjhjigQ@mail.gmail.com>

>
> > Do you have any plans for making a similar way to postpone loading of
> > many-to-many keys and arrays?
>
> Yes, this is on our TODO list. There also doesn't seem to be any
> architectural/technical issues in supporting something like this.
> Container loading is already a separate statement execution so
> doing that at a later time shouldn't be a problem. Probably the
> biggest effort will be to implement lazy variants for all the
> standard, Boost, and Qt containers. Just need to find the time.
>
>
Hi Boris,

Excellent to hear. Now that I have pondered the ramifications of my own
suggestion, I think that making real lazy variants of lists, vectors etc.
might be overdoing it. What is needed is a (programmer-friendly) technique
to control loading of parts of an object. E.g.


  class Author
  { ...
    #pragma db lazyloaded
    QList <QLazySharedPointer<Book> >    authorsBooks;

    #pragma db lazyloaded
    vector<QLazySharedPointer<Article> >  authorsArticles;
    ...
  }

Author* A = ... some default load procedure ... will not load "lazyloaded"
members
A->authorsBooks.size() // 0 ... not loaded

db->load_lazymembers (A);

A->authorsBooks.size() // 34


Regards,
Rene
From boris at codesynthesis.com  Thu Aug  2 08:39:53 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug  2 08:27:54 2012
Subject: [odb-users] How to make a "lazy" QList?
In-Reply-To: <CA+AwhR1m2MRC9x6Tw5S_L-b6JXdeUWYEyCW5mW4VTGmKjhjigQ@mail.gmail.com>
References: <CA+AwhR0vC553XwwJd3tQMmNvqJf3C8a-R6wSA5hnkx8ftXiCoQ@mail.gmail.com>
	<boris.20120801153044@codesynthesis.com>
	<CA+AwhR1m2MRC9x6Tw5S_L-b6JXdeUWYEyCW5mW4VTGmKjhjigQ@mail.gmail.com>
Message-ID: <boris.20120802141257@codesynthesis.com>

Hi Rene,

Rene Jensen <rene@catatonic.dk> writes:

> Now that I have pondered the ramifications of my own suggestion, I think
> that making real lazy variants of lists, vectors etc. might be overdoing
> it. What is needed is a (programmer-friendly) technique to control loading
> of parts of an object. E.g.
> 
> 
>   class Author
>   { ...
>     #pragma db lazyloaded
>     QList <QLazySharedPointer<Book> >    authorsBooks;
> 
>     #pragma db lazyloaded
>     vector<QLazySharedPointer<Article> >  authorsArticles;
>     ...
>   }
> 
> Author* A = ... some default load procedure ... will not load "lazyloaded"
> members
> A->authorsBooks.size() // 0 ... not loaded
> 
> db->load_lazymembers (A);
> 
> A->authorsBooks.size() // 34

Yes, we've thought about this approach also. The major problem here is
that there is no good place to store the loaded/unloaded flag. While it
might not be necessary from the users' perspective (i.e., we can say
that it is the user's responsibility to know whether the lazy parts
are loaded or not), it gets tricky when we consider an operation like
update(). ODB has no way of knowing whether the container is loaded
(and thus we need to write the elements to the database) or not (and
thus we should do nothing for this container).

The other issue is that you may have several lazy containers in a class
and may want to load only one of them at a time. With the above approach
there is no way to specify which member to load. Doing something like
db->load_lazymembers (A, &Author::authorsBooks) will work but will 
require authorsBooks to be public. Plus:

A->authorsBooks.load ();

Feels more natural than:

db->load_lazymembers (A, &Author::authorsBooks);

The other alternative that we are considering is to provide a smart
pointer-like class template that implements laziness:

class Author
{ ...
  #pragma db lazyloaded
  odb::lazy<QList <QLazySharedPointer<Book> > >    authorsBooks;

};

A->authorsBooks.load ();
A->authorsBooks->size ();
(*A->authorsBooks)[0]->load (); // Not very elegant.

This is the approach that will probably be most suitable for simple
members if/when we decide to support lazy loading for them as well.

On a more conceptual level, what seems to be the ultimate goal is
to have the ability to segment data members into lazy-loadable
sections. Supporting lazy loading for simple members will require
a separate statement for each member, if we do it member-by-member.
This is a significant overhead. So grouping members into as few
sections as possible is a good idea. Maybe something along these
lines:

class Author
{ 

  odb::lazy_section publications;

  #pragma db lazy(publications)
  QList <QLazySharedPointer<Book> >    authorsBooks;

  #pragma db lazy(publications)
  vector<QLazySharedPointer<Article> >  authorsArticles;  

};

if (!A->publications.loaded ())
  A->publications.load (); // Loads authorsBooks and authorsArticles.

Just some thoughts.

Boris

From rene at catatonic.dk  Thu Aug  2 14:03:49 2012
From: rene at catatonic.dk (Rene Jensen)
Date: Thu Aug  2 14:03:57 2012
Subject: [odb-users] How to make a "lazy" QList?
In-Reply-To: <boris.20120802141257@codesynthesis.com>
References: <CA+AwhR0vC553XwwJd3tQMmNvqJf3C8a-R6wSA5hnkx8ftXiCoQ@mail.gmail.com>
	<boris.20120801153044@codesynthesis.com>
	<CA+AwhR1m2MRC9x6Tw5S_L-b6JXdeUWYEyCW5mW4VTGmKjhjigQ@mail.gmail.com>
	<boris.20120802141257@codesynthesis.com>
Message-ID: <CA+AwhR2XS-gGMxQntvLW8ojVab1tdnXgK7-y62NP-e1phdfEug@mail.gmail.com>

On a more conceptual level, what seems to be the ultimate goal is
> to have the ability to segment data members into lazy-loadable
> sections. Supporting lazy loading for simple members will require
> a separate statement for each member, if we do it member-by-member.
> This is a significant overhead. So grouping members into as few
> sections as possible is a good idea. Maybe something along these
> lines:
>
> class Author
> {
>
>   odb::lazy_section publications;
>
>   #pragma db lazy(publications)
>   QList <QLazySharedPointer<Book> >    authorsBooks;
>
>   #pragma db lazy(publications)
>   vector<QLazySharedPointer<Article> >  authorsArticles;
>
> };
>
> if (!A->publications.loaded ())
>   A->publications.load (); // Loads authorsBooks and authorsArticles.
>
> Just some thoughts.
>
> Boris
>


This is an utterly brilliant idea in every conceivable way that I can
immediately think of.
Cheers
From dctomm at gmail.com  Thu Aug  2 20:39:36 2012
From: dctomm at gmail.com (Douglas Tomm)
Date: Thu Aug  2 20:39:45 2012
Subject: [odb-users] how to represent mysql timestamps?
Message-ID: <501B1DC8.9000002@gmail.com>

what is the correct type to represent mysql timestamps?  the convention 
given in section 19.4.1 of the manual did not work for me.  when i put 
this in my class:

    #pragma db type("timestamp") not_null column("created")
    boost::posix_time::ptime m_created;

the following compiler messages appeared.  what is the correct way to do 
this? my environment is:

  * odb 2.0.0
  * gcc 4.1.2
  * boost 1.46.1
  * centos 5 linux

thank you,
doug

/usr/local/include/odb/mysql/traits.hxx: In static member function 
'static void odb::mysql::default_value_traits<T, <anonymous> 
 >::set_image(typename odb::mysql::image_traits<ID>::image_type&, bool&, 
T) [with T = boost::posix_time::ptime, odb::mysql::database_type_id ID = 
id_timestamp]':
build/cpp/odb/MyTest-odb.cxx:294:   instantiated from here
/usr/local/include/odb/mysql/traits.hxx:320: error: no matching function 
for call to 'st_mysql_time::st_mysql_time(boost::posix_time::ptime&)'
/usr/include/mysql/mysql_time.h:48: note: candidates are: 
st_mysql_time::st_mysql_time()
/usr/include/mysql/mysql_time.h:48: note: 
st_mysql_time::st_mysql_time(const st_mysql_time&)
/usr/local/include/odb/mysql/traits.hxx: In static member function 
'static void odb::mysql::default_value_traits<T, <anonymous> 
 >::set_value(T&, const typename 
odb::mysql::image_traits<ID>::image_type&, bool) [with T = 
boost::posix_time::ptime, odb::mysql::database_type_id ID = id_timestamp]':
build/cpp/odb/MyTest-odb.cxx:379:   instantiated from here
/usr/local/include/odb/mysql/traits.hxx:311: error: no matching function 
for call to 'boost::posix_time::ptime::ptime(const st_mysql_time&)'
/usr/include/boost/date_time/posix_time/ptime.hpp:53: note: candidates 
are: boost::posix_time::ptime::ptime()
/usr/include/boost/date_time/posix_time/ptime.hpp:49: note: 
boost::posix_time::ptime::ptime(boost::date_time::special_values)
/usr/include/boost/date_time/posix_time/ptime.hpp:45: 
note:                 boost::posix_time::ptime::ptime(const 
boost::date_time::counted_time_rep<boost::posix_time::millisec_posix_time_system_config>&)
/usr/include/boost/date_time/posix_time/ptime.hpp:42: note: 
boost::posix_time::ptime::ptime(boost::gregorian::date)
/usr/include/boost/date_time/posix_time/ptime.hpp:39: note: 
boost::posix_time::ptime::ptime(boost::gregorian::date, 
boost::posix_time::time_duration)
/usr/include/boost/date_time/posix_time/ptime.hpp:32: 
note:                 boost::posix_time::ptime::ptime(const 
boost::posix_time::ptime&)

From boris at codesynthesis.com  Fri Aug  3 06:59:10 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Aug  3 06:47:00 2012
Subject: [odb-users] how to represent mysql timestamps?
In-Reply-To: <501B1DC8.9000002@gmail.com>
References: <501B1DC8.9000002@gmail.com>
Message-ID: <boris.20120803125507@codesynthesis.com>

Hi Doug

Douglas Tomm <dctomm@gmail.com> writes:

> When i put this in my class:
>
>    #pragma db type("timestamp") not_null column("created")
>    boost::posix_time::ptime m_created;
>
> the following compiler messages appeared.

My guess is you didn't add the boost/date-time profile to your ODB
compiler command line (you will also need to link to the libodb-boost
library):

odb ... -p boost/date-time ...

Remember that the database mapping for Boost types is provides by
the Boost profile. See Chapter 19, "Boost Profile" in the ODB
Manual for details.

Boris

From doug at plaxo.com  Fri Aug  3 12:48:14 2012
From: doug at plaxo.com (Douglas Tomm)
Date: Fri Aug  3 12:48:23 2012
Subject: [odb-users] how to represent mysql timestamps?
In-Reply-To: <boris.20120803125507@codesynthesis.com>
References: <501B1DC8.9000002@gmail.com>
	<boris.20120803125507@codesynthesis.com>
Message-ID: <501C00CE.1050305@plaxo.com>

yep, that's all i needed to do.  thanks again!

d

On 8/3/12 3:59 AM, Boris Kolpackov wrote:
> Hi Doug
>
> Douglas Tomm <dctomm@gmail.com> writes:
>
>> When i put this in my class:
>>
>>     #pragma db type("timestamp") not_null column("created")
>>     boost::posix_time::ptime m_created;
>>
>> the following compiler messages appeared.
> My guess is you didn't add the boost/date-time profile to your ODB
> compiler command line (you will also need to link to the libodb-boost
> library):
>
> odb ... -p boost/date-time ...
>
> Remember that the database mapping for Boost types is provides by
> the Boost profile. See Chapter 19, "Boost Profile" in the ODB
> Manual for details.
>
> Boris
>

From eric.b.sum at lmco.com  Mon Aug  6 09:09:29 2012
From: eric.b.sum at lmco.com (Sum, Eric B)
Date: Mon Aug  6 09:10:00 2012
Subject: [odb-users] SQLITE unable to open db multiple threads
Message-ID: <D7ACDD7CE9068447B51856EB8A74F5E8077312@HVXDSP45.us.lmco.com>

Hi,

I am persisting objects into a sqlite database from multiple threads.  However, I am getting odd results caught with odb exceptions.  I get sqlite error 14: unable to open database multiple times throughout the execution of my program, and only a fraction of the objects persisted are present in the database.  I assume I am getting this error because one of the threads is unable to establish a connection to the database.  I am aware that you are unable to do concurrent writes to the sqlite database due to its exclusive locking mechanism, but I thought that there would be some scheduling/blocking scheme used inherently by odb to schedule the order of the persists from the different threads.    How do I fix this?  Any ideas on why I am getting this error?

Thanks,
Eric
From eric.b.sum at lmco.com  Mon Aug  6 09:25:26 2012
From: eric.b.sum at lmco.com (Sum, Eric B)
Date: Mon Aug  6 09:25:40 2012
Subject: [odb-users] RE: SQLITE unable to open db multiple threads
Message-ID: <D7ACDD7CE9068447B51856EB8A74F5E8077328@HVXDSP45.us.lmco.com>

Sorry, one more follow-up question.

The odb manual states that by default, a sqlite connection_pool_factory is used, which behind the scenes uses the shared_cache/unlock_notify mechanism.  From reading about the shared_cache/unlock_notify mechanism it seems that "The sqlite3_unlock_notify()<http://sqlite.org/c3ref/unlock_notify.html> interface is designed for use in systems that have a separate thread assigned to each database connection<http://sqlite.org/c3ref/sqlite3.html>"  and the shared_cache mechanism allows the same cache to be shared across the entire process that contains all of its threads.  How do I create a database connection for each individual separate thread as it seems the intended use is?

1.        I tried explicitly calling connection_ptr c(db.connection())  in each thread on the existing db object created in the main function that was passed to each thread.

2.       I tried passing the existing db object to each thread, which I suppose does not create a new connection and uses the existing one across multiple threads.

3.       I tried called the sqlite db constructor in each thread to open the existing database created in the main function in READWRITE mode, which I assume creates a connection in each thread to use.

None of these seemed to work.  Any help or insight on this is much appreciated.

