Changed the way arguments are chained #29

[Killili]

Major Changes and General Overhaul of the everything.

-Refactored
exceptions into its own namespace
-removed rvalue references where they
are not needed
-made prepared statments possible
-made them
reusable
-removed some template magic

[Killili]

This is somewhat related to #29 but it got a bit out of hand so i made a new pull for this.
Its pretty much what i thought would be possible if we changed every lref of database_binder to a unique pointer so stuff like this is possible now:

try {

        { // scope somewhere else
          // creates a database file 'dbfile.db' if it does not exists.
            database db("dbfile.db");

            db << "drop table if exists test"; // still works as intended cause the database_binder dies at the semicolon
            db <<
                "create table if not exists test ("
                "   id integer primary key autoincrement not null,"
                "   val int"
                ");";

        } // destroy db

        database db("dbfile.db");

        std::cout << "auto&\n";
        auto& pps = db << "select ?"; // get a prepared parsed and ready statment

        int test = 4;
        pps << test; // set a bound var

        pps >> test; // execute and reset statment

        pps << 4; // bind a rvalue
        pps >> test; // and execute again

        auto& pps2 = db << "select 1,2,3,4,5"; // multiple extract test

        pps2 >> [](int a, int b, int c, int d, int e) {
            std::cout << "L " << a << b << c << d << e << "\n"; // still works as intended
        };

    } catch(exception& e) { // and here all statments die...
        cout << e.what() << endl;
    }

[aminroosta]

Changes are good 👍
But build is failing.

[Killili]

Ok it took my ~20h but now it looks good!
There is a lot of template magic in the lib so having more builds would be nice.

I can confirm VC2015 and gcc-5.2.1 and gcc-4.8.2(travis) can someone test clang?

Readme follows... ;)

[aminroosta]

@Killili i can confirm that all 5 tests ran successfully on clang 7.0.2
Great job 👍

[aminroosta]

instead of two functions overload .used() // getter and .used(bool) // setter functions

[aminroosta]

@Killili i have added some comments on the code, these are all just suggestions,
you are free to follow or keep names the way they are ;)

[aminroosta]

@Killili Please go one more iteration on readme file ;)
I recommend writing "sql queries" that work for examples.
Please add your database backup example for shared_connections section ( its a useful example).
Also please draft a new release after you've merged this,
I think there have been so many great changes.

[Killili]

Thx.

Those examples are valid SQL and work, maybe they are a bit short. But i like having examples that dont go through the whole "create table... yada yada" and get right to the point.
The backup routine is a bit long for an example but i will try.
I'm undecided on the getter/setter conventions, both have pros/cons.

[Killili]

As much as i hate the word "iteration" in the context of documentation its done ;)

As for the connection setter: I think that would add inconsistencies to prepared statements, as they have a copy of the connection after they are created. Consider:

database db("orig");
auto ps = db << "select ...";

database db2("new");
db.connection(db2.connection());
auto ps2 = db << "select ...";

ps++; // this will run on DB "orig"
ps2++; // this will run on DB "new"

I dont get what you mean with "draft" a release so i leave that "honor" to you ;)

[aminroosta]

@Killili I just wanted to thank you again 😉
Also thanks for database backup example, really useful example 🏆