From boris at codesynthesis.com  Mon Dec  2 05:43:38 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Dec  2 05:46:03 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
Message-ID: <boris.20131202124025@codesynthesis.com>

Hi Troy,

Troy Heron <troy.heron@hixxy.org> writes:

> Is there much involved in the implementation or is it mostly a matter
> of turning it on?

All the underlying machinery is already there (we use ON DELETE CASCADE
internally to implement containers). So it is a matter of adding the
pragmas, diagnosing all the corner cases, adding a test, and writing
the documentation. In fact, I've done the first three this morning.

Can you let me know which ODB compiler package you are using?

Boris

From astek.rbasset at tisseo.fr  Mon Dec  2 06:09:51 2013
From: astek.rbasset at tisseo.fr (astek.rbasset)
Date: Mon Dec  2 06:48:55 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
Message-ID: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>

Hello,

  I'm trying to build my first ODB project on Windows XP with MinGW
*.hpp sources DO compile on linux platform, but Windows compilation gets an internal compiler error :

8< ------------------------------------- >8
D:\users\<Project>\src>odb -d sqlite -p boost -I D:\users\<Project>\includes -I D:\boost_1_55_0 --default-pointer boost::shared_ptr -o odb_obj_headers -q -s -e odb_obj_headers/*.hpp
*** WARNING *** there are active plugins, do not report this as a bug unless you can reproduce it without enabling any plugins.
Event                            | Plugins
PLUGIN_START_UNIT                | odb
PLUGIN_PRAGMAS                   | odb
PLUGIN_OVERRIDE_GATE             | odb
In file included from <standard-odb-epilogue>:8:0:
D:\users\<Project>\includes/odb/tr1/pointer-traits.hxx: In destructor 'virtual boost::exception_detail::clone_base::~clone_base()':
D:\users\<Project>\includes/odb/tr1/pointer-traits.hxx:116:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://www.codesynthesis.com/products/odb/> for instructions.
8< ------------------------------------- >8

Note that I had to add boost complete library in include path (D:\boost_1_55_0) because of this previous error :

8< ------------------------------------- >8
D:\users\<Project>\src>odb -d sqlite -p boost --default-pointer boost::shared_ptr -o odb_obj_headers -q -s -e odb_obj_headers/*.hpp
In file included from <odb-prologue-1>:1:0:
d:\odb\odb-2.3.0-i686-windows\mingw\bin\../../mingw/include/odb/boost/date-time/sqlite/gregorian-mapping.hxx:8:57: fatal error: boost/date_time/gregorian/gregorian_types.hpp: No such file or directory
compilation terminated.
8< ------------------------------------- >8

ODB : 2.3.0
BOOST : 1.55
MinGW : latest form http://sourceforge.net/projects/mingw/

Any help is welcome !

Rodolphe

From boris at codesynthesis.com  Mon Dec  2 06:55:50 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Dec  2 06:58:14 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
Message-ID: <boris.20131202135325@codesynthesis.com>

Hi Rodolphe,

astek.rbasset <astek.rbasset@tisseo.fr> writes:

> D:\users\<Project>\src>odb -d sqlite -p boost -I \
> D:\users\<Project>\includes -I D:\boost_1_55_0 --default-pointer \
> boost::shared_ptr -o odb_obj_headers -q -s -e odb_obj_headers/*.hpp

Can you try to compile each header file in odb_obj_headers individually
and see which one causes the ODB compiler to crash. If you could then
send this file to me (you can send it off-list if you want to keep it
private), then I could try to figure out what's wrong.

Boris

From astek.rbasset at tisseo.fr  Mon Dec  2 08:38:22 2013
From: astek.rbasset at tisseo.fr (astek.rbasset)
Date: Mon Dec  2 09:15:35 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <boris.20131202135325@codesynthesis.com>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202135325@codesynthesis.com>
Message-ID: <F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>

Hi Boris,

Thanks for your attention.
3 / 7 are not compiling, 
4 / 7 are properly compiling.

Here's two of them not compiling (following private email) 

Control.hpp
Inspector.hpp

I cannot see what makes the difference.

Was I wrong in adding "-I D:\boost_1_55_0", could be another version / implementation of boost lib ?

Thanks for helping,

Rodolphe



-----Message d'origine-----
De?: Boris Kolpackov [mailto:boris@codesynthesis.com] 
Envoy??: lundi 2 d?cembre 2013 12:56
??: astek.rbasset
Cc?: odb-users@codesynthesis.com
Objet?: Re: [odb-users] internal compiler error in pointer-traits.hxx

Hi Rodolphe,

astek.rbasset <astek.rbasset@tisseo.fr> writes:

> D:\users\<Project>\src>odb -d sqlite -p boost -I \ 
> D:\users\<Project>\includes -I D:\boost_1_55_0 --default-pointer \ 
> boost::shared_ptr -o odb_obj_headers -q -s -e odb_obj_headers/*.hpp

Can you try to compile each header file in odb_obj_headers individually and see which one causes the ODB compiler to crash. If you could then send this file to me (you can send it off-list if you want to keep it private), then I could try to figure out what's wrong.

Boris

From boris at codesynthesis.com  Mon Dec  2 09:18:01 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Dec  2 09:20:26 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202135325@codesynthesis.com>
	<F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>
Message-ID: <boris.20131202161322@codesynthesis.com>

Hi Rodolphe,

Ok, I've  managed to reproduce the crash on both Windows and Linux.
The problem is with the combination of the const pointed-to object:

boost::shared_ptr<const LineSpot> _lineSpot;

And Boost 1.55.0. That is, if I remove const from the above line,
everything compiles fine. Similarly, if I use an earlier version
of Boost (I tried 1.52.0) leaving const in place, everything
compiles fine as well.

I am still going to investigate further (and hopefully fix) this
strange combination. In the meantime you can use these two changes
as potential workarounds.

Boris

From astek.rbasset at tisseo.fr  Mon Dec  2 10:32:13 2013
From: astek.rbasset at tisseo.fr (astek.rbasset)
Date: Mon Dec  2 13:24:48 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <boris.20131202161322@codesynthesis.com>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202135325@codesynthesis.com>
	<F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202161322@codesynthesis.com>
Message-ID: <F85A7BD3E12C1E43915F755E08B7E7570200FB1784A8@SRV-MSG01.tisseo-exp.dom>

Hi Boris,

Thanks a lot for this diagnostic !
You're right, well-compiled files has no boost::shared_ptr<const ?> use !

Indeed, linux compiles with libodb-2.2.so and libboost_system.so.5, which seems to be older...

I'll remove temporarily the const keyword to go further with my project.
All *.hpp is compiled !

Thanks a lot !

Rodolphe

-----Message d'origine-----
De?: Boris Kolpackov [mailto:boris@codesynthesis.com] 
Envoy??: lundi 2 d?cembre 2013 15:18
??: astek.rbasset
Cc?: odb-users@codesynthesis.com
Objet?: Re: [odb-users] internal compiler error in pointer-traits.hxx

Hi Rodolphe,

Ok, I've  managed to reproduce the crash on both Windows and Linux.
The problem is with the combination of the const pointed-to object:

boost::shared_ptr<const LineSpot> _lineSpot;

And Boost 1.55.0. That is, if I remove const from the above line, everything compiles fine. Similarly, if I use an earlier version of Boost (I tried 1.52.0) leaving const in place, everything compiles fine as well.

I am still going to investigate further (and hopefully fix) this strange combination. In the meantime you can use these two changes as potential workarounds.

Boris

From troy.heron at hixxy.org  Tue Dec  3 00:46:18 2013
From: troy.heron at hixxy.org (Troy Heron)
Date: Tue Dec  3 00:46:26 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <boris.20131202124025@codesynthesis.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
Message-ID: <CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>

I am using the most recent package from the website:

ODB object-relational mapping (ORM) compiler for C++ 2.3.0

I have another question ;)

Initially I was planning on using the object's 'name' as the primary key,
but my application requires that this no be static (objects can be
renamed). It seems that this would not be possible with foreign key
constraints. So... I've changed all the objects to have a separate id field
as primary key and made the name a unique constraint instead. Now I'm
having troubles with using load() as at the time of loading an object, I
know the name, not the id. Of course, I could query for the object, but
that doesn't return a shared_ptr like load does.

Any idea what the best way to deal with this situation is? I need a primary
key that can change without breaking foreign key constraints or to be able
to load using a field other than the primary key.

I may have missed something in the docs (which are great btw)... any help
appreciated.