Thanks again,
Eric

From: Sum, Eric B
Sent: Monday, August 06, 2012 9:09 AM
To: 'odb-users@codesynthesis.com'
Subject: SQLITE unable to open db multiple threads

Hi,

I am persisting objects into a sqlite database from multiple threads.  However, I am getting odd results caught with odb exceptions.  I get sqlite error 14: unable to open database multiple times throughout the execution of my program, and only a fraction of the objects persisted are present in the database.  I assume I am getting this error because one of the threads is unable to establish a connection to the database.  I am aware that you are unable to do concurrent writes to the sqlite database due to its exclusive locking mechanism, but I thought that there would be some scheduling/blocking scheme used inherently by odb to schedule the order of the persists from the different threads.    How do I fix this?  Any ideas on why I am getting this error?

Thanks,
Eric
From boris at codesynthesis.com  Mon Aug  6 11:00:55 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Aug  6 10:48:14 2012
Subject: [odb-users] Re: Binary Representation of PostgreSQL arrays
In-Reply-To: <5E840D29AD76D2429C1B0FD628D3F360D37C@lcldn-mail12.Liquid-Capital.liquidcap.com>
References: <5E840D29AD76D2429C1B0FD628D3F360D37C@lcldn-mail12.Liquid-Capital.liquidcap.com>
Message-ID: <boris.20120806162839@codesynthesis.com>

Hi Leo,

[CC'ing odb-users to my reply.]

Leonardo Greca <lgreca@liquidcapital.com> writes:

> I have been playing with the new experimental version of ODB and things
> are behaving pretty well.
> 
> Using the example you posted on the blog below, I was able to store stl
> containers using the pgsql types: DOUBLE PRECISION[][], TEXT[][] and
> INTEGER[][]
> http://www.codesynthesis.com/~boris/blog/2012/07/18/custom-database-to-cxx-type-mapping-in-odb/

Glad to hear it is working well.


> All that is done by converting the array to its text representation; e.g.,
> "{n1,n2,...}".
> 
> The problem is that parsing strings is very expensive for both CPU and
> network.

It seems that the PostgreSQL developers don't think this overhead is
significant since their canonical array literal format is {n1,n2,...}.
While there is also a binary (over-the-wire) format for arrays (and other
types), it is not explicitly documented (except in the source code) and
is quite complex. I found this page that has a pretty easy to follow
description for a one-dimensional array of 4-byte integers:

http://stackoverflow.com/questions/4016412/postgresqls-libpq-encoding-for-binary-transport-of-array-data

Based on that information, the binary representation of a 3-integer array
(e.g., {1, 2, 3}) takes 44 bytes. In comparison, the "{1,2,3}" string 
literal takes only 9 bytes. One major overhead of the binary array
format in PG is that each element is accompanied by a 4-byte size.

> I understand it is possible to save any container as bytea but that would
> make the db unreadable.
> 
> I was wondering if it is possible to use ODB to send a binary/compact
> representation of HSTORE or DOUBLE PRECISION[][] to postgres and still
> have it stored as HSTORE or DOUBLE PRECISION[][].

The short answer is, yes, it is possible but is not easy. Below is the
long answer:

First of all, simple cast to/from BYTEA doesn't work. I did some more
digging and found two PG functions that seem relevant: array_send and
array_recv. These functions are used to write and read the binary,
over-the-wire representation of arrays.

First I tried array_send which takes an array as an argument and
returns BYTEA. I added it to the 'from' conversion expression and
indeed I got the binary representation of an array. I tried it on
INTEGER[] and could see exactly the output described on that page
I mentioned above.

Unfortunately, conversion in the other direction is not that easy.
We cannot use array_recv directly because it expects an internal
buffer representation rather than BYTEA. However, this can be
worked around with a simple server-side C function. Below is its
source code. For more information on adding C function extensions,
refer to the PG documentation.

#include "postgres.h"
#include "lib/stringinfo.h"
#include "utils/array.h"

#ifdef PG_MODULE_MAGIC
PG_MODULE_MAGIC;
#endif

PG_FUNCTION_INFO_V1(sql_array_recv);

Datum
sql_array_recv (PG_FUNCTION_ARGS)
{
  bytea *d = PG_GETARG_BYTEA_PP(0);
  StringInfoData si;

  si.data = VARDATA (d);
  si.len = si.maxlen = VARSIZE(d) - VARHDRSZ;
  si.cursor = 0;

  fcinfo->arg[0] = (Datum)&si;

  PG_RETURN_ARRAYTYPE_P(array_recv (fcinfo));
}

Once this function is compiled into a shared library, say array_recv.so,
we can load it into the PG database like this (using GNU/Linux as an
example):

su postgres
psql -d dbname

CREATE FUNCTION int_array_recv(bytea, integer, integer) RETURNS integer[]
     AS '$libdir/array_recv.so', 'sql_array_recv'
     LANGUAGE C STRICT;

GRANT EXECUTE ON FUNCTION int_array_recv(bytea, integer, integer) TO PUBLIC;

Once this is done, we can use int_array_recv in the 'to' conversion
expression. The second argument to this function is the Oid of the
array element (e.g., 23 for INTEGER). And the third argument should
be 0. So for INTEGER[] we can have a mapping pragma like this:

#pragma db map type("INTEGER *\\[(\\d*)\\]")    \
               as("BYTEA")                      \
               to("int_array_recv((?), 23, 0)") \
               from("array_send ((?))")

After that the only thing left is to write the code (in the value_traits
specialization) that can read and write the PG binary array format.

Boris

From boris at codesynthesis.com  Mon Aug  6 13:15:40 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Aug  6 13:02:57 2012
Subject: [odb-users] SQLITE unable to open db multiple threads
In-Reply-To: <D7ACDD7CE9068447B51856EB8A74F5E8077312@HVXDSP45.us.lmco.com>
References: <D7ACDD7CE9068447B51856EB8A74F5E8077312@HVXDSP45.us.lmco.com>
Message-ID: <boris.20120806190341@codesynthesis.com>

Hi Eric,

Sum, Eric B <eric.b.sum@lmco.com> writes:

> I am persisting objects into a sqlite database from multiple threads. 
> However, I am getting odd results caught with odb exceptions. I get
> sqlite error 14: unable to open database multiple times throughout
> the execution of my program, and only a fraction of the objects
> persisted are present in the database.

Can you take a look at the common/threads test in the odb-tests
package and see if there is anything different between that code
and your application. You can either try to model your code after
the test, or you can try to change the test to model your
application and thus reproduce the problem. If you can reproduce
it with the test, then I would be happy to take a look.


> I am aware that you are unable to do concurrent writes to the sqlite
> database due to its exclusive locking mechanism, but I thought that
> there would be some scheduling/blocking scheme used inherently by odb
> to schedule the order of the persists from the different threads.

That's correct. And the common/threads test is the proof that it
works without doing anything special/extra. Besides other things,
this test persists multiple objects from multiple threads into the
same database, simultaneously.


> The odb manual states that by default, a sqlite connection_pool_factory
> is used, which behind the scenes uses the shared_cache/unlock_notify
> mechanism.  From reading about the shared_cache/unlock_notify mechanism
> it seems that "The sqlite3_unlock_notify() interface is designed for
> use in systems that have a separate thread assigned to each database
> connection"  and the shared_cache mechanism allows the same cache to
> be shared across the entire process that contains all of its threads.
> How do I create a database connection for each individual separate
> thread as it seems the intended use is?

This is all done automatically by ODB/connection_pool_factory.


> 1. I tried explicitly calling connection_ptr c(db.connection()) in each
>    thread on the existing db object created in the main function that was
>    passed to each thread.

No, you don't need to do that.


> 2. I tried passing the existing db object to each thread, which I suppose
>    does not create a new connection and uses the existing one across
>    multiple threads.

That's exactly how it should be done: you pass a single sqlite::database
instance to all your threads and this instance will "issue" an individual
connection (via the pool_factory) to each thread. Again, see the 'threads'
test for an example.


> 3. I tried called the sqlite db constructor in each thread to open the
>    existing database created in the main function in READWRITE mode,
>    which I assume creates a connection in each thread to use.

Nop, you shouldn't do that. Just create a single sqlite::database in
main() and pass it to all your threads.

Boris

From christian.fontana at gmail.com  Tue Aug  7 18:31:21 2012
From: christian.fontana at gmail.com (Christian Fontana)
Date: Tue Aug  7 18:31:29 2012
Subject: [odb-users] Ubuntu compiling issue
Message-ID: <CAHspYXPfQqOjaMx7JUXGvA=Gt7U9onKkXjQW9qDhLv82HcFMfQ@mail.gmail.com>

Hello,

I'm trying to compile odb 2.0.1 on my ubuntu machine.
The g++ version is:

Using built-in specs.
COLLECT_GCC=g++
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.6.1/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro
4.6.1-9ubuntu3' --with-bugurl=file:///usr/share/doc/gcc-4.6/README.Bugs
--enable-languages=c,c++,fortran,objc,obj-c++,go --prefix=/usr
--program-suffix=-4.6 --enable-shared --enable-linker-build-id
--with-system-zlib --libexecdir=/usr/lib --without-included-gettext
--enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.6
--libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu
--enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-plugin
--enable-objc-gc --disable-werror --with-arch-32=i686 --with-tune=generic
--enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu
--target=x86_64-linux-gnu
Thread model: posix
gcc version 4.6.1 (Ubuntu/Linaro 4.6.1-9ubuntu3)

The gcc versione is:
Using built-in specs.
COLLECT_GCC=gcc
COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.6.1/lto-wrapper
Target: x86_64-linux-gnu
Configured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro
4.6.1-9ubuntu3' --with-bugurl=file:///usr/share/doc/gcc-4.6/README.Bugs
--enable-languages=c,c++,fortran,objc,obj-c++,go --prefix=/usr
--program-suffix=-4.6 --enable-shared --enable-linker-build-id
--with-system-zlib --libexecdir=/usr/lib --without-included-gettext
--enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.6
--libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu
--enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-plugin
--enable-objc-gc --disable-werror --with-arch-32=i686 --with-tune=generic
--enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu
--target=x86_64-linux-gnu
Thread model: posix
gcc version 4.6.1 (Ubuntu/Linaro 4.6.1-9ubuntu3)


The error is:
checking whether g++ supports plugins... no
configure: error: g++ does not support plugins; reconfigure GCC with
--enable-plugin

But as you can see above the --enable-plugin was used by ubuntu team to
package the compiler.
What can I do?

thanks in advance
Christian F.
From boris at codesynthesis.com  Wed Aug  8 02:16:34 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Aug  8 02:03:31 2012
Subject: [odb-users] Ubuntu compiling issue
In-Reply-To: <CAHspYXPfQqOjaMx7JUXGvA=Gt7U9onKkXjQW9qDhLv82HcFMfQ@mail.gmail.com>
References: <CAHspYXPfQqOjaMx7JUXGvA=Gt7U9onKkXjQW9qDhLv82HcFMfQ@mail.gmail.com>
Message-ID: <boris.20120808081459@codesynthesis.com>

Hi Christian,

Christian Fontana <christian.fontana@gmail.com> writes:

> The error is:
> checking whether g++ supports plugins... no
> configure: error: g++ does not support plugins; reconfigure GCC with
> --enable-plugin
> 
> But as you can see above the --enable-plugin was used by ubuntu team to
> package the compiler.

I think you need to install the gcc-4.6-plugin-dev package. Quoting
the INSTALL file:

"
Note also that for pre-packaged GCC, plugin headers are usually
distributed in a separate package, normally called gcc-plugin-dev
or similar. You will need to install this package in order to build
ODB. For Debian/Ubuntu, this package is called gcc-X.Y-plugin-dev,
for example:

apt-get install gcc-4.7-plugin-dev

For RedHat/Fedora, this package is called gcc-plugin-devel, for
example:

yum install gcc-plugin-devel
"

Boris

From christian.fontana at gmail.com  Wed Aug  8 03:36:21 2012
From: christian.fontana at gmail.com (Christian Fontana)
Date: Wed Aug  8 03:36:28 2012
Subject: [odb-users] Ubuntu compiling issue
In-Reply-To: <boris.20120808081459@codesynthesis.com>
References: <CAHspYXPfQqOjaMx7JUXGvA=Gt7U9onKkXjQW9qDhLv82HcFMfQ@mail.gmail.com>
	<boris.20120808081459@codesynthesis.com>
Message-ID: <CAHspYXMOdWDUumSFkZKDvWxVws4mDa2R5jVQoJq5qMgnGYeFBg@mail.gmail.com>

Thanks,

Probably I read the INSTALL file to fast. It SOLVED the problem.
Regards

Christian

On Wed, Aug 8, 2012 at 8:16 AM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Christian,
>
> Christian Fontana <christian.fontana@gmail.com> writes:
>
> > The error is:
> > checking whether g++ supports plugins... no
> > configure: error: g++ does not support plugins; reconfigure GCC with
> > --enable-plugin
> >
> > But as you can see above the --enable-plugin was used by ubuntu team to
> > package the compiler.
>
> I think you need to install the gcc-4.6-plugin-dev package. Quoting
> the INSTALL file:
>
> "
> Note also that for pre-packaged GCC, plugin headers are usually
> distributed in a separate package, normally called gcc-plugin-dev
> or similar. You will need to install this package in order to build
> ODB. For Debian/Ubuntu, this package is called gcc-X.Y-plugin-dev,
> for example:
>
> apt-get install gcc-4.7-plugin-dev
>
> For RedHat/Fedora, this package is called gcc-plugin-devel, for
> example:
>
> yum install gcc-plugin-devel
> "
>
> Boris
>
From yb947835908 at gmail.com  Fri Aug 10 22:14:18 2012
From: yb947835908 at gmail.com (=?GB2312?B?1KzUrLHz?=)
Date: Fri Aug 10 22:14:26 2012
Subject: [odb-users] how to install ODB on windows
Message-ID: <CAAdSf_nH-at5Uc=tVce0_kucejM-hHqXDx_XXniWXvRZvXE7hA@mail.gmail.com>

I'm a new learner. What exactly do I have to do to install the ODB on
windows? It would be very helpful to tell me more details about it. Many
thanks!!
From boris at codesynthesis.com  Sun Aug 12 05:54:35 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Sun Aug 12 05:40:49 2012
Subject: [odb-users] how to install ODB on windows
In-Reply-To: <CAAdSf_nH-at5Uc=tVce0_kucejM-hHqXDx_XXniWXvRZvXE7hA@mail.gmail.com>
References: <CAAdSf_nH-at5Uc=tVce0_kucejM-hHqXDx_XXniWXvRZvXE7hA@mail.gmail.com>
Message-ID: <boris.20120812115214@codesynthesis.com>

Hi,

yb947835908@gmail.com writes:

> I'm a new learner. What exactly do I have to do to install the ODB on
> windows? It would be very helpful to tell me more details about it.

