Subject: Re: [boost] sqlpp11: SQL for C++
From: Roland Bock (rbock_at_[hidden])
Date: 2013-11-09 20:12:00
On 2013-11-10 00:12, Mateusz Loskot wrote:
> On 9 November 2013 22:58, Roland Bock <rbock_at_[hidden]> wrote:
>> On 2013-11-09 23:03, Michael Marcin wrote:
>>> On 11/9/2013 4:03 PM, Roland Bock wrote:
>>>> Please let me know your questions/thoughts/suggestions/rants.
>>>> Contributions welcome, of course :-)
>>>>
>>> Could you compare your library with SOCI?
>>>
>>>
>> I consider SOCI to be a string and position based approach, meaning that
>> you (as a library user) have to use strings to construct your queries
>> and positions to extract results.
> Yes, that's correct.
>
> In fact, SQL commands play an important role as part of DBMS access
> abstraction. SQL is a part of SOCI interface, by design.
>
>> sqlpp11 takes care of this responsibility for you and gives you rows
>> with appropriately named and typed member variables. It is much harder
>> to use those the wrong way without the compiler yelling at you.
>>
>> sqlpp11 assumes that you know your tables at compile time. Thus you can
>> declare types representing tables and columns with appropriate names and
>> types. You can then construct SQL queries and analyze the results with
>> the full armory of syntax and type checking that C++ and template meta
>> programming have to offer.
> Thank you for this clarification.
>
> I think sqlpp11 is an extremely interesting exercise.
:-)
>
> I have been considering to add non-string layer to SOCI directly based
One option would be to write a sqlpp11 connector library for SOCI,
similar to the existing ones for using MySQL/MariaDb and Sqlite3.
> on your idea, soon after I saw it some time ago in your initial experiments.
>
> Best regards,
You're referring to the discussion in 2010?
http://comments.gmane.org/gmane.comp.lib.boost.devel/208623
Cool, I wasn't aware of the impact :-)
sqlpp11 is much more mature, than what I had back then. Among other things
* auto allows for a much leaner perceived API
* table definitions are also much simpler. I am still using a code
generator for those (DDL->C++, which I'll add to the repository
soon), but the types are quite comprehensible even for a casual
user, I guess
* compile times are better
* sqlpp11 "understands" sub-selects, which were a nightmare in all my
previous attempts. Sub-selects can be used very naturally now, i.e.
as selected values, as input for functions like exists(), as
operands in where-conditions and as pseudo-tables in from(). The
documentation on that is still pretty thin, I guess, but there are
usage hints here and there:
https://github.com/rbock/sqlpp11/wiki/Select#sub-select
https://github.com/rbock/sqlpp11/wiki/Select#aliased-sub-select
* you can add columns to your select at runtime now, if required
* I would not call the library code simple, but it is certainly much
easier to understand and to extend than any of the earlier versions
Cheers,
Roland