Thanks,
Troy


On Mon, Dec 2, 2013 at 8:43 PM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Troy,
>
> Troy Heron <troy.heron@hixxy.org> writes:
>
> > Is there much involved in the implementation or is it mostly a matter
> > of turning it on?
>
> All the underlying machinery is already there (we use ON DELETE CASCADE
> internally to implement containers). So it is a matter of adding the
> pragmas, diagnosing all the corner cases, adding a test, and writing
> the documentation. In fact, I've done the first three this morning.
>
> Can you let me know which ODB compiler package you are using?
>
> Boris
>
From troy.heron at hixxy.org  Tue Dec  3 00:49:34 2013
From: troy.heron at hixxy.org (Troy Heron)
Date: Tue Dec  3 00:49:41 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
	<CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
Message-ID: <CAMy2XzBvFsePB7-0WxxZySQz2=j9CUQqdxGW4YVx+A9W-O=+Ew@mail.gmail.com>

Further to that previous email, I'm using this horrible hack to work around
to issue until I figure out what to do in the long term:

http://pastebin.com/4ebvY6Pw



On Tue, Dec 3, 2013 at 3:46 PM, Troy Heron <troy.heron@hixxy.org> wrote:

> I am using the most recent package from the website:
>
> ODB object-relational mapping (ORM) compiler for C++ 2.3.0
>
> I have another question ;)
>
> Initially I was planning on using the object's 'name' as the primary key,
> but my application requires that this no be static (objects can be
> renamed). It seems that this would not be possible with foreign key
> constraints. So... I've changed all the objects to have a separate id field
> as primary key and made the name a unique constraint instead. Now I'm
> having troubles with using load() as at the time of loading an object, I
> know the name, not the id. Of course, I could query for the object, but
> that doesn't return a shared_ptr like load does.
>
> Any idea what the best way to deal with this situation is? I need a
> primary key that can change without breaking foreign key constraints or to
> be able to load using a field other than the primary key.
>
> I may have missed something in the docs (which are great btw)... any help
> appreciated.
>
> Thanks,
> Troy
>
>
> On Mon, Dec 2, 2013 at 8:43 PM, Boris Kolpackov <boris@codesynthesis.com>wrote:
>
>> Hi Troy,
>>
>> Troy Heron <troy.heron@hixxy.org> writes:
>>
>> > Is there much involved in the implementation or is it mostly a matter
>> > of turning it on?
>>
>> All the underlying machinery is already there (we use ON DELETE CASCADE
>> internally to implement containers). So it is a matter of adding the
>> pragmas, diagnosing all the corner cases, adding a test, and writing
>> the documentation. In fact, I've done the first three this morning.
>>
>> Can you let me know which ODB compiler package you are using?
>>
>> Boris
>>
>
>
From boris at codesynthesis.com  Tue Dec  3 01:47:52 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Tue Dec  3 01:50:17 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
	<CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
Message-ID: <boris.20131203084017@codesynthesis.com>

Hi Troy,

Troy Heron <troy.heron@hixxy.org> writes:

> I am using the most recent package from the website

Are you using a pre-built binary for the ODB compiler, a .deb
package, or .rpm? Also which platform (Linux, Windows, etc.),
and architecture (i686, x86_64, etc)?

I will need to build you one for the beta-testing.


> Initially I was planning on using the object's 'name' as the primary key,
> but my application requires that this no be static (objects can be
> renamed). It seems that this would not be possible with foreign key
> constraints. So... I've changed all the objects to have a separate id field
> as primary key and made the name a unique constraint instead. Now I'm
> having troubles with using load() as at the time of loading an object, I
> know the name, not the id. Of course, I could query for the object, but
> that doesn't return a shared_ptr like load does.
> 
> Any idea what the best way to deal with this situation is? I need a primary
> key that can change without breaking foreign key constraints or to be able
> to load using a field other than the primary key.

Having a separate id (probably auto) makes sense. What you are missing
is the fact that you can load the object from the query result iterator:

auto r = db->query<object>(query<object>::name == name);
shared_ptr<object> o = r.begin ().load ();

This is discussed in Section 4.4, "Query Result".

Also, if you are concerned with performance of this approach (compared
to database::load()), then you may want to consider preparing and
caching the query (Section 4.5, "Prepared Queries"). This way it will
be as efficient as database::load().

Boris

From troy.heron at hixxy.org  Tue Dec  3 02:03:14 2013
From: troy.heron at hixxy.org (Troy Heron)
Date: Tue Dec  3 02:03:21 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <boris.20131203084017@codesynthesis.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
	<CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
	<boris.20131203084017@codesynthesis.com>
Message-ID: <CAMy2XzD63wwGJ4p8qUTJWWQD0S=Pq6hiW3yWvwUL8GSNufiv5A@mail.gmail.com>

I'm using the pre-built .deb ODB compiler on x68_64 Linux.

I completely missed that load() member on the iterator. Thanks, exactly
what I was looking for.


On Tue, Dec 3, 2013 at 4:47 PM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Troy,
>
> Troy Heron <troy.heron@hixxy.org> writes:
>
> > I am using the most recent package from the website
>
> Are you using a pre-built binary for the ODB compiler, a .deb
> package, or .rpm? Also which platform (Linux, Windows, etc.),
> and architecture (i686, x86_64, etc)?
>
> I will need to build you one for the beta-testing.
>
>
> > Initially I was planning on using the object's 'name' as the primary key,
> > but my application requires that this no be static (objects can be
> > renamed). It seems that this would not be possible with foreign key
> > constraints. So... I've changed all the objects to have a separate id
> field
> > as primary key and made the name a unique constraint instead. Now I'm
> > having troubles with using load() as at the time of loading an object, I
> > know the name, not the id. Of course, I could query for the object, but
> > that doesn't return a shared_ptr like load does.
> >
> > Any idea what the best way to deal with this situation is? I need a
> primary
> > key that can change without breaking foreign key constraints or to be
> able
> > to load using a field other than the primary key.
>
> Having a separate id (probably auto) makes sense. What you are missing
> is the fact that you can load the object from the query result iterator:
>
> auto r = db->query<object>(query<object>::name == name);
> shared_ptr<object> o = r.begin ().load ();
>
> This is discussed in Section 4.4, "Query Result".
>
> Also, if you are concerned with performance of this approach (compared
> to database::load()), then you may want to consider preparing and
> caching the query (Section 4.5, "Prepared Queries"). This way it will
> be as efficient as database::load().
>
> Boris
>
From Mikhail.Tomilov at infotecs.ru  Tue Dec  3 08:32:48 2013
From: Mikhail.Tomilov at infotecs.ru (Mikhail Tomilov)
Date: Tue Dec  3 08:37:17 2013
Subject: [odb-users] Postgresql column with default value
Message-ID: <529DDD80.5060506@infotecs.ru>

Hello!

We have postgresql 9.1 table:

CREATE TABLE test (
     id bigserial primary key,
     some_field text not null default some_heavy_computation_procedure()
);

C++ code:
#pragma db object
struct test
{
     #pragma id auto
     long long id;

     std::string some_field;
};

test t;
db.persist( t );
std::cout << t.some_field; // expecting to get actual DB value 
calculated by some_heavy_computation_procedure()

With code above we get this SQL:
INSERT INTO test(id, some_field) VALUES(DEFAULT, $1) RETURNING id;

But we need to get SQL query like this:
INSERT INTO test(id, some_field) VALUES(DEFAULT, DEFAULT) RETURNING id, 
some_field;

We want to persist this class instance with DB default value for 
some_field. some_heavy_computation_procedure() is stored procedure that 
performs complex calculation we cannot implement in C++ code (for some 
reason).

We considered to use callbacks but didn't get it worked the way we want 
(described above). Also we tried to use struct like this:

#pragma db object table("test")
struct test_ins
{
     #pragma id auto
     long long id;

     // Intentionally absent some_field
};

This doesn't suit for polymorphic objects (because of typeid column and 
we had to perform additional select query after persist call).

What is the best approach to solve this problem?

Regards, Mikhail.

From boris at codesynthesis.com  Wed Dec  4 03:02:29 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Dec  4 03:04:57 2013
Subject: [odb-users] Postgresql column with default value
In-Reply-To: <529DDD80.5060506@infotecs.ru>
References: <529DDD80.5060506@infotecs.ru>
Message-ID: <boris.20131204094602@codesynthesis.com>

Hi Mikhail,

Mikhail Tomilov <Mikhail.Tomilov@infotecs.ru> writes:

> We considered to use callbacks but didn't get it worked the way we
> want (described above).

Unfortunately there is still no support for the RETURNING clause
in ODB. It is on our TODO list but is not yet implemented.

For now, the workaround is the same as what I have described in
my previous reply:

http://www.codesynthesis.com/pipermail/odb-users/2013-July/001457.html

That is, send the NULL value on persist() and load the computed
data member in a callback. Unfortunately, this will require a
separate query, but that's the best that can be done until proper
support for RETURNING is implemented.

Here is an outline:

// file: test.hxx
//

#include <odb/nullable.hxx>

#pragma db object callback(init)
struct test
{
  #pragma id auto
  long long id;

  odb::nullable<std::string> some_field;

  void
  init (odb::callback_event, odb::database&) const;
};

#pragma db view object(test)
struct test_computed
{
  std::string some_field;
};


// file: test.cxx
//

void test::
init (odb::callback_event e, odb::database& db) const
{
  switch (e)
  {
  case odb::callback_event::pre_persist:
    {
      // Reset some_field to NULL so that the default value is
      // assigned by the database.
      //
      const_cast<test*> (this)->some_field.reset ();
      break;
    }
  case odb::callback_event::post_persist:
    {
      // Load the computed fields using the test_computed view.
      //
      test_computed v;
      db.query<test_computed> (
        odb::query<test_computed>::id == id).begin ().load (v);
      const_cast<test*> (this)->some_field = v.some_field;
      break;
    }
  }
}

You can further optimize this by preparing and caching the query
used to load computed members (Section 4.5, "Prepared Queries").