There is the "Installing ODB on Windows" page which gives a high-
level overview of the steps:

http://www.codesynthesis.com/products/odb/doc/install-windows.xhtml

Also make sure that for every package that you choose to install you
read the INSTALL file that comes inside the package.

Boris

From oded at geek.co.il  Sun Aug 12 05:52:48 2012
From: oded at geek.co.il (Oded Arbel)
Date: Sun Aug 12 05:52:55 2012
Subject: [odb-users] ODB roadmap to 2.1?
Message-ID: <CAOqhUvw6_UbGOpxtEOexqmHvxxeT3xbUJ4WDCJD7OdDuj_pdLA@mail.gmail.com>

To ODB maintainers.

I'm considering using ODB to replace a custom database abstraction layer in
a high-risk project with a rather longish release cycle.

I'm wondering if I should use ODB 2.0 or should I wait for 2.1 (I'm
interested in custom type mapping and better Oracle support, but its not a
must for me), which mostly depends if I I can develop with the 2.1 alphas
now and have the 2.1 release in time for my release date - so my question
is, can you guys provide some info about the timeline for the 2.1 release?
I've looked on the web site and didn't see a roadmap document for ODB (I
may have missed it?)

Thanks in advance,
-- 
Oded
From duhai_lee at qq.com  Sun Aug 12 21:22:40 2012
From: duhai_lee at qq.com (=?ISO-8859-1?B?SnVzdGxpdmU=?=)
Date: Mon Aug 13 02:00:39 2012
Subject: [odb-users] directly generate .h and .cxx by connected db?
Message-ID: <tencent_00C40A020CE431FA4A3D0A27@qq.com>

could odb command only generate cxx files from .h file?   Can I connect to the database directly generated . H files and .cxx files?
From boris at codesynthesis.com  Mon Aug 13 02:22:38 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Aug 13 02:08:44 2012
Subject: [odb-users] ODB roadmap to 2.1?
In-Reply-To: <CAOqhUvw6_UbGOpxtEOexqmHvxxeT3xbUJ4WDCJD7OdDuj_pdLA@mail.gmail.com>
References: <CAOqhUvw6_UbGOpxtEOexqmHvxxeT3xbUJ4WDCJD7OdDuj_pdLA@mail.gmail.com>
Message-ID: <boris.20120813082121@codesynthesis.com>

Hi Oded,

Oded Arbel <oded@geek.co.il> writes:

> so my question is, can you guys provide some info about the timeline
> for the 2.1 release?

The plan is to release 2.1.0 around mid-September.

Boris

From boris at codesynthesis.com  Mon Aug 13 02:24:21 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Aug 13 02:10:28 2012
Subject: [odb-users] directly generate .h and .cxx by connected db?
In-Reply-To: <tencent_00C40A020CE431FA4A3D0A27@qq.com>
References: <tencent_00C40A020CE431FA4A3D0A27@qq.com>
Message-ID: <boris.20120813082311@codesynthesis.com>

Hi,

Justlive <duhai_lee@qq.com> writes:

> Can I connect to the database directly generated . H files and .cxx
> files?

No, not yet. Though this feature is on our TODO list so at some point
we plan to provide this functionality.

Boris




From yb947835908 at gmail.com  Mon Aug 13 13:54:10 2012
From: yb947835908 at gmail.com (=?GB2312?B?1KzUrLHz?=)
Date: Mon Aug 13 13:54:17 2012
Subject: [odb-users] compiling error
Message-ID: <CAAdSf_=EexTC+WUK+Uo924g0yACsKf7Aer4N9AQ6bPw_YJ0KVQ@mail.gmail.com>

Thank u so much,it helped a lot!!
Since I've install ODB successfully, I start to learn to use it by
compiling and running the examples. And I encounter with this problem:
While compiling the mapping example,it says:  (traits.hxx':error:unable to
open in read mode ).
What dose that mean and how can I solve this?
On the other hand, I tried to remove the (--hxx-prologue '#include
"traits.hxx"' ) option while compiling and manually added text (#include
"traits.hxx") into the generated file person-odb.hxx , it worked  just well.

PS: My compiling and running environment:
Windows 7 , Microsoft Visual Studio 2008 , sqlite3.

By the way, there is a tiny error: in the
file \odb-examples-2.0.0\c++11\README, you miss a '-' in the command
line,that is,it should be "odb ... --std c++11..." , not  "odb ... -std
c++11...".
From Davide.Anastasia at qualitycapital.com  Tue Aug 14 04:49:18 2012
From: Davide.Anastasia at qualitycapital.com (Davide Anastasia)
Date: Tue Aug 14 04:50:11 2012
Subject: [odb-users] Unit Testing and ODB
Message-ID: <B4934D86F0A1B8419AB680C1B95886D10E13CE51@EXVS01.nasstar-t1.net>

Hi All,

I am wondering whether there is any way to programmatically control the
content of a database in order to write "well define" unit testing.

In particular, I wrote a function that retrieves an object if available,
or adds it to the database and I want to be sure that the database
doesn't contain that object before running my tests (using some kind of
fixture).

 

Best,

Davide

 

Davide Anastasia
Analyst, Research & Development


Quality Capital Management Ltd.
QCM House * Horizon Business Village
No. 1 Brooklands Road
Weybridge * Surrey KT13 0TJ
United Kingdom

Tel: +44 (0) 1932 334 400
Fax: +44 (0) 1932 334 415
Email: Davide.Anastasia@QualityCapital.com
<mailto:Davide.Anastasia@QualityCapital.com> 

www.qualitycapital.com <http://www.qualitycapital.com/> 

________________________________

This email and any attachments are confidential and intended solely for
the use of the individual(s) to whom it is addressed. 

Any views or opinions presented are solely those of the author and do
not necessarily represent those of Quality Capital Management Ltd. If
you are not the intended recipient, be advised that you have received
this email in error and that any use, dissemination, printing,
forwarding or copying of this email is strictly prohibited. Please
contact the sender if you have received this email in error. You should
also be aware that emails are susceptible to interference and you should
not assume that the contents of this email originated from the sender
above or that they have been accurately reproduced in their original
form. Quality Capital Management Ltd is authorised and regulated by the
Financial Services Authority in the UK and is a member of the National
Futures Association in the US. 

________________________________

 

From interceptor at 126.com  Mon Aug 13 21:58:10 2012
From: interceptor at 126.com (=?GBK?B?0e63sg==?=)
Date: Tue Aug 14 08:38:04 2012
Subject: [odb-users] help!how to generate database
Message-ID: <43e64fc7.1aa0a.13922da3b26.Coremail.interceptor@126.com>

Hi,Every one,I'm trying to use odb right now,and  I have red manual,and did some test.but I just find that odb don't support index column(not  ordered container?s index column of a data member,just index to speed search),it means i should manual add it?
Thans for your reply!
panda
From boris at codesynthesis.com  Tue Aug 14 09:00:58 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Aug 14 08:46:51 2012
Subject: [odb-users] compiling error
In-Reply-To: <CAAdSf_=EexTC+WUK+Uo924g0yACsKf7Aer4N9AQ6bPw_YJ0KVQ@mail.gmail.com>
References: <CAAdSf_=EexTC+WUK+Uo924g0yACsKf7Aer4N9AQ6bPw_YJ0KVQ@mail.gmail.com>
Message-ID: <boris.20120814145729@codesynthesis.com>

Hi,

yb947835908@gmail.com writes:

> And I encounter with this problem: While compiling the mapping example,
> it says: (traits.hxx':error:unable to open in read mode).
> What dose that mean and how can I solve this?
> On the other hand, I tried to remove the (--hxx-prologue '#include
> "traits.hxx"' ) option while compiling and manually added text (#include
> "traits.hxx") into the generated file person-odb.hxx , it worked  just well.

The command line in the README was not valid for Windows (I've fixed
that). Try this variant instead, which works for both POSIX and Windows
shells:

--hxx-prologue "#include \"traits.hxx\""

> By the way, there is a tiny error: in the file 
> \odb-examples-2.0.0\c++11\README, you miss a '-' in the command
> line, that is, it should be "odb ... --std c++11..." , not  "odb ... -std
> c++11...".

Fixed. Thanks for reposting this!

Boris

From boris at codesynthesis.com  Tue Aug 14 09:03:55 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Aug 14 08:49:51 2012
Subject: [odb-users] help!how to generate database
In-Reply-To: <43e64fc7.1aa0a.13922da3b26.Coremail.interceptor@126.com>
References: <43e64fc7.1aa0a.13922da3b26.Coremail.interceptor@126.com>
Message-ID: <boris.20120814150131@codesynthesis.com>

Hi,

interceptor@126.com writes:

> But I just find that odb don't support index column.

This feature has been added for the upcoming 2.1.0 release. Also
there is a pre-release available, if you don't want to wait:

http://www.codesynthesis.com/pipermail/odb-announcements/2012/000017.html

Boris

From boris at codesynthesis.com  Tue Aug 14 09:15:40 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Aug 14 09:01:36 2012
Subject: [odb-users] Unit Testing and ODB
In-Reply-To: <B4934D86F0A1B8419AB680C1B95886D10E13CE51@EXVS01.nasstar-t1.net>
References: <B4934D86F0A1B8419AB680C1B95886D10E13CE51@EXVS01.nasstar-t1.net>
Message-ID: <boris.20120814150643@codesynthesis.com>

Hi Davide,

Davide Anastasia <Davide.Anastasia@qualitycapital.com> writes:

> I am wondering whether there is any way to programmatically control the
> content of a database in order to write "well define" unit testing.
> 
> In particular, I wrote a function that retrieves an object if available,
> or adds it to the database and I want to be sure that the database
> doesn't contain that object before running my tests (using some kind of
> fixture).

I am not clear what kind of functionality you are looking for here. You
can make sure that the database is empty by re-creating the schema (i.e.,
dropping all the tables and re-creating them). After that, if your tests
expect some content in the database, then you will need to add it using
the standard ODB mechanisms. This is how ODB's own test suite does it.

Now, if you want to do this many times during your test execution,
then embedded (into C++) schema creation code will probably be more
convenient than a standalone SQL file. You can also partition your
schema into segments if you don't want to re-create the while
database for every test (which could be slow). For more information
on this functionality refer to Section 3.3, "Database" in the ODB
manual.

Boris

From eric.b.sum at lmco.com  Wed Aug 15 11:09:34 2012
From: eric.b.sum at lmco.com (Sum, Eric B)
Date: Wed Aug 15 11:12:35 2012
Subject: [odb-users] SQL Statement Tracer
Message-ID: <D7ACDD7CE9068447B51856EB8A74F5E807AA4A@HVXDSP45.us.lmco.com>

Hi,

I have been trying out the SQL Statement execution tracer.  It's a very neat and useful tool for debugging.  I have a question about it (I am using a sqlite database).

When I use the tracer to print out the statements (either with the built-in stderr_tracer or my own odb::sqlite::tracer implementation), I have been unable to print out the values associated with the SQL statement.  For example, if I persist objects into a table.  It prints something along the lines of

INSERT INTO table VALUES (?,?,?)

It looks like the parameters for the sqlite_statement are being print out unbound.  I was wondering if it is possible to get it to print something like:

INSERT INTO table VALUES(1,2,3) where the bound values appear rather than "?"

I have looked through the odb::sqlite::statement class to see if there are any particular functions specific to sqlite that I can use for this, but I am unable to find any.  I might have missed it.

Thanks so much,
Eric
From yb947835908 at gmail.com  Wed Aug 15 11:59:11 2012
From: yb947835908 at gmail.com (=?GB2312?B?1KzUrLHz?=)
Date: Wed Aug 15 11:59:19 2012
Subject: [odb-users] error:Unknown pragma;
Message-ID: <CAAdSf_mvotLa87qQQgZGRkKjgowG=ovYxL_BnAvqz7WSyhWTQw@mail.gmail.com>

Hi!
I've run all the examples successfully.After that,I start to build my own
project using odb.
Those files(.hxx,.c.xx,.ixx) are generated successfully,but after that,when
compiling the project on VS2008,
it says"error :Unknown pragma" in those lines that contain "#pragma db
....",
also it can't recognise the "db" from text " auto_ptr<database> db (new
odb::mysql::database (argc, argv));".
By the way, when adding the .ixx into the project, I ignored to create new
.rules file, since I don't know what that means.
So,my problem is, after the files are generated, how to build a new project
, how to compile , and at last create the .exe file.
Thanks!
Environment : vs2008, windows 7, sqlites.
From boris at codesynthesis.com  Thu Aug 16 08:33:40 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug 16 08:19:15 2012
Subject: [odb-users] SQL Statement Tracer
In-Reply-To: <D7ACDD7CE9068447B51856EB8A74F5E807AA4A@HVXDSP45.us.lmco.com>
References: <D7ACDD7CE9068447B51856EB8A74F5E807AA4A@HVXDSP45.us.lmco.com>
Message-ID: <boris.20120816142611@codesynthesis.com>

Hi Eric,

Sum, Eric B <eric.b.sum@lmco.com> writes:

> It prints something along the lines of
> 
> INSERT INTO table VALUES (?,?,?)
> 
> It looks like the parameters for the sqlite_statement are being print
> out unbound. I was wondering if it is possible to get it to print 
> something like:
> 
> INSERT INTO table VALUES(1,2,3) where the bound values appear rather
> than "?"

