From: Date: April 15 2006 6:03pm
Subject: Re: getting table meta data (primary key, in this case)
List-Archive: http://lists.mysql.com/internals/33517
Message-Id: <20060415160310.GF6355@w3.org>
MIME-Version: 1.0
Content-Type: multipart/signed; micalg=pgp-sha1;
	protocol="application/pgp-signature"; boundary="V4b9U9vrdWczvw78"

--V4b9U9vrdWczvw78
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Mon, Apr 10, 2006 at 12:17:54PM +0200, Sergei Golubchik wrote:
> Hi!
>=20
> On Apr 06, Eric Prud'hommeaux wrote:
> > On Thu, Apr 06, 2006 at 10:24:32AM +0200, Sergei Golubchik wrote:
> > > Hi!
> > >=20
> > > On Apr 05, Eric Prud'hommeaux wrote:
> > > > On Mon, Apr 03, 2006 at 02:53:08AM -0400, Eric Prud'hommeaux wrote:
> > > > > On Mon, Apr 03, 2006 at 12:41:30AM -0400, SGreen@stripped wrote:
> > > > >=20
> > > > > The problem is that I can't express higher-order-logic (symbols as
> > > > > values) in relational calculus. That is, I can't express like
> > > > > (inventing a $var notation here):
> > > > >   SELECT Orders.$field FROM Orders WHERE $field=3D(
> > > > >     SELECT Column_name SHOW INDEXES FROM Orders WHERE Key_name=3D=
'PRIMARY')
> > > > > I don't know if SQL limits itself to relational calculus, so perh=
aps
> > > > > there is some magic to do this.
> > >=20
> > > Not really :(
> > > Not with one SQL query, at least.
> > > But unless you limit yourself to rewriting-into-SQL approach, you sho=
uld
> > > not really care what SQL limits are.
> >=20
> > true. this approach (finding the key and joining on that field) is
> > limited only by the expressivity of the MySQL query structure.
> > However, as far as I can see, I'd need to do something like:
> >       Item* l =3D new Item_field(lex->current_context(), NullS,
> >                                this_alias, primary_key_name);
> >       Item* r =3D new Item_field(lex->current_context(), NullS,
> >                                linked_alias, linked_foreign_key_name);
> >       chooser_compare_func_creator eq =3D &comp_eq_creator;
> >       Item* on =3D eq(0)->create(l, r);
> >       add_join_on(table_list, on);
> >
> > and late bind primary_key_name. I don't think that's possible, at
> > least with an Item_field. I'd make a Item_field_late_binding and add a
> > schema table to the mix, but I can't see how I'd get select to call
> > some bind method on it while executing.
>=20
> It's possible.
>=20
> First - one can write SELECT * FROM t1, t2, ...
> and * will be expanded to a full list of fields after all tables are
> opened.
>=20
> Then - in fact all Item_field's always use this "late binding".
> Before Item_field::fix_fields(), Item_field only knows its name,
> sometimes database name and table name, but not always. And nothing
> else. That's all that the parser can tell about the field.
> For example:
>=20
>   SELECT f1, t1.f2, db1.t2.f3 FROM t1, db1.t2, t3;
>=20
> and only after all tables are opened, MySQL calls
> Item_field::fix_fields() for all Item_field's, and it's the method where
> "binding" happens - a field finds what table and database it belongs to,
> finds all metadata about itself, data type, constrains, indexes, and so
> on.

worky (and quite easily)!!! Thank you SO much for the pointer!

[[
bool Item_variable::fix_fields(THD *thd, Item **reference) {
  if (primary_key)
    field_name =3D get_primary_key();
  return Item_field::fix_fields(thd, reference);
}

const char* sparqlFrob::get_primary_key(const char *table_name) {
  TABLE_LIST *entry =3D thd->main_lex.select_lex.context.table_list;
  while (entry) {
    if (!strcmp(table_name, entry->table_name)) {
      KEY *key =3D entry->table->key_info;
      for (unsigned i =3D 0; i < entry->table->s->keys; i++) { // from sql_=
show.cc::2905 get_schema_stat_record
	if (!strcmp(key[i].name, "PRIMARY")) {
	  assert(key[i].key_parts =3D=3D 1);
	  return key[i].key_part->field->field_name;
	}
      }
      assert(0); // couldn't find PRIMARY KEY
    }
    entry =3D entry->next_global;
  }
  assert(0); // couldn't find table
}
]]
It brings up 2 questions:

  Is there a way to kill a thread ('cause something went horribly
  wrong), without killing the whole server? What's the polite way to
  assert(0)?

  Is there a better way to find the table_list for a thread? I feel
  like thd->main_lex.select_lex.context.table_list should really
  include some inode numbers, the process id and maybe some astrology.
--=20
-eric

office: +81.466.49.1170 W3C, Keio Research Institute at SFC,
                        Shonan Fujisawa Campus, Keio University,
                        5322 Endo, Fujisawa, Kanagawa 252-8520
                        JAPAN
        +1.617.258.5741 NE43-344, MIT, Cambridge, MA 02144 USA
cell:   +81.90.6533.3882

(eric@stripped)
Feel free to forward this message to any list for any purpose other than
email address distribution.

--V4b9U9vrdWczvw78
Content-Type: application/pgp-signature; name="signature.asc"
Content-Description: Digital signature
Content-Disposition: inline

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (GNU/Linux)

iQEVAwUBREEZPpZX2p1ccTnpAQJMnAgAge/Ss0ElNulPDaQV7as5vxfzUBPrFV/l
5L7q4MAAlGxhwR/Ji/DtDXr2+q37HWNx30fxJU5ENrkS/NeYzEN3aGsXmJestNXK
XZxaajSVvwGQtl6Fj5jhO5hNgWtblnfB7LLi5RkXl59IMz4co17rAdI57R/26QQ3
2hVP8/d9qRfD6izc7PoXJqJ+jxk2FezFXIL39kj3uapbiibTz1MVcaFxx5A4m+Yo
TqBE3+baZaNNT5vvpZTt0c1/rrxdz/G2cflK40/gtfuugtbwihPQ9l6UrnfI/1Np
lSLviDlUSEpGrsKz2J1KQ+kp5O79zbl3MQ0VTGR8Q00OfXK9OhBd/A==
=rP6N
-----END PGP SIGNATURE-----

--V4b9U9vrdWczvw78--