Boris

From boris at codesynthesis.com  Wed Dec  4 05:28:38 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Dec  4 05:31:02 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <boris.20131202161322@codesynthesis.com>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202135325@codesynthesis.com>
	<F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202161322@codesynthesis.com>
Message-ID: <boris.20131204122555@codesynthesis.com>

Hi Rodolphe,

Boris Kolpackov <boris@codesynthesis.com> writes:

> I am still going to investigate further (and hopefully fix) this
> strange combination.

Ok, I've fixed this. I am not planning to make a bug-fix release
yet for this issue since it is fairly obscure. You can continue
using one of the workarounds I mentioned or I can build you a
pre-release binary.

There is also the patch if anyone is interested:

http://scm.codesynthesis.com/?p=odb/odb.git;a=commit;h=ab4abe514a617ef530a18d5f71950d082ebd1ac4

Boris

From astek.rbasset at tisseo.fr  Thu Dec  5 04:13:14 2013
From: astek.rbasset at tisseo.fr (astek.rbasset)
Date: Thu Dec  5 06:00:03 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <boris.20131204122555@codesynthesis.com>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202135325@codesynthesis.com>
	<F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202161322@codesynthesis.com>
	<boris.20131204122555@codesynthesis.com>
Message-ID: <F85A7BD3E12C1E43915F755E08B7E7570200FB1785E3@SRV-MSG01.tisseo-exp.dom>

Hi Boris,

  Good work !
'Hope I made things progress... though it wasn't my first goal ;-)
A pre-release is welcome, avoid me to fork from linux sources on my project to build on windows.

My current build is "odb-2.3.0-i686-windows" (Win32/XP)

Thanks a lot,

Rodolphe

-----Message d'origine-----
De?: Boris Kolpackov [mailto:boris@codesynthesis.com] 
Envoy??: mercredi 4 d?cembre 2013 11:29
??: astek.rbasset
Cc?: odb-users@codesynthesis.com
Objet?: Re: [odb-users] internal compiler error in pointer-traits.hxx

Hi Rodolphe,

Boris Kolpackov <boris@codesynthesis.com> writes:

> I am still going to investigate further (and hopefully fix) this 
> strange combination.

Ok, I've fixed this. I am not planning to make a bug-fix release yet for this issue since it is fairly obscure. You can continue using one of the workarounds I mentioned or I can build you a pre-release binary.

There is also the patch if anyone is interested:

http://scm.codesynthesis.com/?p=odb/odb.git;a=commit;h=ab4abe514a617ef530a18d5f71950d082ebd1ac4

Boris

From Aleksey.Enakaev at infotecs.ru  Fri Dec  6 01:27:34 2013
From: Aleksey.Enakaev at infotecs.ru (=?UTF-8?B?0JXQvdCw0LrQsNC10LIg0JDQu9C10LrRgdC10Lk=?=)
Date: Fri Dec  6 02:42:37 2013
Subject: [odb-users] ODB pgsql compiling problem
Message-ID: <52A16E56.2010006@infotecs.ru>

Hi.

First, I want to thank you for answers to our previous emails.

Now, I'd like to inform you about a problem we discovered.
If we describe some DB view in a separate file, then generate ODB code 
(files *.odb.cpp) and try compiling them we will get a compiler error 
message. (Note: it's all about pgsql code generation). Compiler says:

".../odb/pgsql/view-result.hxx:68:7: error: ?view_traits_calls? does not 
name a type"

Investigation we had taken showed that there was a private field 
view_traits_calls<view_type> tc_ added to class view_result_impl. Class 
declaration view_traits_calls<T> was found in odb/pgsql/traits-calls.hxx 
file.
The problem has been solved after we added #include 
<odb/pgsql/traits-calls.hxx> to "odb/pgsql/view-result.hxx".
(Note: the same ODB code generated for the sqlite database is compiled 
without error.)

We suppose you'd like to know about this problem to fix it in next versions.

Best regards, Aleksey, Mikhail.

From boris at codesynthesis.com  Fri Dec  6 03:14:05 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Dec  6 03:16:29 2013
Subject: [odb-users] libodb-pgsql 2.3.1 bugfix released
Message-ID: <boris.20131206100911@codesynthesis.com>

Hi,

A new bugfix release for the PostgreSQL ODB runtime library (libodb-pgsql)
is now available. It adds a missing include file that affected compilation
of generated files that contain only views. For more details on this issue,
see the following mailing list thread:

http://www.codesynthesis.com/pipermail/odb-users/2013-December/001653.html

You can download the new packages from the ODB download page:

http://www.codesynthesis.com/products/odb/download.xhtml

SHA1 checksums for the files in this release are as follows:

91f5688decb0269f630f7a88d16a5a73b9ebb284  libodb-pgsql-2.3.1.tar.bz2
8b1874747580ac840f4b22c4b7b13a1523a9f270  libodb-pgsql-2.3.1.tar.gz
985b191ff1c84fcf94400db70b219023bfdfffd3  libodb-pgsql-2.3.1.zip

Boris

From boris at codesynthesis.com  Fri Dec  6 03:22:29 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Dec  6 03:24:54 2013
Subject: [odb-users] ODB pgsql compiling problem
In-Reply-To: <52A16E56.2010006@infotecs.ru>
References: <52A16E56.2010006@infotecs.ru>
Message-ID: <boris.20131206102035@codesynthesis.com>

Hi Aleksey,

??????? ??????? <Aleksey.Enakaev@infotecs.ru> writes:

> The problem has been solved after we added #include
> <odb/pgsql/traits-calls.hxx> to "odb/pgsql/view-result.hxx".

Yes, that's a bug. I've just released a bugfix for libodb-pgsql:

http://www.codesynthesis.com/pipermail/odb-announcements/2013/000038.html

Thanks for reporting it!

Boris

From boris at codesynthesis.com  Fri Dec  6 05:36:24 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Dec  6 05:38:48 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <CAMy2XzD63wwGJ4p8qUTJWWQD0S=Pq6hiW3yWvwUL8GSNufiv5A@mail.gmail.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
	<CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
	<boris.20131203084017@codesynthesis.com>
	<CAMy2XzD63wwGJ4p8qUTJWWQD0S=Pq6hiW3yWvwUL8GSNufiv5A@mail.gmail.com>
Message-ID: <boris.20131206123329@codesynthesis.com>

Hi Troy,

Troy Heron <troy.heron@hixxy.org> writes:

> I'm using the pre-built .deb ODB compiler on x68_64 Linux.

Ok, the pre-release is ready:

http://www.codesynthesis.com/~boris/tmp/odb/pre-release/

Note that besides installing the ODB compiler, you will also need
to build and install new runtime libraries (libodb, etc).

The documentation for the new on_delete pragma can be found in Section
14.4.15, "on_delete" in the ODB manual that comes with the ODB compiler
(installed into /usr/share/doc/odb/).

Let me know how it goes.

Boris

From boris at codesynthesis.com  Fri Dec  6 05:37:43 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Dec  6 05:40:08 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <CAJA0JPAVN57wLE7CRx2zdm7r=yUFHmMgKbXu04R+TyvRcaqA4A@mail.gmail.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAJA0JPAVN57wLE7CRx2zdm7r=yUFHmMgKbXu04R+TyvRcaqA4A@mail.gmail.com>
Message-ID: <boris.20131206123626@codesynthesis.com>

Hi Per,

Per Edin <info@peredin.com> writes:

> I heavily rely on CASCADE in other projects although I have yet to
> experience any problems with ODB's lack of supportt. If you implement
> support I would happily beta-test it!

Thanks. The pre-release is here:

http://www.codesynthesis.com/~boris/tmp/odb/pre-release/

If you need a different package for the ODB compiler, let me know.

Boris

From boris at codesynthesis.com  Fri Dec  6 05:39:47 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Dec  6 05:42:12 2013
Subject: [odb-users] internal compiler error in pointer-traits.hxx
In-Reply-To: <F85A7BD3E12C1E43915F755E08B7E7570200FB1785E3@SRV-MSG01.tisseo-exp.dom>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB17847F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202135325@codesynthesis.com>
	<F85A7BD3E12C1E43915F755E08B7E7570200FB17848F@SRV-MSG01.tisseo-exp.dom>
	<boris.20131202161322@codesynthesis.com>
	<boris.20131204122555@codesynthesis.com>
	<F85A7BD3E12C1E43915F755E08B7E7570200FB1785E3@SRV-MSG01.tisseo-exp.dom>
Message-ID: <boris.20131206123749@codesynthesis.com>

Hi Rodolphe,

astek.rbasset <astek.rbasset@tisseo.fr> writes:

> A pre-release is welcome, avoid me to fork from linux sources on my
> project to build on windows.
> 
> My current build is "odb-2.3.0-i686-windows" (Win32/XP)

I've packaged a pre-release with the fix which you can find here:

http://www.codesynthesis.com/~boris/tmp/odb/pre-release/

Note that besides the ODB compiler itself, you will also need to
build new runtimes (libodb, etc).