No, this is not possible. ODB uses prepared statements and passes all
the parameters in binary form (i.e., ODB actually never constructs a
string like "INSERT INTO table VALUES(1,2,3)"). Adding this functionality
would involve quite a bit of effort/code bloat. It is also a fairly
"hairy" problem if you start to think about it. For example, how do
we print BLOBs. What if the data is very long (say multiple megabytes)?
What about national characters?

If you want to see actual values, then I suggest that you look for
this support in your database logs. In case of SQLite, there is
the sqlite3_trace() function. However, I don't think it will print
values for parameters in prepared statements either.

Boris

From boris at codesynthesis.com  Thu Aug 16 08:48:50 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug 16 08:34:22 2012
Subject: [odb-users] error:Unknown pragma;
In-Reply-To: <CAAdSf_mvotLa87qQQgZGRkKjgowG=ovYxL_BnAvqz7WSyhWTQw@mail.gmail.com>
References: <CAAdSf_mvotLa87qQQgZGRkKjgowG=ovYxL_BnAvqz7WSyhWTQw@mail.gmail.com>
Message-ID: <boris.20120816143748@codesynthesis.com>

Hi,

yb947835908@gmail.com writes:

> I start to build my own project using odb.
>
> Those files (.hxx,.c.xx,.ixx) are generated successfully, but after that,
> when compiling the project on VS2008, it says
>
> "error :Unknown pragma" in those lines that contain "#pragma db ...."

Normally this is a warning. I think you instructed your VC++ compiler to
treat all warnings as errors. For more information on how to suppress
these warnings/errors, see Section 12.6, "C++ Compiler Warnings" in the
ODB manual.


> also it can't recognise the "db" from text " auto_ptr<database> db (new
> odb::mysql::database (argc, argv));".

I assume you are using the database.hxx file from one of the examples.
If so, then you need to define one of the DATABASE_* macros to select
the database that you would like to use (e.g., DATABASE_SQLITE).
Alternatively, you can remove database.hxx and create the database
directly. For example:

#include <odb/sqlite/database.hxx>

int main (int argc, char* argv[])
{
  odb::sqlite::database db (argc, argv);

  ...
}

Boris

From yb947835908 at gmail.com  Fri Aug 17 22:43:01 2012
From: yb947835908 at gmail.com (=?GB2312?B?1KzUrLHz?=)
Date: Fri Aug 17 22:43:08 2012
Subject: [odb-users] Can I define views wherever I want ?
Message-ID: <CAAdSf_nYdcuabGPrH1XMtrxOf1DhibJt+_J9mcbW0YBZVvhD_w@mail.gmail.com>

Hi!
Must I define all the views in the .hxx file(the input file of the ODB
compiler) ?
What if I don't know what views I may use at first?
Can I define views wherever I want to ,for example, in my main function?
Thanks!
From yo at miguelrevilla.com  Sat Aug 18 06:23:04 2012
From: yo at miguelrevilla.com (=?UTF-8?Q?Miguel_Revilla_Rodr=C3=ADguez?=)
Date: Sat Aug 18 06:23:13 2012
Subject: [odb-users] std::set<odb::lazy_shared_ptr<foo>> fails
Message-ID: <CAG9fDsWubZJ7Xfq3g44aPH_ee9Ei3JkFaBygJX2F+ByGptVAQg@mail.gmail.com>

Hi,

I'm trying to declare an object with a
std::set<odb::lazy_shared_ptr<foo>> member, and while the odb compiler
works nicely on it and doesn't complain, later the generated code
(object-odb.cpp) fails to compile with G++ 4.7.1 with options -O2
-std=c++11. If I change std::set with std::vector, then everything is
fine, but I'd prefer to use a set rather than a vector in this case.
Any thoughts?

Thanks,

Miguel

From yo at miguelrevilla.com  Sat Aug 18 20:06:16 2012
From: yo at miguelrevilla.com (=?UTF-8?Q?Miguel_Revilla_Rodr=C3=ADguez?=)
Date: Sat Aug 18 20:06:24 2012
Subject: [odb-users] Trouble with lazy_ptr
Message-ID: <CAG9fDsUBDnUkM5_yK2xy7YeoDOKih95jMykdzcuykPAkrxkv_A@mail.gmail.com>

I'm completely lost on this. Examples compile fine, so it's not an
environment problem. Basically I use this files (sorry if it is a bit
long, but I'm really, really lost):

node.h:

#ifndef NODE_H
#define NODE_H

#include <string>
#include <vector>
#include <odb/core.hxx>
#include <odb/lazy-ptr.hxx>

#include "activity.h"
#include "userProfile.h"
#include "avatar.h"

class group ;

#pragma db object polymorphic
class node {
public:
        node(   const std::string &hash,
                unsigned long creationTime) :
                hash_(hash), creationTime_(creationTime) {}
        virtual ~node() = 0 ;

        void nodeavatar(const odb::lazy_shared_ptr<avatar> &n) ;
        const odb::lazy_shared_ptr<avatar> nodeavatar() ;
        const std::string nodeHash() ;


protected:
        friend class odb::access ;
        node() {}

        #pragma db id auto
        unsigned long id_ ;
        std::string hash_ ;
        unsigned long creationTime_ ;
        #pragma db value_not_null
        odb::lazy_shared_ptr<avatar> nodeavatar_ ;
        #pragma db value_not_null
        std::vector<odb::lazy_shared_ptr<activity> > activities_ ;
        #pragma db value_not_null inverse(nodes_)
        std::vector<odb::lazy_weak_ptr<group> > groups_ ;
} ;

#pragma db object
class user : public node {
typedef std::vector<odb::lazy_weak_ptr<user>> friends_type ;
public:
        user(   const std::string &hash,
                unsigned long creationTime,
                const std::string &username,
                const std::string &password,
                const std::string &email,
                const std::string &creationIP) :
                node(hash,creationTime),
                username_(username),
                password_(password),
                email_(email),
                creationIP_(creationIP) {}

        const unsigned long id() ;
        const bool verified() ;
        const bool locked() ;
        const std::string username() ;
        const std::string email() ;
        void verified(const bool &v) ;
        void locked(const bool &l) ;
        void email(const std::string &e) ;
        const std::shared_ptr<userProfile> profile() ;
        void profile(const std::shared_ptr<userProfile> &p) ;
        friends_type& friends() ;
        const friends_type& friends() const ;
        void addFriend(const std::shared_ptr<user> &u) ;

private:
        friend class odb::access ;
        user() {}

        std::string username_ ;
        std::string password_ ;
        std::string email_ ;
        #pragma db default(false)
        bool verified_ ;
        #pragma db default(false)
        bool locked_ ;
        std::string creationIP_ ;
        #pragma db not_null
        std::shared_ptr<userProfile> userProfile_ ;
        #pragma db value_not_null
        friends_type friends_ ;
} ;

#endif // NODE_H

----------------------------------------------------------------
node.cpp:

#include "node.h"

node::~node() {}

void node::nodeavatar(const odb::lazy_shared_ptr<avatar> &n) {
        nodeavatar_ = n ;
}

const odb::lazy_shared_ptr<avatar> node::nodeavatar() {
        return nodeavatar_ ;
}

const std::string node::nodeHash() {
        return hash_ ;
}

const unsigned long user::id() {
        return id_ ;
}

const bool user::verified() {
        return verified_ ;
}

const bool user::locked() {
        return locked_ ;
}

const std::string user::username() {
        return username_ ;
}

const std::string user::email() {
        return email_ ;
}

void user::email(const std::string &e) {
        email_ = e ;
}

void user::verified(const bool &v) {
        verified_ = v ;
}

void user::locked(const bool &l) {
        locked_ = l ;
}

const std::shared_ptr<userProfile> user::profile() {
        return userProfile_ ;
}

void user::profile(const std::shared_ptr<userProfile> &p) {
        userProfile_ = p ;
}

typedef std::vector<odb::lazy_weak_ptr<user>> friends_type ;

friends_type& user::friends() {
        return friends_ ;
}

const friends_type& user::friends() const {
        return friends_ ;
}

void user::addFriend(const std::shared_ptr<user> &u) {
//      friends_.insert(u) ;
}

------------------------------------
apiUserFriends.cpp:


#include <odb/database.hxx>
#include <odb/pgsql/database.hxx>
#include <odb/transaction.hxx>
#include <Wt/Utils>
#include <vector>
#include <sstream>
#include <string>
#include "apiUserFriends.h"
#include "odbClasses/node.h"
#include "odbClasses/node-odb.h"
#include "odbClasses/session.h"
#include "odbClasses/session-odb.h"

apiUserFriends::apiUserFriends(serviceTools *serviceT, WObject
*parent) : WResource(parent), serviceT_(serviceT) {

}

apiUserFriends::~apiUserFriends() {
        beingDeleted() ;
}

void apiUserFriends::handleRequest(const Http::Request& request,
Http::Response& response) {

        response.setMimeType("application/json") ;

        if(
serviceT_->checkClientAuth(Utils::urlDecode(*request.getParameter("token")),permissions::permFrontend)
&&

serviceT_->checkSessionAuth(Utils::urlDecode(*request.getParameter("sessionid")),Utils::hexEncode(Utils::md5(request.clientAddress()))))
{

                typedef odb::query<session> query ;
                typedef odb::result<session> result ;

                odb::database *db_ = serviceT_->database() ;

                {
                        typedef std::vector<odb::lazy_weak_ptr<user>>
friends_type ;
                        odb::transaction t (db_->begin()) ;
                        result g(db_->query<session> ( query::token ==
Utils::urlDecode(*request.getParameter("sessionid")))) ;

                        if(g.size()==1) {
                                for(session& s: g) {
                                        std::shared_ptr<user>
u(s.sessionUser()) ;
                                        std::stringstream friendsStr ;
                                        friends_type& f (u->friends()) ;
                                        friends_type::iterator it ;
                                        for(auto it (f.begin()); it !=
f.end() ; it++) {

odb::lazy_weak_ptr<user>& tmpPtr (*it) ;
                                                tmpPtr.load() ;
                                        }
                                        response.out() <<
"{\"result\":\"true\",\"friends\":[" ;
                                        response.out() <<
friendsStr.str().substr(0,friendsStr.str().size()-1) ;
                                        response.out() << "]}\n" ;
                                }
                        } else {
                                response.out() <<
"{\"result\":\"error\",\"data\":{\"errormsg\":\"bad_session\"}}\n" ;
                        }

                        t.commit() ;
                }
        } else {
                response.out()  << "{\"result\":\"false\"}\n" ;
        }
}


When I try to compile I get:

[  2%] Building CXX object CMakeFiles/yajpws.dir/apiUserFriends.cpp.o
In file included from /usr/include/odb/lazy-ptr.hxx:643:0,
                 from /home/mine/bettercodes/yajp-backend/odbClasses/node.h:7,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:9:
/usr/include/odb/lazy-ptr.ixx: In instantiation of
'std::shared_ptr<_Tp> odb::lazy_weak_ptr<T>::load() const [with T =
user]':
/home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:48:19:
required from here
/usr/include/odb/lazy-ptr.ixx:1511:5: error: no match for 'operator='
in 'r = ((const
odb::lazy_weak_ptr<user>*)this)->odb::lazy_weak_ptr<user>::i_.odb::lazy_ptr_impl<T>::load<user>(0)'
/usr/include/odb/lazy-ptr.ixx:1511:5: note: candidates are:
In file included from
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/memory:87:0,
                 from /usr/include/odb/pointer-traits.hxx:11,
                 from /usr/include/odb/traits.hxx:11,
                 from /usr/include/odb/database.hxx:13,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:1:
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:269:19:
note: std::shared_ptr<_Tp>& std::shared_ptr<_Tp>::operator=(const
std::shared_ptr<_Tp>&) [with _Tp = user; std::shared_ptr<_Tp> =
std::shared_ptr<user>]
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:269:19:
note:   no known conversion for argument 1 from
'odb::object_traits<user>::pointer_type {aka user*}' to 'const
std::shared_ptr<user>&'
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:273:2:
note: template<class _Tp1> std::shared_ptr&
std::shared_ptr::operator=(const std::shared_ptr<_Tp1>&) [with _Tp1 =
_Tp1; _Tp = user]
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:273:2:
note:   template argument deduction/substitution failed:
In file included from /usr/include/odb/lazy-ptr.hxx:643:0,
                 from /home/mine/bettercodes/yajp-backend/odbClasses/node.h:7,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:9:
/usr/include/odb/lazy-ptr.ixx:1511:5: note:   mismatched types 'const
std::shared_ptr<_Tp>' and 'odb::object_traits<user>::pointer_type {aka
user*}'
In file included from
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/memory:87:0,
                 from /usr/include/odb/pointer-traits.hxx:11,
                 from /usr/include/odb/traits.hxx:11,
                 from /usr/include/odb/database.hxx:13,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:1:
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:282:2:
note: template<class _Tp1> std::shared_ptr&
std::shared_ptr::operator=(std::auto_ptr<_Tp1>&&) [with _Tp1 = _Tp1;
_Tp = user]
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:282:2:
note:   template argument deduction/substitution failed:
In file included from /usr/include/odb/lazy-ptr.hxx:643:0,
                 from /home/mine/bettercodes/yajp-backend/odbClasses/node.h:7,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:9:
/usr/include/odb/lazy-ptr.ixx:1511:5: note:   mismatched types
'std::auto_ptr<T>' and 'odb::object_traits<user>::pointer_type {aka
user*}'
In file included from
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/memory:87:0,
                 from /usr/include/odb/pointer-traits.hxx:11,
                 from /usr/include/odb/traits.hxx:11,
                 from /usr/include/odb/database.hxx:13,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:1:
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:290:7:
note: std::shared_ptr<_Tp>&
std::shared_ptr<_Tp>::operator=(std::shared_ptr<_Tp>&&) [with _Tp =
user; std::shared_ptr<_Tp> = std::shared_ptr<user>]
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:290:7:
note:   no known conversion for argument 1 from
'odb::object_traits<user>::pointer_type {aka user*}' to
'std::shared_ptr<user>&&'
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:298:2:
note: template<class _Tp1> std::shared_ptr&
std::shared_ptr::operator=(std::shared_ptr<_Tp1>&&) [with _Tp1 = _Tp1;
_Tp = user]
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:298:2:
note:   template argument deduction/substitution failed:
In file included from /usr/include/odb/lazy-ptr.hxx:643:0,
                 from /home/mine/bettercodes/yajp-backend/odbClasses/node.h:7,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:9:
/usr/include/odb/lazy-ptr.ixx:1511:5: note:   mismatched types
'std::shared_ptr<_Tp>' and 'odb::object_traits<user>::pointer_type
{aka user*}'
In file included from
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/memory:87:0,
                 from /usr/include/odb/pointer-traits.hxx:11,
                 from /usr/include/odb/traits.hxx:11,
                 from /usr/include/odb/database.hxx:13,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:1:
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:306:2:
note: template<class _Tp1, class _Del> std::shared_ptr&
std::shared_ptr::operator=(std::unique_ptr<_Up, _Ep>&&) [with _Tp1 =
_Tp1; _Del = _Del; _Tp = user]
/usr/lib/gcc/x86_64-unknown-linux-gnu/4.7.1/../../../../include/c++/4.7.1/bits/shared_ptr.h:306:2:
note:   template argument deduction/substitution failed:
In file included from /usr/include/odb/lazy-ptr.hxx:643:0,
                 from /home/mine/bettercodes/yajp-backend/odbClasses/node.h:7,
                 from /home/mine/bettercodes/yajp-backend/apiUserFriends.cpp:9:
/usr/include/odb/lazy-ptr.ixx:1511:5: note:   mismatched types
'std::unique_ptr<_Tp, _Dp>' and
'odb::object_traits<user>::pointer_type {aka user*}'
make[2]: *** [CMakeFiles/yajpws.dir/apiUserFriends.cpp.o] Error 1
make[1]: *** [CMakeFiles/yajpws.dir/all] Error 2
make: *** [all] Error 2