Let me know if there are any problems.

Boris

From scott.deerwester at gmail.com  Fri Dec  6 10:37:22 2013
From: scott.deerwester at gmail.com (Scott Deerwester)
Date: Fri Dec  6 10:37:51 2013
Subject: [odb-users] ODB segfaults in Qt profile
Message-ID: <CAK-De8cFyvcL51FaUbf_kHAKZp9tP6Th9CSw3aL_DN_12RJGCA@mail.gmail.com>

I am receiving the following error:

odb --database mysql --profile qt --generate-schema --generate-query
--generate-session -I/opt/Qt/5.1.1/gcc_64/include
-I/opt/Qt/5.1.1/gcc_64/include/QtCore ../ODBExample/ChildClass.h
*** WARNING *** there are active plugins, do not report this as a bug
unless you can reproduce it without enabling any plugins.
Event                            | Plugins
PLUGIN_START_UNIT                | odb
PLUGIN_PRAGMAS                   | odb
PLUGIN_OVERRIDE_GATE             | odb
In file included from <standard-odb-epilogue>:8:0:
/usr/local/include/odb/tr1/pointer-traits.hxx: In member function ?virtual
const char* __gnu_cxx::__concurrence_lock_error::what() const?:
/usr/local/include/odb/tr1/pointer-traits.hxx:116:1: internal compiler
error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
make: *** [ChildClass-odb.cxx] Error 1

Here is file ChildClass.h:

#ifndef CHILDCLASS_H

#define CHILDCLASS_H


#include <QString>

#include <QSharedPointer>


#include <odb/core.hxx>


class ParentClass;


#pragma db object

class ChildClass

{

public:

    ChildClass();

    const QString getID() const;

    void setID(QString &aID);

private:

    friend class odb::access;

#pragma db id

    QString id_;

    QSharedPointer<const ParentClass> parent_;

};


#endif // CHILDCLASS_H


and here is file ParentClass.h:


#ifndef PARENTCLASS_H

#define PARENTCLASS_H

#include <QString>

#include <QMap>

#include <QSharedPointer>

#include <odb/core.hxx>

#include "ChildClass.h"

#pragma db object

class ParentClass

{

public:

    ParentClass();

    ParentClass(QString aID);

    const QString getID() const;

    void setID(QString &aID);

private:

    friend class odb::access;

#pragma db id

    QString id_;

    bool bool_;

    int int_;

#pragma db not_null inverse(parent_)

    QMap<QString,QSharedPointer<const ChildClass> > children_;

};

#endif // PARENTCLASS_H

I am running Qt 5.1.1, odb 2.3.0 and gcc 4.6.3 on Ubuntu 12.04. Any ideas?
From boris at codesynthesis.com  Fri Dec  6 11:20:51 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Fri Dec  6 11:23:16 2013
Subject: [odb-users] ODB segfaults in Qt profile
In-Reply-To: <CAK-De8cFyvcL51FaUbf_kHAKZp9tP6Th9CSw3aL_DN_12RJGCA@mail.gmail.com>
References: <CAK-De8cFyvcL51FaUbf_kHAKZp9tP6Th9CSw3aL_DN_12RJGCA@mail.gmail.com>
Message-ID: <boris.20131206181503@codesynthesis.com>

Hi Scott,

Scott Deerwester <scott.deerwester@gmail.com> writes:

>     QSharedPointer<const ParentClass> parent_;>
>
> [...]
>
>     QMap<QString,QSharedPointer<const ChildClass> > children_;

I am pretty sure this is the same bug that was reported a few
days ago:

http://www.codesynthesis.com/pipermail/odb-users/2013-December/001640.html

Can you try to remove the const's and see if it help. If it does then
there is a pre-release available with the fix:

http://www.codesynthesis.com/~boris/tmp/odb/pre-release/

Note that you will also need to update the runtime libraries (libodb,
etc) besides the ODB compiler.

Or, if you built the ODB compiler from source, you can apply the patch:

http://scm.codesynthesis.com/?p=odb/odb.git;a=commit;h=ab4abe514a617ef530a18d5f71950d082ebd1ac4

Boris

From scott.deerwester at gmail.com  Fri Dec  6 11:47:03 2013
From: scott.deerwester at gmail.com (Scott Deerwester)
Date: Fri Dec  6 11:47:31 2013
Subject: [odb-users] ODB segfaults in Qt profile
In-Reply-To: <boris.20131206181503@codesynthesis.com>
References: <CAK-De8cFyvcL51FaUbf_kHAKZp9tP6Th9CSw3aL_DN_12RJGCA@mail.gmail.com>
	<boris.20131206181503@codesynthesis.com>
Message-ID: <CAK-De8fvbiSQF-yGOTjiCz-3-ZoJjEVS9auD4qXyRk+xqoxLPQ@mail.gmail.com>

That did it, thank you. I'll load the prerelease and try putting the consts
back.

On Fri, Dec 6, 2013 at 11:20 AM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Scott,
>
> Scott Deerwester <scott.deerwester@gmail.com> writes:
>
> >     QSharedPointer<const ParentClass> parent_;>
> >
> > [...]
> >
> >     QMap<QString,QSharedPointer<const ChildClass> > children_;
>
> I am pretty sure this is the same bug that was reported a few
> days ago:
>
> http://www.codesynthesis.com/pipermail/odb-users/2013-December/001640.html
>
> Can you try to remove the const's and see if it help. If it does then
> there is a pre-release available with the fix:
>
> http://www.codesynthesis.com/~boris/tmp/odb/pre-release/
>
> Note that you will also need to update the runtime libraries (libodb,
> etc) besides the ODB compiler.
>
> Or, if you built the ODB compiler from source, you can apply the patch:
>
>
> http://scm.codesynthesis.com/?p=odb/odb.git;a=commit;h=ab4abe514a617ef530a18d5f71950d082ebd1ac4
>
> Boris
>
From troy.heron at hixxy.org  Fri Dec  6 16:00:52 2013
From: troy.heron at hixxy.org (Troy Heron)
Date: Fri Dec  6 16:01:01 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <boris.20131206123329@codesynthesis.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
	<CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
	<boris.20131203084017@codesynthesis.com>
	<CAMy2XzD63wwGJ4p8qUTJWWQD0S=Pq6hiW3yWvwUL8GSNufiv5A@mail.gmail.com>
	<boris.20131206123329@codesynthesis.com>
Message-ID: <CAMy2XzAcQuXByu=ivCckDTR14RAUXMnajrvWoGPo=heC_Exkeg@mail.gmail.com>

Thanks Boris. Will give it a work over sometime today or tomorrow.

I'm just looking at using --include-regex at the moment. What regex engine
is it using? Does it support negative lookbehind?

Thanks


On Fri, Dec 6, 2013 at 8:36 PM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Troy,
>
> Troy Heron <troy.heron@hixxy.org> writes:
>
> > I'm using the pre-built .deb ODB compiler on x68_64 Linux.
>
> Ok, the pre-release is ready:
>
> http://www.codesynthesis.com/~boris/tmp/odb/pre-release/
>
> Note that besides installing the ODB compiler, you will also need
> to build and install new runtime libraries (libodb, etc).
>
> The documentation for the new on_delete pragma can be found in Section
> 14.4.15, "on_delete" in the ODB manual that comes with the ODB compiler
> (installed into /usr/share/doc/odb/).
>
> Let me know how it goes.
>
> Boris
>
From troy.heron at hixxy.org  Fri Dec  6 16:54:48 2013
From: troy.heron at hixxy.org (Troy Heron)
Date: Fri Dec  6 16:54:55 2013
Subject: [odb-users] Problems with foreign key constraints when erasing
In-Reply-To: <CAMy2XzAcQuXByu=ivCckDTR14RAUXMnajrvWoGPo=heC_Exkeg@mail.gmail.com>
References: <CAMy2XzCs5f0aN26Xn0DzEmbf549garRuLok16xzJ9zmfG5C-iw@mail.gmail.com>
	<boris.20131130114625@codesynthesis.com>
	<CAMy2XzB5_EeaMHmWR47DVToCanOM3bPJ=-4taqnGXfOc7QJH=g@mail.gmail.com>
	<boris.20131202124025@codesynthesis.com>
	<CAMy2XzAPUpMn8F8uMThe8E-uYEqxscusFJ0bvMcO1Ne4=G+ffQ@mail.gmail.com>
	<boris.20131203084017@codesynthesis.com>
	<CAMy2XzD63wwGJ4p8qUTJWWQD0S=Pq6hiW3yWvwUL8GSNufiv5A@mail.gmail.com>
	<boris.20131206123329@codesynthesis.com>
	<CAMy2XzAcQuXByu=ivCckDTR14RAUXMnajrvWoGPo=heC_Exkeg@mail.gmail.com>
Message-ID: <CAMy2XzDAU_7mhWBfWa+-g_yyBu-Fg09uDUJYEFq-nDJqfsf6tw@mail.gmail.com>

Nevermind. I've discovered what I need to know. It does support negative
look-behind and was only user-error of course :)


On Sat, Dec 7, 2013 at 7:00 AM, Troy Heron <troy.heron@hixxy.org> wrote:

> Thanks Boris. Will give it a work over sometime today or tomorrow.
>
> I'm just looking at using --include-regex at the moment. What regex engine
> is it using? Does it support negative lookbehind?
>
> Thanks
>
>
> On Fri, Dec 6, 2013 at 8:36 PM, Boris Kolpackov <boris@codesynthesis.com>wrote:
>
>> Hi Troy,
>>
>> Troy Heron <troy.heron@hixxy.org> writes:
>>
>> > I'm using the pre-built .deb ODB compiler on x68_64 Linux.
>>
>> Ok, the pre-release is ready:
>>
>> http://www.codesynthesis.com/~boris/tmp/odb/pre-release/
>>
>> Note that besides installing the ODB compiler, you will also need
>> to build and install new runtime libraries (libodb, etc).
>>
>> The documentation for the new on_delete pragma can be found in Section
>> 14.4.15, "on_delete" in the ODB manual that comes with the ODB compiler
>> (installed into /usr/share/doc/odb/).
>>
>> Let me know how it goes.
>>
>> Boris
>>
>
>
From lidia at lemur-soft.com  Tue Dec 10 15:05:55 2013
From: lidia at lemur-soft.com (Lidia Kalinovsky)
Date: Tue Dec 10 15:06:01 2013
Subject: [odb-users] is it possible to use same relationship table for 2
 class data members ?
Message-ID: <CAG4H3oO42mWLb7DCNw8uDiQa4VqtUuDLMZaj05LErBm=r0fypg@mail.gmail.com>

Hello

For example, something like this :

// contained object
#pragma db object
class B
{
private:
friend class odb::access;
  int m;
};

// container
#pragma db object
class A
{
  public:
...
   private:

    friend class odb::access;

   #pragma db table("A_bs")
  std::unordered_map<unsigned long, std::shared_ptr<B> , std::hash<unsigned
long>> m_bsMap;

 std::vector<std::shared_ptr<B> > m_bs;
};


( i see solution with defining one member as transient and using db
callback functions to update/load it ).

Any advice/suggestion is really appreciated.

Thanks.
Lidia.
From arnoud.van-der-wal at nz.abb.com  Tue Dec 10 15:37:37 2013
From: arnoud.van-der-wal at nz.abb.com (Arnoud van-der-Wal)
Date: Tue Dec 10 15:37:54 2013
Subject: [odb-users] Sqlite auto ID persists with ID=NULL returns NULL?
Message-ID: <OF5C10ED67.3BB0F796-ONCC257C3D.006E5DF3-CC257C3D.00714EB3@au.abb.com>

We are trying to use the nullable ID entries but hitting issues with the 
returned ID.

Sqlite has the feature that it can create unique primarily keys when the 
ID is set to NULL - without needing to use AUTOINCREMENT  the "auto" 
pragma .
In our app we like to avoid to use the AUTOINCREMENT.

When doing a persists()  and selecting NULL for ID  in ODB 2.3.0  then 
Sqlite assigns an correct ID and writes the table rows in the database.
But for some reason the db->persists()  does not return the assigned ID 
and it just returns the id as NULL ( null_ set true).  That is a problem, 
because we need to use the ID.
When selecting  pragma 'auto'  using  autoincrement it works correctly and 
returns the created ID.

Example class

#pragma db object  table("test")  pointer(*) 
class tOrm_test_table 
{
public:
  friend class odb::access;
  #pragma db id  type("INTEGER") column("id") 
  odb::nullable<int> m_id; 
  #pragma db type("INTEGER") column("is_integer") 
  int       mInteger; 
};

In the code used the sequence is similar to

  odb::database * db (  new odb::sqlite::database ( "\\test.db", 
(int)SQLITE_OPEN_READWRITE | SQLITE_OPEN_CREATE, false, "" ));
  tOrm_test_table test;
  test.m_id.reset();
  odb::transaction t ( db->begin() );
  odb::nullable<int> id;
  id = db->persist ( &test );
  t.commit();


What do we wrong? How can we obtain the ID of the just persist class?

Thanks,

Arnoud
From boris at codesynthesis.com  Wed Dec 11 00:07:51 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Dec 11 00:10:15 2013
Subject: [odb-users] Sqlite auto ID persists with ID=NULL returns NULL?
In-Reply-To: <OF5C10ED67.3BB0F796-ONCC257C3D.006E5DF3-CC257C3D.00714EB3@au.abb.com>
References: <OF5C10ED67.3BB0F796-ONCC257C3D.006E5DF3-CC257C3D.00714EB3@au.abb.com>
Message-ID: <boris.20131211070335@codesynthesis.com>

Hi Arnoud,

Arnoud van-der-Wal <arnoud.van-der-wal@nz.abb.com> writes:

> When doing a persists()  and selecting NULL for ID  in ODB 2.3.0  then 
> Sqlite assigns an correct ID and writes the table rows in the database.
> But for some reason the db->persists()  does not return the assigned ID 
> and it just returns the id as NULL ( null_ set true).

The reason for this is that without auto ODB doesn't know that you are
using database-generated IDs.

The ODB way of doing what you want is to use the --sqlite-lax-auto-id
option. Quoting the documentation:

"Do not force monotonically increasing automatically-assigned object ids. In
 this mode the generated database schema omits the AUTOINCREMENT keyword which
 results in faster object persistence but may lead to automatically-assigned
 ids not being in a strictly ascending order. Refer to the SQLite
 documentation for details."

The good thing with this approach is that you don't need to mess with NULLs;
ODB will do everything automatically for you.

Boris

From boris at codesynthesis.com  Wed Dec 11 00:14:15 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Dec 11 00:16:38 2013
Subject: [odb-users] is it possible to use same relationship table for 2
	class data members ?
In-Reply-To: <CAG4H3oO42mWLb7DCNw8uDiQa4VqtUuDLMZaj05LErBm=r0fypg@mail.gmail.com>
References: <CAG4H3oO42mWLb7DCNw8uDiQa4VqtUuDLMZaj05LErBm=r0fypg@mail.gmail.com>
Message-ID: <boris.20131211070903@codesynthesis.com>

Hi Lidia,

Lidia Kalinovsky <lidia@lemur-soft.com> writes:

>  #pragma db table("A_bs")
>  std::unordered_map<unsigned long, std::shared_ptr<B> , std::hash<unsigned long>> m_bsMap;
> 
>  std::vector<std::shared_ptr<B> > m_bs;
>
> ( i see solution with defining one member as transient and using db
> callback functions to update/load it ).

Yes, this is probably the most straightforward way.

If you have many objects with such containers (or containers of
this template), then it may be better to create your own container
wrapper that maintains the two underlying containers. You can then
add a custom mapping for such a container for ODB. Let me know if
you are interested in this approach and I can provide more details.

Boris

From lidia at lemur-soft.com  Wed Dec 11 01:44:05 2013
From: lidia at lemur-soft.com (Lidia Kalinovsky)
Date: Wed Dec 11 01:44:13 2013
Subject: [odb-users] is it possible to use same relationship table for 2
	class data members ?
In-Reply-To: <boris.20131211070903@codesynthesis.com>
References: <CAG4H3oO42mWLb7DCNw8uDiQa4VqtUuDLMZaj05LErBm=r0fypg@mail.gmail.com>
	<boris.20131211070903@codesynthesis.com>
Message-ID: <CAG4H3oOaukPp+dDvsU-O7mdERku0kLyS16eLKP61jXxJmG0NXQ@mail.gmail.com>

Hello Boris.

We definitely will try to use custom mapping odb functionality.
If you could recommend some known payment free solution for map
implementation
that works well as std::map and has directs "values" function ( get vector
of values without iterating ), it will be great.

Thanks a lot for quick response and great library.

Lidia.