odb files are compiled with:

/usr/bin/odb --std c++11 --database pgsql --hxx-suffix .h --ixx-suffix
.i --cxx-suffix .cpp --output-dir
${CMAKE_CURRENT_SOURCE_DIR}/odbClasses --generate-query
--generate-schema --schema-format sql

and C++:

g++ -O2 -std=c++11 (4.7.1)

I really don't understand those errors. As a clue, if I comment out
the line tmpPtr.load(), everything compiles fine; and, if instead of
load() I use object_id(), the lazy_ptr returns the correct object ID,
so the lazy_ptr looks well formed and well instantiated. Another
things is that if I use std::weak_ptr (or shared_ptr for the matter)
everything works as expected, except that in some cases it gets on a
recursive loop that causes a crash (that's why I need to use lazy
pointers).

Thanks in advance if you help me out of this :) :) :)

Miguel

From boris at codesynthesis.com  Sun Aug 19 10:08:57 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Sun Aug 19 09:54:02 2012
Subject: [odb-users] Can I define views wherever I want ?
In-Reply-To: <CAAdSf_nYdcuabGPrH1XMtrxOf1DhibJt+_J9mcbW0YBZVvhD_w@mail.gmail.com>
References: <CAAdSf_nYdcuabGPrH1XMtrxOf1DhibJt+_J9mcbW0YBZVvhD_w@mail.gmail.com>
Message-ID: <boris.20120819155954@codesynthesis.com>

Hi,

yb947835908@gmail.com writes:

> Must I define all the views in the .hxx file(the input file of the ODB
> compiler) ? What if I don't know what views I may use at first? Can I
> define views wherever I want to, for example, in my main function?

No, this is not supported.

While I agree, it would neat to be able to define function-local views
(e.g., in main(), as you suggested), this would pose a chicken-and-egg
problem. Consider that in order to use the view, you will need to include
the header file that contains the database support code for this view
(i.e., the -odb.hxx header). So if you define the view in a file, say
main.cxx, where you also use it then this file will need to include
the main-odb.hxx file that is produced by the ODB compiler. But when
you invoke the ODB compiler on main.cxx, no main-odb.hxx is yet
available. So you will get a compile error.

There are also other problems, such as the need for main-odb.hxx to
include the definition of the view (including main.cxx is probably
a bad idea).

Boris

From boris at codesynthesis.com  Sun Aug 19 10:26:37 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Sun Aug 19 10:11:41 2012
Subject: [odb-users] std::set<odb::lazy_shared_ptr<foo>> fails
In-Reply-To: <CAG9fDsWubZJ7Xfq3g44aPH_ee9Ei3JkFaBygJX2F+ByGptVAQg@mail.gmail.com>
References: <CAG9fDsWubZJ7Xfq3g44aPH_ee9Ei3JkFaBygJX2F+ByGptVAQg@mail.gmail.com>
Message-ID: <boris.20120819161101@codesynthesis.com>

Hi Miguel,

Miguel Revilla Rodr?guez <yo@miguelrevilla.com> writes:

> I'm trying to declare an object with a
> std::set<odb::lazy_shared_ptr<foo>> member, and while the odb compiler
> works nicely on it and doesn't complain, later the generated code
> (object-odb.cpp) fails to compile with G++ 4.7.1 with options -O2
> -std=c++11.

When reporting that something fails to compile, it is always a good
idea to include the diagnostics from the compiler. Without it we can
only guess what could have gone wrong.


> If I change std::set with std::vector, then everything is fine, but
> I'd prefer to use a set rather than a vector in this case.

My guess is that the compiler complains there is no operator< defined
for odb::lazy_shared_ptr. The reason for this is that it is tricky to
come up with a universally-acceptable semantics. For example:

1. How do we compare a transient object to a persistent one? Do
   we use the underlying pointers and if so what about unloaded
   pointers? Do we use object ids?

2. When comparing unloaded persistent objects, do we take into
   account the database?

However, if you have some application-specific knowledge that
allows you to provide a sensible less-than comparison, then you
can easily add it as a second template argument to std::set.

For example, if you know that all the objects in your set will
be non-NULL and will have valid object ids (i.e., no object with
yet unassigned auto id will ever be added to the set), and they
will also be from the same database, then you can define an id-
based comparison:

template <typename T>
struct object_id_comparator
{
  bool operator (const odb::lazy_shared_ptr<T>& x,
                 const odb::lazy_shared_ptr<T>& y) const
  {
    return x.object_id<T> () < y.object_id<T> ();
  }
};

Then, in your persistent class, you can have:

std::set<odb::lazy_shared_ptr<foo>, object_id_comparator<foo>> foo_;

Boris

From yo at miguelrevilla.com  Sun Aug 19 10:31:12 2012
From: yo at miguelrevilla.com (=?UTF-8?Q?Miguel_Revilla_Rodr=C3=ADguez?=)
Date: Sun Aug 19 10:31:21 2012
Subject: [odb-users] std::set<odb::lazy_shared_ptr<foo>> fails
In-Reply-To: <boris.20120819161101@codesynthesis.com>
References: <CAG9fDsWubZJ7Xfq3g44aPH_ee9Ei3JkFaBygJX2F+ByGptVAQg@mail.gmail.com>
	<boris.20120819161101@codesynthesis.com>
Message-ID: <CAG9fDsWcFWgD8OYW2J28y2EjmkOm+cnvYbC0qrhurMePhywq9A@mail.gmail.com>

Hi,

I see. Will try it right now. Sorry for not including the compilator
messages. Will inform here if this works.

Miguel

2012/8/19 Boris Kolpackov <boris@codesynthesis.com>:
> Hi Miguel,
>
> Miguel Revilla Rodr?guez <yo@miguelrevilla.com> writes:
>
>> I'm trying to declare an object with a
>> std::set<odb::lazy_shared_ptr<foo>> member, and while the odb compiler
>> works nicely on it and doesn't complain, later the generated code
>> (object-odb.cpp) fails to compile with G++ 4.7.1 with options -O2
>> -std=c++11.
>
> When reporting that something fails to compile, it is always a good
> idea to include the diagnostics from the compiler. Without it we can
> only guess what could have gone wrong.
>
>
>> If I change std::set with std::vector, then everything is fine, but
>> I'd prefer to use a set rather than a vector in this case.
>
> My guess is that the compiler complains there is no operator< defined
> for odb::lazy_shared_ptr. The reason for this is that it is tricky to
> come up with a universally-acceptable semantics. For example:
>
> 1. How do we compare a transient object to a persistent one? Do
>    we use the underlying pointers and if so what about unloaded
>    pointers? Do we use object ids?
>
> 2. When comparing unloaded persistent objects, do we take into
>    account the database?
>
> However, if you have some application-specific knowledge that
> allows you to provide a sensible less-than comparison, then you
> can easily add it as a second template argument to std::set.
>
> For example, if you know that all the objects in your set will
> be non-NULL and will have valid object ids (i.e., no object with
> yet unassigned auto id will ever be added to the set), and they
> will also be from the same database, then you can define an id-
> based comparison:
>
> template <typename T>
> struct object_id_comparator
> {
>   bool operator (const odb::lazy_shared_ptr<T>& x,
>                  const odb::lazy_shared_ptr<T>& y) const
>   {
>     return x.object_id<T> () < y.object_id<T> ();
>   }
> };
>
> Then, in your persistent class, you can have:
>
> std::set<odb::lazy_shared_ptr<foo>, object_id_comparator<foo>> foo_;
>
> Boris

From boris at codesynthesis.com  Sun Aug 19 10:58:16 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Sun Aug 19 10:43:19 2012
Subject: [odb-users] Trouble with lazy_ptr
In-Reply-To: <CAG9fDsUBDnUkM5_yK2xy7YeoDOKih95jMykdzcuykPAkrxkv_A@mail.gmail.com>
References: <CAG9fDsUBDnUkM5_yK2xy7YeoDOKih95jMykdzcuykPAkrxkv_A@mail.gmail.com>
Message-ID: <boris.20120819164907@codesynthesis.com>

Hi Miguel,

> /usr/include/odb/lazy-ptr.ixx:1511:5: error: no match for 'operator='
> in 'r = ((const odb::lazy_weak_ptr<user>*)this)->odb::lazy_weak_ptr<user>::i_.odb::lazy_ptr_impl<T>::load<user>(0)'

I believe this error is due to the incompatible canonical object
pointer used for your 'node' and 'user' classes (see Section 3.2,
"Object and View Pointers" for more information). Specifically,
your 'node' class uses the raw pointer (node*) as the object pointer.
This is incompatible with all the C++11 shared/weak pointers that
you are using in object relationships. So what you need to do is
use std::shared_ptr as your object pointer:

#pragma db object polymorphic pointer(std::shared_ptr)
class node {
  ...
};

Alternatively, you may want to make all your objects use std::shared_ptr
as object pointer by default (this way, if you add another class, you
won't need to worry about forgetting to add the pointer pragma). This
can be done with the --default-pointer ODB compiler option, for example:

--default-pointer std::shared_ptr

Boris

From yo at miguelrevilla.com  Sun Aug 19 10:50:59 2012
From: yo at miguelrevilla.com (=?UTF-8?Q?Miguel_Revilla_Rodr=C3=ADguez?=)
Date: Sun Aug 19 10:51:07 2012
Subject: [odb-users] Trouble with lazy_ptr
In-Reply-To: <boris.20120819164907@codesynthesis.com>
References: <CAG9fDsUBDnUkM5_yK2xy7YeoDOKih95jMykdzcuykPAkrxkv_A@mail.gmail.com>
	<boris.20120819164907@codesynthesis.com>
Message-ID: <CAG9fDsXU+VH=gN7YSd+ELL7Q=xvy=F9dXSVG57RrSJ7k-bqc6A@mail.gmail.com>

Thank you, Boris. You're the best. It's amazing the amount of C++ that
one can learn just by looking at your code and your messages.

Thanks again, will put a huge THANKS in my site when it opens.