On Wed, Dec 11, 2013 at 7:14 AM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Lidia,
>
> Lidia Kalinovsky <lidia@lemur-soft.com> writes:
>
> >  #pragma db table("A_bs")
> >  std::unordered_map<unsigned long, std::shared_ptr<B> ,
> std::hash<unsigned long>> m_bsMap;
> >
> >  std::vector<std::shared_ptr<B> > m_bs;
> >
> > ( i see solution with defining one member as transient and using db
> > callback functions to update/load it ).
>
> Yes, this is probably the most straightforward way.
>
> If you have many objects with such containers (or containers of
> this template), then it may be better to create your own container
> wrapper that maintains the two underlying containers. You can then
> add a custom mapping for such a container for ODB. Let me know if
> you are interested in this approach and I can provide more details.
>
> Boris
>



-- 
Software integration and outsourcing services,
Lemur-Soft, Giv'at Nili
Israel, 37825
Phone : (+972) 545748325
Fax : (+972) 775345383
Email : lidia@lemur-soft.com
Web: www.lemur-soft.com
From boris at codesynthesis.com  Wed Dec 11 02:03:04 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Dec 11 02:05:27 2013
Subject: [odb-users] is it possible to use same relationship table for 2
	class data members ?
In-Reply-To: <CAG4H3oOaukPp+dDvsU-O7mdERku0kLyS16eLKP61jXxJmG0NXQ@mail.gmail.com>
References: <CAG4H3oO42mWLb7DCNw8uDiQa4VqtUuDLMZaj05LErBm=r0fypg@mail.gmail.com>
	<boris.20131211070903@codesynthesis.com>
	<CAG4H3oOaukPp+dDvsU-O7mdERku0kLyS16eLKP61jXxJmG0NXQ@mail.gmail.com>
Message-ID: <boris.20131211090049@codesynthesis.com>

Hi Lidia,

Lidia Kalinovsky <lidia@lemur-soft.com> writes:

> If you could recommend some known payment free solution for map
> implementation that works well as std::map and has directs "values"
> function ( get vector of values without iterating ), it will be great.

Boost multi-index container. The best part is that the ODB Boost profile
already includes a mapping for this container (Section 23.3, "Multi-Index
Container Library").

Boris

From lidia at lemur-soft.com  Wed Dec 11 03:06:19 2013
From: lidia at lemur-soft.com (Lidia Kalinovsky)
Date: Wed Dec 11 03:06:26 2013
Subject: [odb-users] is it possible to use same relationship table for 2
	class data members ?
In-Reply-To: <boris.20131211090049@codesynthesis.com>
References: <CAG4H3oO42mWLb7DCNw8uDiQa4VqtUuDLMZaj05LErBm=r0fypg@mail.gmail.com>
	<boris.20131211070903@codesynthesis.com>
	<CAG4H3oOaukPp+dDvsU-O7mdERku0kLyS16eLKP61jXxJmG0NXQ@mail.gmail.com>
	<boris.20131211090049@codesynthesis.com>
Message-ID: <CAG4H3oN49-m=S6a05SuPXDit7KQ=qdG_9Sq1gSh8ay-D_ZtXOg@mail.gmail.com>

looks great and fit our needs - thanks !


On Wed, Dec 11, 2013 at 9:03 AM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Lidia,
>
> Lidia Kalinovsky <lidia@lemur-soft.com> writes:
>
> > If you could recommend some known payment free solution for map
> > implementation that works well as std::map and has directs "values"
> > function ( get vector of values without iterating ), it will be great.
>
> Boost multi-index container. The best part is that the ODB Boost profile
> already includes a mapping for this container (Section 23.3, "Multi-Index
> Container Library").
>
> Boris
>



-- 
Software integration and outsourcing services,
Lemur-Soft, Giv'at Nili
Israel, 37825
Phone : (+972) 545748325
Fax : (+972) 775345383
Email : lidia@lemur-soft.com
Web: www.lemur-soft.com
From scott.deerwester at gmail.com  Wed Dec 11 10:30:59 2013
From: scott.deerwester at gmail.com (Scott Deerwester)
Date: Wed Dec 11 10:31:28 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
Message-ID: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>

I'm receiving the following in a unit test:

tst_MyTest: statement.cxx:377: odb::mysql::select_statement::result
odb::mysql::select_statement::fetch(bool): Assertion
`mysql_stmt_field_count (stmt_) == count' failed.

The situation is that I've just called db.find() with a key that doesn't
exist, and then try again with a key that does. If you need me to, I'll try
to write a minimal set of code that demonstrates the problem.
From boris at codesynthesis.com  Wed Dec 11 10:43:11 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Wed Dec 11 10:45:35 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
In-Reply-To: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
References: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
Message-ID: <boris.20131211174107@codesynthesis.com>

Hi Scott,

Scott Deerwester <scott.deerwester@gmail.com> writes:

> I'm receiving the following in a unit test:
> 
> tst_MyTest: statement.cxx:377: odb::mysql::select_statement::result
> odb::mysql::select_statement::fetch(bool): Assertion
> `mysql_stmt_field_count (stmt_) == count' failed.

The comment just before the assert() explains what it tries to catch:

// Make sure that the number of columns in the result returned by
// the database matches the number that we expect. A common cause
// of this assertion is a native view with a number of data members
// not matching the number of columns in the SELECT-list.

Though it seems like you are loading an object so this can't be that.


> If you need me to, I'll try to write a minimal set of code that
> demonstrates the problem.

Yes, that would be very helpful.

Boris

From scott.deerwester at gmail.com  Wed Dec 11 11:28:39 2013
From: scott.deerwester at gmail.com (Scott Deerwester)
Date: Wed Dec 11 11:29:07 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
In-Reply-To: <boris.20131211174107@codesynthesis.com>
References: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
	<boris.20131211174107@codesynthesis.com>
Message-ID: <CAK-De8cnVh328pRFZURajzPEu8+Hog8Sv6nc2XjkDF1NMjTWCQ@mail.gmail.com>

Hmm. I *do* have a view defined for the class that's not working. I don't
see the view in the code in the schema for this class, and so the view
hasn't been defined in the database. I define the view after the class in
the same header:

#pragma db value(QString) type("VARCHAR(255)")
#pragma db object
class CodeSet
{
...
#pragma db id
    QString key_;		// Identifier
...
};

#pragma db view object(CodeSet)
struct codeSetStat {
#pragma db column("count(" + CodeSet::key_ + ")")
    int count;
};


Do I need to do something special to get the definition for the view?

On Wed, Dec 11, 2013 at 10:43 AM, Boris Kolpackov
<boris@codesynthesis.com>wrote:

> Hi Scott,
>
> Scott Deerwester <scott.deerwester@gmail.com> writes:
>
> > I'm receiving the following in a unit test:
> >
> > tst_MyTest: statement.cxx:377: odb::mysql::select_statement::result
> > odb::mysql::select_statement::fetch(bool): Assertion
> > `mysql_stmt_field_count (stmt_) == count' failed.
>
> The comment just before the assert() explains what it tries to catch:
>
> // Make sure that the number of columns in the result returned by
> // the database matches the number that we expect. A common cause
> // of this assertion is a native view with a number of data members
> // not matching the number of columns in the SELECT-list.
>
> Though it seems like you are loading an object so this can't be that.
>
>
> > If you need me to, I'll try to write a minimal set of code that
> > demonstrates the problem.
>
> Yes, that would be very helpful.
>
> Boris
>
>
From scott.deerwester at gmail.com  Wed Dec 11 16:11:26 2013
From: scott.deerwester at gmail.com (Scott Deerwester)
Date: Wed Dec 11 16:11:53 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
In-Reply-To: <CAK-De8cnVh328pRFZURajzPEu8+Hog8Sv6nc2XjkDF1NMjTWCQ@mail.gmail.com>
References: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
	<boris.20131211174107@codesynthesis.com>
	<CAK-De8cnVh328pRFZURajzPEu8+Hog8Sv6nc2XjkDF1NMjTWCQ@mail.gmail.com>
Message-ID: <CAK-De8fjwiLbc=qidyP1MrUXGDreTysKKJxEW_int2mx9tEX9g@mail.gmail.com>

Okay, I've got a minimal example for you, but it's got nine files in it.
How would you like me to get it to you? I presume that an email attachment
to the list wouldn't be a very good idea...


On Wed, Dec 11, 2013 at 11:28 AM, Scott Deerwester <
scott.deerwester@gmail.com> wrote:

> Hmm. I *do* have a view defined for the class that's not working. I don't
> see the view in the code in the schema for this class, and so the view
> hasn't been defined in the database. I define the view after the class in
> the same header:
>
> #pragma db value(QString) type("VARCHAR(255)")
> #pragma db object
> class CodeSet
> {
> ...
> #pragma db id
>     QString key_;		// Identifier
> ...
> };
>
> #pragma db view object(CodeSet)
> struct codeSetStat {
> #pragma db column("count(" + CodeSet::key_ + ")")
>     int count;
> };
>
>
> Do I need to do something special to get the definition for the view?
>
> On Wed, Dec 11, 2013 at 10:43 AM, Boris Kolpackov <boris@codesynthesis.com
> > wrote:
>
>> Hi Scott,
>>
>> Scott Deerwester <scott.deerwester@gmail.com> writes:
>>
>> > I'm receiving the following in a unit test:
>> >
>> > tst_MyTest: statement.cxx:377: odb::mysql::select_statement::result
>> > odb::mysql::select_statement::fetch(bool): Assertion
>> > `mysql_stmt_field_count (stmt_) == count' failed.
>>
>> The comment just before the assert() explains what it tries to catch:
>>
>> // Make sure that the number of columns in the result returned by
>> // the database matches the number that we expect. A common cause
>> // of this assertion is a native view with a number of data members
>> // not matching the number of columns in the SELECT-list.
>>
>> Though it seems like you are loading an object so this can't be that.
>>
>>
>> > If you need me to, I'll try to write a minimal set of code that
>> > demonstrates the problem.
>>
>> Yes, that would be very helpful.
>>
>> Boris
>>
>>
>
From boris at codesynthesis.com  Thu Dec 12 03:19:40 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Dec 12 03:22:04 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
In-Reply-To: <CAK-De8fjwiLbc=qidyP1MrUXGDreTysKKJxEW_int2mx9tEX9g@mail.gmail.com>
References: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
	<boris.20131211174107@codesynthesis.com>
	<CAK-De8cnVh328pRFZURajzPEu8+Hog8Sv6nc2XjkDF1NMjTWCQ@mail.gmail.com>
	<CAK-De8fjwiLbc=qidyP1MrUXGDreTysKKJxEW_int2mx9tEX9g@mail.gmail.com>
Message-ID: <boris.20131212101343@codesynthesis.com>

Hi Scott,

Thanks for sending the test case. The problematic part is this line:

#pragma db not_null inverse(parent_)
QMap<QString, QSharedPointer<ChildClass> > children_;

An inverse container of pointers cannot be a map (remember that there
is no table for an inverse container so there is no place to store the
key). We should actually diagnose this in the ODB compiler and I've
added this check for the next release.

Once I've changed QMap to QSet, I ran into another problem in your
code: since you have a bi-directional relationship between ChildClass
and ParentClass, you need to use a session (object cache) when loading
them. Otherwise ODB will keep loading them recursively until it runs
out of stack:
    
    transaction t(this->db->begin());

    odb::session s;

    QSharedPointer<ParentClass> parent;

    if (!this->db->find<ParentClass>(aID)) {
        qDebug() << "Failed to find parent" << aID;
    }

    t.commit();

With these two fixes your test runs fine.

Boris

From astek.rbasset at tisseo.fr  Thu Dec 12 05:46:04 2013
From: astek.rbasset at tisseo.fr (astek.rbasset)
Date: Thu Dec 12 07:32:42 2013
Subject: [odb-users] include odb/details/config.h on Windows / Eclipse CDT /
	MinGW ?
Message-ID: <F85A7BD3E12C1E43915F755E08B7E7570200FB1787BF@SRV-MSG01.tisseo-exp.dom>

Hello,

  I'm trying to build C++ application with odb libs to access an sqlite DB.
I want it to be portable Linux / Windows.
Linux building is OK (different version of ODB / Boost, but same source)

On Windows XP, I'm using Eclipse (Kepler) CDT and MinGW as a compiler.

My includes are the following ones :
<language name="C++ Source File">
<includepath>D:\MinGW\msys\1.0\include</includepath>
<includepath workspace_path="true">/PlanControle/includes</includepath>
<includepath>D:\boost_1_55_0</includepath>
<includepath>D:\odb\libodb-2.3.0</includepath>
<includepath>D:\odb\libodb-boost-2.3.0</includepath>
<includepath>D:\odb\libodb-sqlite-2.3.0</includepath>
<includepath>D:\odb\odb-2.3.0-i686-windows\mingw\include</includepath>
</language>

I'm not compiling with MSVC, but I am on windows target.
My question : which conf should be preprocessed in config.hxx ?

#ifdef _MSC_VER
#  include <odb/details/config-vc.h>
#elif defined(ODB_COMPILER)
#  ...
#else
#  include <odb/details/config.h>
#  ...
#endif

Do I need to define _MSC_VER ? ODB_COMPILER ?

Nowadays, building cannot find
odb/details/config.h
nor
odb/details/config-vc.h


I'm a bit confused...

Thanks for any help.

Rodolphe
From boris at codesynthesis.com  Thu Dec 12 07:37:35 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Dec 12 07:39:59 2013
Subject: [odb-users] include odb/details/config.h on Windows / Eclipse
	CDT / MinGW ?
In-Reply-To: <F85A7BD3E12C1E43915F755E08B7E7570200FB1787BF@SRV-MSG01.tisseo-exp.dom>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB1787BF@SRV-MSG01.tisseo-exp.dom>
Message-ID: <boris.20131212143405@codesynthesis.com>

Hi Rodolphe,

astek.rbasset <astek.rbasset@tisseo.fr> writes:

> <includepath>D:\odb\odb-2.3.0-i686-windows\mingw\include</includepath>

You should not include this directory. It is internal ODB compiler
details.


> My question : which conf should be preprocessed in config.hxx ?

It should be autoconf-generated odb/details/config.h found in
D:\odb\libodb-2.3.0. Have you actually built libodb for MinGW?


> Do I need to define _MSC_VER ? ODB_COMPILER ?

No, you should not define either of those.

Boris

From scott.deerwester at gmail.com  Thu Dec 12 10:18:02 2013
From: scott.deerwester at gmail.com (Scott Deerwester)
Date: Thu Dec 12 10:18:29 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
In-Reply-To: <boris.20131212101343@codesynthesis.com>
References: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
	<boris.20131211174107@codesynthesis.com>
	<CAK-De8cnVh328pRFZURajzPEu8+Hog8Sv6nc2XjkDF1NMjTWCQ@mail.gmail.com>
	<CAK-De8fjwiLbc=qidyP1MrUXGDreTysKKJxEW_int2mx9tEX9g@mail.gmail.com>
	<boris.20131212101343@codesynthesis.com>
Message-ID: <CAK-De8f8BKg=CWuUXmz2asSti8NYqRC1=YdcppqWkw3bHA1sUQ@mail.gmail.com>

Thanks very much, Boris. One more question. What do I do if I need to get
an ordered list of the ChildClass ID's that are associated with a
ParentClass object? That was the motivation for using a QMap. I was using a
QMap iterator, but QSet doesn't do that. Any ideas?


On Thu, Dec 12, 2013 at 3:19 AM, Boris Kolpackov <boris@codesynthesis.com>wrote:

> Hi Scott,
>
> Thanks for sending the test case. The problematic part is this line:
>
> #pragma db not_null inverse(parent_)
> QMap<QString, QSharedPointer<ChildClass> > children_;
>
> An inverse container of pointers cannot be a map (remember that there
> is no table for an inverse container so there is no place to store the
> key). We should actually diagnose this in the ODB compiler and I've
> added this check for the next release.
>
> Once I've changed QMap to QSet, I ran into another problem in your
> code: since you have a bi-directional relationship between ChildClass
> and ParentClass, you need to use a session (object cache) when loading
> them. Otherwise ODB will keep loading them recursively until it runs
> out of stack:
>
>     transaction t(this->db->begin());
>
>     odb::session s;
>
>     QSharedPointer<ParentClass> parent;
>
>     if (!this->db->find<ParentClass>(aID)) {
>         qDebug() << "Failed to find parent" << aID;
>     }
>
>     t.commit();
>
> With these two fixes your test runs fine.
>
> Boris
>
From astek.rbasset at tisseo.fr  Thu Dec 12 12:11:02 2013
From: astek.rbasset at tisseo.fr (astek.rbasset)
Date: Thu Dec 12 12:54:57 2013
Subject: [odb-users] include odb/details/config.h on Windows / Eclipse
	CDT / MinGW ?
In-Reply-To: <boris.20131212143405@codesynthesis.com>
References: <F85A7BD3E12C1E43915F755E08B7E7570200FB1787BF@SRV-MSG01.tisseo-exp.dom>
	<boris.20131212143405@codesynthesis.com>
Message-ID: <F85A7BD3E12C1E43915F755E08B7E7570200FB1787DF@SRV-MSG01.tisseo-exp.dom>

Hi Boris,

>> <includepath>D:\odb\odb-2.3.0-i686-windows\mingw\include</includepath>
> You should not include this directory. It is internal ODB compiler details.

OK

>> My question : which conf should be preprocessed in config.hxx ?
> It should be autoconf-generated odb/details/config.h found in D:\odb\libodb-2.3.0. Have you actually built libodb for MinGW?

'should have actually read the INSTALL file... ;-)
I didn't notice that I had to build : libodb, sqlite, boost, libodb-sqlite, libodb-boost

And that's ok (compiled) !
Now trying to link... :-)

Thanks a lot.

Rodolphe

From boris at codesynthesis.com  Thu Dec 12 13:02:13 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Thu Dec 12 13:04:44 2013
Subject: [odb-users] Assertion fails in mysql/statement.cxx
In-Reply-To: <CAK-De8f8BKg=CWuUXmz2asSti8NYqRC1=YdcppqWkw3bHA1sUQ@mail.gmail.com>
References: <CAK-De8fsYqBoasuCeMiMf=UdYh+yZVH5Djh8Pz0JEj8xe9mCVQ@mail.gmail.com>
	<boris.20131211174107@codesynthesis.com>
	<CAK-De8cnVh328pRFZURajzPEu8+Hog8Sv6nc2XjkDF1NMjTWCQ@mail.gmail.com>
	<CAK-De8fjwiLbc=qidyP1MrUXGDreTysKKJxEW_int2mx9tEX9g@mail.gmail.com>
	<boris.20131212101343@codesynthesis.com>
	<CAK-De8f8BKg=CWuUXmz2asSti8NYqRC1=YdcppqWkw3bHA1sUQ@mail.gmail.com>
Message-ID: <boris.20131212195428@codesynthesis.com>

Hi Scott,

Scott Deerwester <scott.deerwester@gmail.com> writes:

> What do I do if I need to get an ordered list of the ChildClass ID's
> that are associated with a ParentClass object? That was the motivation
> for using a QMap. I was using a QMap iterator, but QSet doesn't do that.