Miguel

2012/8/19 Boris Kolpackov <boris@codesynthesis.com>:
> Hi Miguel,
>
>> /usr/include/odb/lazy-ptr.ixx:1511:5: error: no match for 'operator='
>> in 'r = ((const odb::lazy_weak_ptr<user>*)this)->odb::lazy_weak_ptr<user>::i_.odb::lazy_ptr_impl<T>::load<user>(0)'
>
> I believe this error is due to the incompatible canonical object
> pointer used for your 'node' and 'user' classes (see Section 3.2,
> "Object and View Pointers" for more information). Specifically,
> your 'node' class uses the raw pointer (node*) as the object pointer.
> This is incompatible with all the C++11 shared/weak pointers that
> you are using in object relationships. So what you need to do is
> use std::shared_ptr as your object pointer:
>
> #pragma db object polymorphic pointer(std::shared_ptr)
> class node {
>   ...
> };
>
> Alternatively, you may want to make all your objects use std::shared_ptr
> as object pointer by default (this way, if you add another class, you
> won't need to worry about forgetting to add the pointer pragma). This
> can be done with the --default-pointer ODB compiler option, for example:
>
> --default-pointer std::shared_ptr
>
> Boris

From yo at miguelrevilla.com  Sun Aug 19 11:53:34 2012
From: yo at miguelrevilla.com (=?UTF-8?Q?Miguel_Revilla_Rodr=C3=ADguez?=)
Date: Sun Aug 19 11:53:43 2012
Subject: [odb-users] std::set<odb::lazy_shared_ptr<foo>> fails
In-Reply-To: <CAG9fDsWcFWgD8OYW2J28y2EjmkOm+cnvYbC0qrhurMePhywq9A@mail.gmail.com>
References: <CAG9fDsWubZJ7Xfq3g44aPH_ee9Ei3JkFaBygJX2F+ByGptVAQg@mail.gmail.com>
	<boris.20120819161101@codesynthesis.com>
	<CAG9fDsWcFWgD8OYW2J28y2EjmkOm+cnvYbC0qrhurMePhywq9A@mail.gmail.com>
Message-ID: <CAG9fDsXhELDZewZzC6YMwibBZdn7ZJYzPRgxLJrti8m3Hu62rQ@mail.gmail.com>

Everything works perfectly now (including the subject in the other
mail). Thanks for you patience with such a noob :)

Miguel

2012/8/19 Miguel Revilla Rodr?guez <yo@miguelrevilla.com>:
> Hi,
>
> I see. Will try it right now. Sorry for not including the compilator
> messages. Will inform here if this works.
>
> Miguel
>
> 2012/8/19 Boris Kolpackov <boris@codesynthesis.com>:
>> Hi Miguel,
>>
>> Miguel Revilla Rodr?guez <yo@miguelrevilla.com> writes:
>>
>>> I'm trying to declare an object with a
>>> std::set<odb::lazy_shared_ptr<foo>> member, and while the odb compiler
>>> works nicely on it and doesn't complain, later the generated code
>>> (object-odb.cpp) fails to compile with G++ 4.7.1 with options -O2
>>> -std=c++11.
>>
>> When reporting that something fails to compile, it is always a good
>> idea to include the diagnostics from the compiler. Without it we can
>> only guess what could have gone wrong.
>>
>>
>>> If I change std::set with std::vector, then everything is fine, but
>>> I'd prefer to use a set rather than a vector in this case.
>>
>> My guess is that the compiler complains there is no operator< defined
>> for odb::lazy_shared_ptr. The reason for this is that it is tricky to
>> come up with a universally-acceptable semantics. For example:
>>
>> 1. How do we compare a transient object to a persistent one? Do
>>    we use the underlying pointers and if so what about unloaded
>>    pointers? Do we use object ids?
>>
>> 2. When comparing unloaded persistent objects, do we take into
>>    account the database?
>>
>> However, if you have some application-specific knowledge that
>> allows you to provide a sensible less-than comparison, then you
>> can easily add it as a second template argument to std::set.
>>
>> For example, if you know that all the objects in your set will
>> be non-NULL and will have valid object ids (i.e., no object with
>> yet unassigned auto id will ever be added to the set), and they
>> will also be from the same database, then you can define an id-
>> based comparison:
>>
>> template <typename T>
>> struct object_id_comparator
>> {
>>   bool operator (const odb::lazy_shared_ptr<T>& x,
>>                  const odb::lazy_shared_ptr<T>& y) const
>>   {
>>     return x.object_id<T> () < y.object_id<T> ();
>>   }
>> };
>>
>> Then, in your persistent class, you can have:
>>
>> std::set<odb::lazy_shared_ptr<foo>, object_id_comparator<foo>> foo_;
>>
>> Boris

From yb947835908 at gmail.com  Wed Aug 22 11:20:08 2012
From: yb947835908 at gmail.com (=?GB2312?B?1KzUrLHz?=)
Date: Wed Aug 22 11:20:16 2012
Subject: [odb-users] supporting Chinese?
Message-ID: <CAAdSf_nV5OxH-0W7YZiB4WCPmoJdks-ovuG+aRvEpXkPC-nsvg@mail.gmail.com>

Hi!
Does ODB support Chinese word , as the names of the table or column?
If it does, how can I get such supporting?
If it doesn't , what do U suggest to solve this problem if I absolutely
have to use Chinese words as those names?
THX!!!
From oded at geek.co.il  Tue Aug 21 08:55:57 2012
From: oded at geek.co.il (Oded Arbel)
Date: Thu Aug 23 04:52:13 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
Message-ID: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>

Hi ODB people.

I'm writing an application that I want to use with multiple database
servers, from multiple vendors (actually only two - I use MS-SQL as a
"serious database" and SQLite for "database server is optional" mode).

I was thinking of writing database agnostic model (with some #ifdefs in the
more hairy parts, though for most cases ODB vendor agnostic pragmas are
enough), then run the ODB compiler twice to generate two code sets - one
for each database vendor, then build two DLLs out of the whole shebang and
load either one in the application depending on the runtime configuration.

When I started to work on that, I got into a lot of trouble (I'm a C++
newbie, so it may as well be all my fault). Has anyone tried to do
something like this, and if so - what approach did you take?

The main problem I have is that, building for MS-Windows, I have to use
__declspec(dllexport) on the model API, and that causes the VC++ compiler
to get really upset that ODB types do not declare dllexports as well.

-- 
Oded
From boris at codesynthesis.com  Thu Aug 23 06:20:35 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug 23 06:04:57 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
In-Reply-To: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
References: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
Message-ID: <boris.20120823120244@codesynthesis.com>

Hi Oded,

Oded Arbel <oded@geek.co.il> writes:

> I'm writing an application that I want to use with multiple database
> servers, from multiple vendors (actually only two - I use MS-SQL as a
> "serious database" and SQLite for "database server is optional" mode).

This is not yet properly supported by ODB, though it is high up on our
TODO list.


> I was thinking of writing database agnostic model (with some #ifdefs in the
> more hairy parts, though for most cases ODB vendor agnostic pragmas are
> enough), then run the ODB compiler twice to generate two code sets - one
> for each database vendor, then build two DLLs out of the whole shebang and
> load either one in the application depending on the runtime configuration.

That will probably work if you don't use queries. The problem with
queries is that they are a very, very thin layer over the native
SQL queries (and this is done for good reasons, specifically to
minimize code bloat and maintain good performance).

The result of this design is that the actual query code that you
use in your application is very-database specific (i.e., a lot of
database-specific non-virtual/inline functions).

This is actually the biggest obstacle that we need to overcome
in order to provide proper multi-database support. The challenge
here is on the one hand to maintain good performance for applications
that use only one database (probably the majority of them) and on the
other to "abstract" the API sufficiently to allow for multi-database
support.

So if you need queries, the most straightforward approach would
probably be to just build two versions of your application (you
can also have a "starter program" that decided which version to
run). You could also probably go the DLL route, though in this
case you will need to package your whole database-using code 
(and not just the ODB-generated code) into the DLLs.


> The main problem I have is that, building for MS-Windows, I have to use
> __declspec(dllexport) on the model API, and that causes the VC++ compiler
> to get really upset that ODB types do not declare dllexports as well.

That's something that we can fix. However, as I mentioned above, it
probably won't help you much.

Boris

From boris at codesynthesis.com  Thu Aug 23 06:23:12 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug 23 06:07:33 2012
Subject: [odb-users] supporting Chinese?
In-Reply-To: <CAAdSf_nV5OxH-0W7YZiB4WCPmoJdks-ovuG+aRvEpXkPC-nsvg@mail.gmail.com>
References: <CAAdSf_nV5OxH-0W7YZiB4WCPmoJdks-ovuG+aRvEpXkPC-nsvg@mail.gmail.com>
Message-ID: <boris.20120823122049@codesynthesis.com>

Hi,

yb947835908@gmail.com writes:

> Does ODB support Chinese word , as the names of the table or column?

This really depends on the database you are using. Specifically, whether
the native C API supports Unicode in query text and if so, in what form.

So, which database are you using?

Boris

From boris at codesynthesis.com  Mon Aug 27 08:01:30 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Aug 27 07:45:10 2012
Subject: [odb-users] supporting Chinese?
In-Reply-To: <CAAdSf_khp-NzO7GMFNMP7huU8k-5qqkozPx0OdfHK0z7MoykOw@mail.gmail.com>
References: <CAAdSf_nV5OxH-0W7YZiB4WCPmoJdks-ovuG+aRvEpXkPC-nsvg@mail.gmail.com>
	<boris.20120823122049@codesynthesis.com>
	<CAAdSf_khp-NzO7GMFNMP7huU8k-5qqkozPx0OdfHK0z7MoykOw@mail.gmail.com>
Message-ID: <boris.20120827135346@codesynthesis.com>

Hi,

In the future please keep your replies CC'ed to the odb-users mailing
list as discussed in the posting guidelines:

http://www.codesynthesis.com/support/posting-guidelines.xhtml

yb947835908@gmail.com writes:

> Hi,Boris! I'm using Sqlite.As you may know, it's not difficult to write
> some codes to make some conversions so that we could use Chinese words on
> the Sqlite database. But I don't think those codes works on ODB because of
> the generating-code thing. Actaully, I'm totally lost,I don't even know
> where to start with, to write the conversion-codes. So, have you got any
> ideas?

The SQLite C API allows one to execute Unicode statements as long as
they are encoded in UTF-8. This means that you can use Unicode table
and column names as long as you specify them in UTF-8. I modified the
'hello' example to use letter '?' (e with acute, UTF-8 2-byte encoding
is 0xC3, 0x88) in both table and column names:

#pragma db object table("p\xC3\x88rson")
class person
{
  ...

  #pragma db column("ag\xC3\x88")
  unsigned short age_;
};

Everything seems to work fine. So if you want to use Chinese in your
table/column names with SQLite, then you will need to get their UTF-8
encodings and specify those in the table/column pragmas.

Boris

From duhai_lee at qq.com  Mon Aug 27 21:35:25 2012
From: duhai_lee at qq.com (=?gb18030?B?SnVzdGxpdmU=?=)
Date: Tue Aug 28 01:59:26 2012
Subject: [odb-users] =?gb18030?q?support_Multi-field_primary_key=A3=BF?=
Message-ID: <tencent_787E1A504B6FAB4846E0D9BE@qq.com>

Hi,
    there are multiple field primary key table, how can I do?
   THX
From oded at geek.co.il  Tue Aug 28 02:47:31 2012
From: oded at geek.co.il (Oded Arbel)
Date: Tue Aug 28 03:21:45 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
In-Reply-To: <boris.20120823120244@codesynthesis.com>
References: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
	<boris.20120823120244@codesynthesis.com>
Message-ID: <CAOqhUvxrs5N2WqBy4dGFm6=d+9WGOCE==XXV7TxJAkYKd+Qa_w@mail.gmail.com>

On Thu, Aug 23, 2012 at 1:20 PM, Boris Kolpackov
<boris@codesynthesis.com> wrote:
>
> Oded Arbel <oded@geek.co.il> writes:
>
> > I was thinking of writing database agnostic model (with some #ifdefs in
> > the
> > more hairy parts, though for most cases ODB vendor agnostic pragmas are
> > enough), then run the ODB compiler twice to generate two code sets - one
> > for each database vendor, then build two DLLs out of the whole shebang
> > and
> > load either one in the application depending on the runtime
> > configuration.
>
> That will probably work if you don't use queries. The problem with
> queries is that they are a very, very thin layer over the native
> SQL queries (and this is done for good reasons, specifically to
> minimize code bloat and maintain good performance).

The way I'm doing this is to have the model implementation hide all
the queries and expose a business logic API to the rest of the system
(as it is done in MVC). While I expect most queries to be quite
generic (I don't really see how "SELECT * FROM table" is much
different from vendor to vendor), if there is some queries that happen
to be vendor specific I can #ifdef it appropriately. The model
implementation will be packaged with the ODB vendor-specific generated
code into the vendor-specific DLL.

> That's something that we can fix.

I've noticed that ODB code has a lot of LIBODB_EXPORT symbols all over
the place - is thins something I can use to allow VC++ to compile ODB
code into a DLL?

--
Oded

From duhai_lee at qq.com  Tue Aug 28 03:04:22 2012
From: duhai_lee at qq.com (=?gb18030?B?SnVzdGxpdmU=?=)
Date: Tue Aug 28 08:01:46 2012
Subject: [odb-users] =?gb18030?q?How_to_use_wstring_instead_of_string?=
 =?gb18030?b?o78=?=
Message-ID: <tencent_3DFA8FF54B77BDA53B58CA32@qq.com>

Hi there,
    in my .h file, I define a class member of type is std::wstring. 
   then generate code, it show errors: 
error: unable to map C++ type '::std::wstring' used in data member 'Report_Id' to a database type
   how can I do ?
   thx very much.
From boris at codesynthesis.com  Tue Aug 28 08:22:45 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Aug 28 08:07:51 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
In-Reply-To: <CAOqhUvxrs5N2WqBy4dGFm6=d+9WGOCE==XXV7TxJAkYKd+Qa_w@mail.gmail.com>
References: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
	<boris.20120823120244@codesynthesis.com>
	<CAOqhUvxrs5N2WqBy4dGFm6=d+9WGOCE==XXV7TxJAkYKd+Qa_w@mail.gmail.com>
Message-ID: <boris.20120828141910@codesynthesis.com>

Hi Oded,

Oded Arbel <oded@geek.co.il> writes:

> The way I'm doing this is to have the model implementation hide all
> the queries and expose a business logic API to the rest of the system
> (as it is done in MVC). [...] The model implementation will be packaged
> with the ODB vendor-specific generated code into the vendor-specific
> DLL.

Yes, that will work. In other words, your "model implementation" will
wrap the ODB functionality which will not be exposed from the DLL.


> I've noticed that ODB code has a lot of LIBODB_EXPORT symbols all over
> the place - is thins something I can use to allow VC++ to compile ODB
> code into a DLL?

No, these are used from the libodb runtime DLL.

Note that if you use the above approach, you don't need to export any
ODB-generated code from the DLL. The only thing that you should export
is the business logic API.

Boris

From boris at codesynthesis.com  Tue Aug 28 08:30:07 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Aug 28 08:13:38 2012
Subject: [odb-users] support Multi-field primary key?
In-Reply-To: <tencent_787E1A504B6FAB4846E0D9BE@qq.com>
References: <tencent_787E1A504B6FAB4846E0D9BE@qq.com>
Message-ID: <boris.20120828142739@codesynthesis.com>

Hi,

Justlive <duhai_lee@qq.com> writes:

> there are multiple field primary key table, how can I do?

You can use a composite value type as an object id:

#pragma db value
struct name
{
  std::string first;
  std::string last;
};

#pragma db object
class person
{
  ...

  #pragma db id
  name name_;
};

For more information, refer to Section 7.2.1, "Composite Object Ids"
in the ODB manual.