With std::set you can provide a custom comparator to order the
objects using IDs rather than pointers (second template argument
to std::set). This way iteration over set will implicitly give you
an ordered list of IDs.

Unfortunately, QSet doesn't seem to provide this functionality and
I am not sure if any other Qt container supported by ODB does.

Boris

From alexis.mestag at epitech.eu  Sun Dec 15 14:40:04 2013
From: alexis.mestag at epitech.eu (Alexis Mestag)
Date: Sun Dec 15 14:45:11 2013
Subject: [odb-users] Using Custom Container (with template) makes odb crash
Message-ID: <loom.20131215T185036-956@post.gmane.org>

Hello everyone,

I use ODB for a school C++ project. I need relationships for my persistent
objects, for example, a list of pointers.
ODB recognizes relationships made with list of pointers.
The problem is, in my class, I want to store a POINTER to a list of pointers
to make the relationship, not the list directly.
So far, in order to do that, I created a wrapper to a list, and implemented
the templated odb::access::container_traits class to make ODB recognize my
custom container (as it's explained in the documentation).

For example, I need that :

// Person.hh
class Person
{
  friend class odb::access;

private:
  // ODB will not recognize this since it's a pointer to a list, not
directly the list.
  std::list<Person *>  *_friends;
};

#ifdef ODB_COMPILER
#  pragma db object(Person)
#endif

So I transform the class as followed :

// Person.hh

#include "Friends.hh"

class Person
{
  friend class odb::access;

private:
  Friends  *_friends;
};

#ifdef ODB_COMPILER
#  pragma db object(Person)
#endif

I add this class :

// Friends.hh

class Friends
{
private:
  std::list<Person *>  *_elts;
};

I implement the class odb::access::container_traits<Friends *> in a
Friends-traits.hxx file, create the Friends.options file and add it in the
parameters of the compilation command.
That works well.
The problem of this solution is that I need to recreate a wrapper to a list,
reimplement the odb::access::container_traits<Whatever *> class, recreate a
Whatever.options file and also add it in the command each time I make a
relationship.

To solve this problem, I wanted to create a templated class which would wrap
a list, as followed :

#ifndef			__LISTCONTAINER_HH__
# define		__LISTCONTAINER_HH__

# include		<list>

template<typename T>
class			ListContainer
{
public:
  typedef std::list<T>				ListSubType;
  typedef typename ListSubType::size_type	IndexType;

private:
  ListSubType		*_elts;

public:
  ListContainer() : _elts(new ListSubType) {}
  virtual ~ListContainer() {
    delete _elts;
  }
};

#endif

And I implemented the odb::access::container_traits<ListContainer<T> *>
class (big file) :

#ifndef			__LISTCONTAINER_TRAITS_HXX__
# define		__LISTCONTAINER_TRAITS_HXX__

# include		<odb/pre.hxx>
# include		<odb/container-traits.hxx>
# include		"ListContainer.hpp"

namespace		odb
{
  template<typename T>
  class			access::container_traits<ListContainer<T> *>
  {
  public:
    static container_kind const				kind = ck_ordered;
    static bool const					smart = false;

    typedef ListContainer<T> *				container_type;

    typedef T						value_type;
    typedef typename ListContainer<T>::IndexType	index_type;

    typedef ordered_functions<index_type, value_type>	functions;

  public:
    static void		persist(container_type const &c, functions const &f) {
      index_type	i = 0;

      for (auto it = c->_elts->cbegin() ; it != c->_elts->cend() ; ++it) {
	f.insert(i++, *it);
      }
    }

    static void		load(container_type &c, bool more, functions const &f) {
      c->_elts->clear();

      while (more) {
	index_type	dummy = 0;

	c->_elts->push_back(value_type());
	more = f.insert(dummy, c->_elts->back());
      }
    }

    static void		update(container_type const &c, functions const &f) {
      index_type	i = 0;

      f.delete_();
      for (auto it = c->_elts->cbegin() ; it != c->_elts->cend() ; ++it) {
	f.insert(i++, *it);
      }
    }

    static void		erase(functions const &f) {
      f.delete_();
    }
  };
}

# include		<odb/post.hxx>

#endif

The ListContainer.options file :

--odb-epilogue '#include "ListContainer-traits.hxx"'
--hxx-prologue '#include "ListContainer-traits.hxx"'

In theory, theses classes would allow me to do this :

#ifndef                 __PERSON_HH__
# define                __PERSON_HH__

# include               "Persistent.hh"
# include               "Nameable.hh"
# include               "ListContainer.hpp"

class                   Person : public Persistent, public Nameable
{
  friend class          odb::access;

private:
  unsigned char                 _age;
  ListContainer<Person *>       *_friends;

private:
  Person();

public:
  Person(std::string const &name, unsigned char const age = 0);
  Person(Person const &rhs);

  virtual ~Person();

  Person                &operator=(Person const &rhs);

  unsigned char         getAge() const;
  void                  setAge(unsigned char const age);
};

# ifdef ODB_COMPILER
#  pragma db object(Person)
# endif

#endif

I don't know if I did something wrong but when I try to compile this with
the command :
odb  -d mysql --generate-schema-only --std c++11 --at-once --options-file
ListContainer.options --input-name schema.sql Persistent.hh Nameable.hh
Person.hh

ODB crashes (Segmentation Fault), that's what it prints :
*** WARNING *** there are active plugins, do not report this as a bug unless
you can reproduce it without enabling any plugins.
Event                            | Plugins
PLUGIN_START_UNIT                | odb
PLUGIN_PRAGMAS                   | odb
PLUGIN_OVERRIDE_GATE             | odb
In file included from /usr/include/c++/4.8.2/bits/stl_map.h:63:0,
                 from /usr/include/c++/4.8.2/map:61,
                 from /usr/include/odb/std-map-traits.hxx:10,
                 from /usr/include/odb/container-traits.hxx:207,
                 from ListContainer-traits.hxx:15,
                 from <odb-epilogue-1>:1:
/usr/include/c++/4.8.2/tuple: In member function ?void
std::tuple<>::swap(std::tuple<>&)?:
/usr/include/c++/4.8.2/tuple:515:12: internal compiler error: Segmentation fault
       void swap(tuple&) noexcept { /* no-op */ }
            ^
Please submit a full bug report,
with preprocessed source if appropriate.
See <https://bugs.archlinux.org/> for instructions.

Here's my problem. Note that odb doesn't crashes when I comment the line
'ListContainer<Person *> *_friends;' in the Person class.
I hope I made my problem clear, if not, feel free to ask more details.

So, is it possible to make a templated custom container with ODB ?
If it is, what did I miss ?

Thanks in advance.

Alexis MESTAG

From boris at codesynthesis.com  Mon Dec 16 03:28:17 2013
From: boris at codesynthesis.com (Boris Kolpackov)
Date: Mon Dec 16 03:30:42 2013
Subject: [odb-users] Using Custom Container (with template) makes odb crash
In-Reply-To: <loom.20131215T185036-956@post.gmane.org>
References: <loom.20131215T185036-956@post.gmane.org>
Message-ID: <boris.20131216101452@codesynthesis.com>

Hi Alexis,

Alexis Mestag <alexis.mestag@epitech.eu> writes:

> class Person
> {
>   ...
>
>   Friends  *_friends;
> };
> 
> class Friends
> {
>   std::list<Person *>  *_elts;
> };

Your wrapper approach is not exactly equivalent to the original version.
If we "unwrap" it, we will get std::list<Person*>** instead of 
std::list<Person*>*. To me, it seems the equivalent will be:

class Person
{
  ...

  Friends _friends;
};


> I implement the class odb::access::container_traits<Friends *>

The problem is, container_traits should be specialized for a container.
You are still specializing it for a pointer to a container. This won't
work.

Besides containers, ODB also has a notion of wrappers. That is, something
that wraps or holds the underlying value. I think this will be more inline
with what you are trying to do. For example, you can do something like this
out of the box:

class Person
{
  ...

  std::unique_ptr<std::list<Person*>> _friends;
};

Because ODB includes a specialization of odb::warpper_traits for 
std::unique_ptr. While there is no default specialization for the raw
pointer, you can add one yourself. In fact, the odb/wrapper-traits.hxx
file on libodb contains a commented-out implementation that has the
following comment:

// Sample specialization for raw pointers. It is not enabled by default
// since it makes many assumptions that may not always hold true (such
// as that instances are allocated with new and freed with delete).
// This makes it too dangerous to be enabled unconditionally. If you
// need this functionality, you can copy the below code into your
// application. Also consider changing it to only specialize for
// specific types instead of for any pointer (it will almost always
// do the wrong thing for char*).

Boris