Boris

From boris at codesynthesis.com  Tue Aug 28 08:38:44 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Aug 28 08:22:15 2012
Subject: [odb-users] How to use =?utf-8?Q?wstri?=
	=?utf-8?Q?ng_instead_of_string=EF=BC=9F?=
In-Reply-To: <tencent_3DFA8FF54B77BDA53B58CA32@qq.com>
References: <tencent_3DFA8FF54B77BDA53B58CA32@qq.com>
Message-ID: <boris.20120828143445@codesynthesis.com>

Hi,

Justlive <duhai_lee@qq.com> writes:

> in my .h file, I define a class member of type is std::wstring. 
> then generate code, it show errors: 
> error: unable to map C++ type '::std::wstring' used in data member 'Report_Id' to a database type

Which database are you using? 

Note that it is often impossible to say why things don't work without
at least some information about your setup. At the minimum you should
include the target database, ODB compiler version, operating system
name and version, and C++ compiler name and version that you are using.
For more on this, see the "Posting Guidelines" page:

http://www.codesynthesis.com/support/posting-guidelines.xhtml

Boris

From oded at geek.co.il  Wed Aug 29 11:17:59 2012
From: oded at geek.co.il (Oded Arbel)
Date: Wed Aug 29 11:18:08 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
In-Reply-To: <boris.20120828141910@codesynthesis.com>
References: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
	<boris.20120823120244@codesynthesis.com>
	<CAOqhUvxrs5N2WqBy4dGFm6=d+9WGOCE==XXV7TxJAkYKd+Qa_w@mail.gmail.com>
	<boris.20120828141910@codesynthesis.com>
Message-ID: <CAOqhUvzY1tc3cYM5vwOVR3Bi0w8-U+pT2=5TputsYQWicuAq0g@mail.gmail.com>

Hi Boris.

On Tue, Aug 28, 2012 at 3:22 PM, Boris Kolpackov
<boris@codesynthesis.com> wrote:
>
> Oded Arbel <oded@geek.co.il> writes:
>> The way I'm doing this is to have the model implementation hide all
>> the queries and expose a business logic API to the rest of the system [...]
>
> Yes, that will work. In other words, your "model implementation" will
> wrap the ODB functionality which will not be exposed from the DLL.

It will not expose any ODB functionality through the business logic API, but...

> Note that if you use the above approach, you don't need to export any
> ODB-generated code from the DLL. The only thing that you should export
> is the business logic API.

I still want to expose the model classes themselves, so that the
application can say things like:
vector<Employee> employees = Employee::getAll();
for (vector<Employee>::iterator it = employees.begin(); it !=
employees.end(); ++it)
  it->employer->processSalary(*it);

Where Employee and Employer are the model classes used in ODB, and
getAll() and processSalary() are business logic methods on these
classes.

I don't want to have the application compile against ODB headers
because that causes VC++ to freak out. It should work fine - unless I
use ODB datatypes like odb::nullable, which I need to represent my
model correctly, and at which point application code start seeing
libodb headers and I get export warnings.
-- 
Oded

From oded at geek.co.il  Wed Aug 29 14:13:57 2012
From: oded at geek.co.il (Oded Arbel)
Date: Wed Aug 29 14:14:06 2012
Subject: [odb-users] Self persisting objects?
Message-ID: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>

Hi guys.

I'm trying to write a model class that is self contained - i.e. it
includes all business logic to manipulate the ODB API and does not
require the application to know ODB to operate it.

As part of that, I thought it will be cool (and OOish) to be able to
create and persist objects using a simple API like this:
Employee *e = new Employee("John","Doe");
e->persist();

The implementation of Employee::persist() could be pretty trivial
(assuming someone knows how to initialize a database connection. I'm
going to leave that as an exercise to the user and assume an already
inited static field):
void Employee::persist() {
	odb::transaction t(s_db->begin());
	s_db->persist(this);
	t.commit();
}

Unfortunately Visual Studio 10 doesn't like me much and breaks when
trying to compile.

I've seen database.hxx has a definition for persist(T* o), but the
"this" pointer is "const Class*" which won't fit.

This is what the compiler has to say (actual class names have been
modified to protect the cryptic):
---8<---
1>------ Build started: Project: data-model, Configuration: Debug Win32 ------
1>  employee.cpp
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(103): error C2504:
'odb::access::object_traits<T>' : base class undefined
1>          with
1>          [
1>              T=model::Employee
1>          ]
1>          c:\XXX\data-model\src\employee.cpp(17) : see reference to
class template instantiation 'odb::object_traits<T>' being compiled
1>          with
1>          [
1>              T=model::Employee
1>          ]
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(104): error C2146: syntax error :
missing ',' before identifier 'pointer_type'
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(104): error C2065: 'pointer_type'
: undeclared identifier
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(105): error C2955:
'odb::access::object_factory' : use of class template requires
template argument list
1>          c:\XXX\libodb-2.0.1\odb\traits.hxx(17) : see declaration
of 'odb::access::object_factory'
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2146: syntax error :
missing ';' before identifier 'const_pointer_type'
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C3254:
'odb::object_traits<T>' : class contains explicit override
'const_pointer_type' but does not derive from an interface that
contains the function declaration
1>          with
1>          [
1>              T=model::Employee
1>          ]
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2838:
'const_pointer_type' : illegal qualified name in member declaration
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C4430: missing type
specifier - int assumed. Note: C++ does not support default-int
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2602:
'odb::object_traits<T>::const_pointer_type' is not a member of a base
class of 'odb::object_traits<T>'
1>          with
1>          [
1>              T=model::Employee
1>          ]
1>          c:\XXX\libodb-2.0.1\odb\traits.hxx(119) : see declaration
of 'odb::object_traits<T>::const_pointer_type'
1>          with
1>          [
1>              T=model::Employee
1>          ]
1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2868:
'odb::object_traits<T>::const_pointer_type' : illegal syntax for
using-declaration; expected qualified-name
1>          with
1>          [
1>              T=model::Employee
1>          ]
========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========
---8<---

Any idea what I'm doing wrong here?

-- 
Oded

From oded at geek.co.il  Wed Aug 29 14:39:39 2012
From: oded at geek.co.il (Oded Arbel)
Date: Wed Aug 29 14:39:47 2012
Subject: [odb-users] Re: Self persisting objects?
In-Reply-To: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>
References: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>
Message-ID: <CAOqhUvxubG0GoGu-cyQjfTb5yMHNfaHJNbpwi9oo7Cxhxmr_jg@mail.gmail.com>

Actually, NM - sorry to bother. My mistake that I didn't include the
generated employee-odb.hxx. After that fix it works fine.

On Wed, Aug 29, 2012 at 9:13 PM, Oded Arbel <oded@geek.co.il> wrote:
> Hi guys.
>
> I'm trying to write a model class that is self contained - i.e. it
> includes all business logic to manipulate the ODB API and does not
> require the application to know ODB to operate it.
>
> As part of that, I thought it will be cool (and OOish) to be able to
> create and persist objects using a simple API like this:
> Employee *e = new Employee("John","Doe");
> e->persist();
>
> The implementation of Employee::persist() could be pretty trivial
> (assuming someone knows how to initialize a database connection. I'm
> going to leave that as an exercise to the user and assume an already
> inited static field):
> void Employee::persist() {
>         odb::transaction t(s_db->begin());
>         s_db->persist(this);
>         t.commit();
> }
>
> Unfortunately Visual Studio 10 doesn't like me much and breaks when
> trying to compile.
>
> I've seen database.hxx has a definition for persist(T* o), but the
> "this" pointer is "const Class*" which won't fit.
>
> This is what the compiler has to say (actual class names have been
> modified to protect the cryptic):
> ---8<---
> 1>------ Build started: Project: data-model, Configuration: Debug Win32 ------
> 1>  employee.cpp
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(103): error C2504:
> 'odb::access::object_traits<T>' : base class undefined
> 1>          with
> 1>          [
> 1>              T=model::Employee
> 1>          ]
> 1>          c:\XXX\data-model\src\employee.cpp(17) : see reference to
> class template instantiation 'odb::object_traits<T>' being compiled
> 1>          with
> 1>          [
> 1>              T=model::Employee
> 1>          ]
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(104): error C2146: syntax error :
> missing ',' before identifier 'pointer_type'
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(104): error C2065: 'pointer_type'
> : undeclared identifier
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(105): error C2955:
> 'odb::access::object_factory' : use of class template requires
> template argument list
> 1>          c:\XXX\libodb-2.0.1\odb\traits.hxx(17) : see declaration
> of 'odb::access::object_factory'
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2146: syntax error :
> missing ';' before identifier 'const_pointer_type'
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C3254:
> 'odb::object_traits<T>' : class contains explicit override
> 'const_pointer_type' but does not derive from an interface that
> contains the function declaration
> 1>          with
> 1>          [
> 1>              T=model::Employee
> 1>          ]
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2838:
> 'const_pointer_type' : illegal qualified name in member declaration
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C4430: missing type
> specifier - int assumed. Note: C++ does not support default-int
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2602:
> 'odb::object_traits<T>::const_pointer_type' is not a member of a base
> class of 'odb::object_traits<T>'
> 1>          with
> 1>          [
> 1>              T=model::Employee
> 1>          ]
> 1>          c:\XXX\libodb-2.0.1\odb\traits.hxx(119) : see declaration
> of 'odb::object_traits<T>::const_pointer_type'
> 1>          with
> 1>          [
> 1>              T=model::Employee
> 1>          ]
> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2868:
> 'odb::object_traits<T>::const_pointer_type' : illegal syntax for
> using-declaration; expected qualified-name
> 1>          with
> 1>          [
> 1>              T=model::Employee
> 1>          ]
> ========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========
> ---8<---
>
> Any idea what I'm doing wrong here?
>
> --
> Oded



-- 
Oded

From oded at geek.co.il  Thu Aug 30 05:31:53 2012
From: oded at geek.co.il (Oded Arbel)
Date: Thu Aug 30 05:32:01 2012
Subject: [odb-users] Re: Self persisting objects?
In-Reply-To: <CAOqhUvxubG0GoGu-cyQjfTb5yMHNfaHJNbpwi9oo7Cxhxmr_jg@mail.gmail.com>
References: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>
	<CAOqhUvxubG0GoGu-cyQjfTb5yMHNfaHJNbpwi9oo7Cxhxmr_jg@mail.gmail.com>
Message-ID: <CAOqhUvyP68W_+H0y3mweUv-wQU+3wxfmN5pmAjnv5z6MSjc84A@mail.gmail.com>

Re-correction - apparently it didn't work (I'm not sure why I though
it does), as I get this error during compilation:

libodb-2.0.1\odb\database.ixx(49): error C2440: 'initializing' :
cannot convert from 'model::Employee *' to 'const object_pointer &'


On Wed, Aug 29, 2012 at 9:39 PM, Oded Arbel <oded@geek.co.il> wrote:
> Actually, NM - sorry to bother. My mistake that I didn't include the
> generated employee-odb.hxx. After that fix it works fine.
>
> On Wed, Aug 29, 2012 at 9:13 PM, Oded Arbel <oded@geek.co.il> wrote:
>> Hi guys.
>>
>> I'm trying to write a model class that is self contained - i.e. it
>> includes all business logic to manipulate the ODB API and does not
>> require the application to know ODB to operate it.
>>
>> As part of that, I thought it will be cool (and OOish) to be able to
>> create and persist objects using a simple API like this:
>> Employee *e = new Employee("John","Doe");
>> e->persist();
>>
>> The implementation of Employee::persist() could be pretty trivial
>> (assuming someone knows how to initialize a database connection. I'm
>> going to leave that as an exercise to the user and assume an already
>> inited static field):
>> void Employee::persist() {
>>         odb::transaction t(s_db->begin());
>>         s_db->persist(this);
>>         t.commit();
>> }
>>
>> Unfortunately Visual Studio 10 doesn't like me much and breaks when
>> trying to compile.
>>
>> I've seen database.hxx has a definition for persist(T* o), but the
>> "this" pointer is "const Class*" which won't fit.
>>
>> This is what the compiler has to say (actual class names have been
>> modified to protect the cryptic):
>> ---8<---
>> 1>------ Build started: Project: data-model, Configuration: Debug Win32 ------
>> 1>  employee.cpp
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(103): error C2504:
>> 'odb::access::object_traits<T>' : base class undefined
>> 1>          with
>> 1>          [
>> 1>              T=model::Employee
>> 1>          ]
>> 1>          c:\XXX\data-model\src\employee.cpp(17) : see reference to
>> class template instantiation 'odb::object_traits<T>' being compiled
>> 1>          with
>> 1>          [
>> 1>              T=model::Employee
>> 1>          ]
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(104): error C2146: syntax error :
>> missing ',' before identifier 'pointer_type'
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(104): error C2065: 'pointer_type'
>> : undeclared identifier
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(105): error C2955:
>> 'odb::access::object_factory' : use of class template requires
>> template argument list
>> 1>          c:\XXX\libodb-2.0.1\odb\traits.hxx(17) : see declaration
>> of 'odb::access::object_factory'
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2146: syntax error :
>> missing ';' before identifier 'const_pointer_type'
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C3254:
>> 'odb::object_traits<T>' : class contains explicit override
>> 'const_pointer_type' but does not derive from an interface that
>> contains the function declaration
>> 1>          with
>> 1>          [
>> 1>              T=model::Employee
>> 1>          ]
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2838:
>> 'const_pointer_type' : illegal qualified name in member declaration
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C4430: missing type
>> specifier - int assumed. Note: C++ does not support default-int
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2602:
>> 'odb::object_traits<T>::const_pointer_type' is not a member of a base
>> class of 'odb::object_traits<T>'
>> 1>          with
>> 1>          [
>> 1>              T=model::Employee
>> 1>          ]
>> 1>          c:\XXX\libodb-2.0.1\odb\traits.hxx(119) : see declaration
>> of 'odb::object_traits<T>::const_pointer_type'
>> 1>          with
>> 1>          [
>> 1>              T=model::Employee
>> 1>          ]
>> 1>c:\XXX\libodb-2.0.1\odb\traits.hxx(119): error C2868:
>> 'odb::object_traits<T>::const_pointer_type' : illegal syntax for
>> using-declaration; expected qualified-name
>> 1>          with
>> 1>          [
>> 1>              T=model::Employee
>> 1>          ]
>> ========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========
>> ---8<---
>>
>> Any idea what I'm doing wrong here?
>>
>> --
>> Oded
>
>
>
> --
> Oded



-- 
Oded

From boris at codesynthesis.com  Thu Aug 30 09:25:18 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug 30 09:08:27 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
In-Reply-To: <CAOqhUvzY1tc3cYM5vwOVR3Bi0w8-U+pT2=5TputsYQWicuAq0g@mail.gmail.com>
References: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
	<boris.20120823120244@codesynthesis.com>
	<CAOqhUvxrs5N2WqBy4dGFm6=d+9WGOCE==XXV7TxJAkYKd+Qa_w@mail.gmail.com>
	<boris.20120828141910@codesynthesis.com>
	<CAOqhUvzY1tc3cYM5vwOVR3Bi0w8-U+pT2=5TputsYQWicuAq0g@mail.gmail.com>
Message-ID: <boris.20120830152226@codesynthesis.com>

Hi Oded,

Oded Arbel <oded@geek.co.il> writes:

> I still want to expose the model classes themselves, so that the
> application can say things like:
> vector<Employee> employees = Employee::getAll();
> for (vector<Employee>::iterator it = employees.begin(); it !=
> employees.end(); ++it)
>   it->employer->processSalary(*it);
> 
> Where Employee and Employer are the model classes used in ODB, and
> getAll() and processSalary() are business logic methods on these
> classes.
> 
> I don't want to have the application compile against ODB headers
> because that causes VC++ to freak out. It should work fine - unless I
> use ODB datatypes like odb::nullable, which I need to represent my
> model correctly, and at which point application code start seeing
> libodb headers and I get export warnings.

That is very strange since libodb itself is built as a DLL and
includes all the necessary export/import machinery.

Can you show the errors/warnings you are getting? Better yet,
can you create a test solution that reproduces this problem?

Boris

From boris at codesynthesis.com  Thu Aug 30 09:37:42 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Aug 30 09:20:53 2012
Subject: [odb-users] Re: Self persisting objects?
In-Reply-To: <CAOqhUvyP68W_+H0y3mweUv-wQU+3wxfmN5pmAjnv5z6MSjc84A@mail.gmail.com>
References: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>
	<CAOqhUvxubG0GoGu-cyQjfTb5yMHNfaHJNbpwi9oo7Cxhxmr_jg@mail.gmail.com>
	<CAOqhUvyP68W_+H0y3mweUv-wQU+3wxfmN5pmAjnv5z6MSjc84A@mail.gmail.com>
Message-ID: <boris.20120830152755@codesynthesis.com>

Hi Oded,

Oded Arbel <oded@geek.co.il> writes:

> libodb-2.0.1\odb\database.ixx(49): error C2440: 'initializing' :
> cannot convert from 'model::Employee *' to 'const object_pointer &'

If you look at that line, there is a helpful comment preceding it
that reads:

    // The passed pointer should be the same or implicit-convertible
    // to the object pointer. This way we make sure the object pointer
    // does not assume ownership of the passed object.
    //

So what most likely happens is you are using a smart pointer (e.g.,
shared_ptr) as an object pointer. If that's the case then you may
want to use the shared_from_this (or similar) machinery to get the
shared pointer from this.

Also note that the distinction between persisting via object
pointer and via reference only comes into play if you are using
sessions (if object pointer is not the raw pointer then by-
reference persist() won't add the object to the session). So if
you are not planning to use sessions, then you can instead
replace:

s_db->persist(this);

With:

s_db->persist(*this);


> The implementation of Employee::persist() could be pretty trivial
> (assuming someone knows how to initialize a database connection. I'm
> going to leave that as an exercise to the user and assume an already
> inited static field):
>
> void Employee::persist() {
> 	odb::transaction t(s_db->begin());
> 	s_db->persist(this);
> 	t.commit();
> }

Just a side note: a better design might be to leave it to the caller
of Employee::persist() to create the transaction. This way the caller
will be able to persist several objects at once:

Employee e1, e2;

odb::transaction t (db->begin ());
e1.persist ();
e2.persist ();
t.commit ();

The added benefit of this approach is that you can get the database
from the current transaction without having to store it in a static
field or similar:

void Employee::persist() {
  odb::transaction& t (odb::transaction::current ());
  odb::database& db (t.database ());
  db.persist (*this);
}

Boris

From oded at geek.co.il  Thu Aug 30 12:14:39 2012
From: oded at geek.co.il (Oded Arbel)
Date: Thu Aug 30 12:14:50 2012
Subject: [odb-users] Re: Self persisting objects?
In-Reply-To: <boris.20120830152755@codesynthesis.com>
References: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>
	<CAOqhUvxubG0GoGu-cyQjfTb5yMHNfaHJNbpwi9oo7Cxhxmr_jg@mail.gmail.com>
	<CAOqhUvyP68W_+H0y3mweUv-wQU+3wxfmN5pmAjnv5z6MSjc84A@mail.gmail.com>
	<boris.20120830152755@codesynthesis.com>
Message-ID: <CAOqhUvzoLNSOJEjaKj2NRc6e+JsFd_GeJrnNKYF7V=S2DCd+4Q@mail.gmail.com>

On Thu, Aug 30, 2012 at 4:37 PM, Boris Kolpackov
<boris@codesynthesis.com> wrote:
> Oded Arbel <oded@geek.co.il> writes:
>> libodb-2.0.1\odb\database.ixx(49): error C2440: 'initializing' :
>> cannot convert from 'model::Employee *' to 'const object_pointer &'
>
> So what most likely happens is you are using a smart pointer [...]

Not in this specific piece of code - its just a "this" pointer. s_db
is a smart pointer though - does it matter?

> [...] So if
> you are not planning to use sessions, then you can instead
> replace:
>
> s_db->persist(this);
>
> With:
>
> s_db->persist(*this);

I changed it as you said and it looks to be working fine now. I'm not
using sessions, but its something that I might want to look into in
the future (possibly as an optimization), and I'm a bit confused
regarding why is there a difference in the behavior of "this"
dereferencing in regard to it being held by a smart pointer or not.

> Just a side note: a better design might be to leave it to the caller
> of Employee::persist() to create the transaction. This way the caller
> will be able to persist several objects at once:
> [...]
> The added benefit of this approach is that you can get the database
> from the current transaction without having to store it in a static
> field or similar:
>
> void Employee::persist() {
>   odb::transaction& t (odb::transaction::current ());
>   odb::database& db (t.database ());
>   db.persist (*this);
> }

That is very interesting - I'll have to consider that in the future,
but in the mean time I'm trying to complete a basic implementation
before doing more complex things :-)

-- 
Oded

From oded at geek.co.il  Thu Aug 30 12:19:22 2012
From: oded at geek.co.il (Oded Arbel)
Date: Thu Aug 30 12:19:31 2012
Subject: [odb-users] Support more then one database vendor in a single
	application?
In-Reply-To: <boris.20120830152226@codesynthesis.com>
References: <CAOqhUvzRa1v+wf5N9NcmXciJcGvOrH-U138z5uT9DtYBiHAOfg@mail.gmail.com>
	<boris.20120823120244@codesynthesis.com>
	<CAOqhUvxrs5N2WqBy4dGFm6=d+9WGOCE==XXV7TxJAkYKd+Qa_w@mail.gmail.com>
	<boris.20120828141910@codesynthesis.com>
	<CAOqhUvzY1tc3cYM5vwOVR3Bi0w8-U+pT2=5TputsYQWicuAq0g@mail.gmail.com>
	<boris.20120830152226@codesynthesis.com>
Message-ID: <CAOqhUvy977FZZrQX2CnjwRSVSBZAwXe=HN4C-FkUSXOhPgzbTQ@mail.gmail.com>

On Thu, Aug 30, 2012 at 4:25 PM, Boris Kolpackov
<boris@codesynthesis.com> wrote:
> Oded Arbel <oded@geek.co.il> writes:
>> I don't want to have the application compile against ODB headers
>> because that causes VC++ to freak out. It should work fine - unless I
>> use ODB datatypes like odb::nullable, which I need to represent my
>> model correctly, and at which point application code start seeing
>> libodb headers and I get export warnings.
>
> That is very strange since libodb itself is built as a DLL and
> includes all the necessary export/import machinery.

Well... what I was trying to do, is to build a single DLL that bundles
both libodb, libodb-<vendor> and my model, so I can have only 1 DLL
that needs to be run-time loaded.
It may be that I defined the LIBODB*STATIC* pre-processor definitions
incorrectly in my setup. :-/

> Can you show the errors/warnings you are getting? Better yet,
> can you create a test solution that reproduces this problem?

Currently I reverted to getting everything compiled statically into
the application (with support only for MS-SQL in this build). When I
get my business logic working properly, I'll re-visit this issue and
let you know what's up with this.

-- 
Oded

From boris at codesynthesis.com  Fri Aug 31 07:15:38 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Aug 31 06:58:40 2012
Subject: [odb-users] Re: Self persisting objects?
In-Reply-To: <CAOqhUvzoLNSOJEjaKj2NRc6e+JsFd_GeJrnNKYF7V=S2DCd+4Q@mail.gmail.com>
References: <CAOqhUvz+qfHUc5UhqPTmaRNoFRyeKiifxFJAFb6vRW+deUmYYw@mail.gmail.com>
	<CAOqhUvxubG0GoGu-cyQjfTb5yMHNfaHJNbpwi9oo7Cxhxmr_jg@mail.gmail.com>
	<CAOqhUvyP68W_+H0y3mweUv-wQU+3wxfmN5pmAjnv5z6MSjc84A@mail.gmail.com>
	<boris.20120830152755@codesynthesis.com>
	<CAOqhUvzoLNSOJEjaKj2NRc6e+JsFd_GeJrnNKYF7V=S2DCd+4Q@mail.gmail.com>
Message-ID: <boris.20120831130415@codesynthesis.com>

Hi Oded,

Oded Arbel <oded@geek.co.il> writes:

> > So what most likely happens is you are using a smart pointer [...]
> 
> Not in this specific piece of code - its just a "this" pointer.

I mean your persistent class (Employee) uses a smart pointer as its
object pointer (normally specified with the 'pointer' pragma, the
--default-pointer ODB compiler option, or if you are using Boost
or Qt smart-pointer profiles). See Section 3.2, "Object and View
Pointers" for details.


> I changed it as you said and it looks to be working fine now. I'm not
> using sessions, but its something that I might want to look into in
> the future (possibly as an optimization), and I'm a bit confused
> regarding why is there a difference in the behavior of "this"
> dereferencing in regard to it being held by a smart pointer or not.

If your persistent class uses a smart pointer, say, shared_ptr, then
initializing it from a reference that was passed to persist() (which
is needed in order to enter this object into the session) is a bad
idea. Think what would happen if that reference is already managed
by another shared_ptr instance:

void 
my_persist (Employee& r)
{
  db.persist (r); // Initializing another shared_ptr from &r
                  // is bad news.
}

shared_ptr<Employee> p (new Employee (...));
my_persist (*p);

That's why if you are using smart pointers, persist() will enter
the object into the session only if you pass the smart pointer:

db.persist (p);

If you need to convert 'this' to a shared_ptr, look into the 
shared_from_this mechanism.

Boris

From andrew at nelless.net  Fri Aug 31 07:49:35 2012
From: andrew at nelless.net (Andrew Nelless)
Date: Fri Aug 31 08:34:09 2012
Subject: [odb-users] std::array<char> doesn't work with BYTEA
Message-ID: <CABoWKxjq=WsGd85U97NsnCweFKAakakPtrYV-7DC2ih4EpimPg@mail.gmail.com>

Hi,

I haven't checked 2.1.0a, but I've noticed std::array<char, N> doesn't
work with Postgres BYTEA types in 2.0.1.

This seems like a notable omission feature given that it's the clean
C++11 replacement for char[N], which is supported.

Cheers,

Andrew

From boris at codesynthesis.com  Fri Aug 31 08:56:28 2012
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Aug 31 08:39:30 2012
Subject: [odb-users] std::array<char> doesn't work with BYTEA
In-Reply-To: <CABoWKxjq=WsGd85U97NsnCweFKAakakPtrYV-7DC2ih4EpimPg@mail.gmail.com>
References: <CABoWKxjq=WsGd85U97NsnCweFKAakakPtrYV-7DC2ih4EpimPg@mail.gmail.com>
Message-ID: <boris.20120831145536@codesynthesis.com>

Hi Andrew,

Andrew Nelless <andrew@nelless.net> writes:

> I haven't checked 2.1.0a, but I've noticed std::array<char, N> doesn't
> work with Postgres BYTEA types in 2.0.1.
> 
> This seems like a notable omission feature given that it's the clean
> C++11 replacement for char[N], which is supported.

Yes, this is still on my TODO list for 2.1.0.

Boris

From candy.chiu.ad at gmail.com  Fri Aug 31 10:52:18 2012
From: candy.chiu.ad at gmail.com (Candy Chiu)
Date: Fri Aug 31 10:52:26 2012
Subject: [odb-users] Can odb be used without touching the domain object
	files?
Message-ID: <CADH8t_WeE2YDJEGw0f3a8mN-HcD1-OLDupHy--E4ZXtYid3v3g@mail.gmail.com>

Hi,

I'd like to add db capability to a project without modifying the existing
.h's.  Is there another way to achieve mapping in this case?

Thanks.
Candy
From hugirat at gmail.com  Fri Aug 31 17:20:26 2012
From: hugirat at gmail.com (Yihya Hugirat)
Date: Fri Aug 31 17:20:34 2012
Subject: [odb-users] odb and QT
Message-ID: <CAHpdtTOva3jKc6V_Sh76-qEogxoT3M5doE4_S_9v1RT6MmqqRw@mail.gmail.com>

Hi all,
how to use odb with qt in qt creator.

thank you.