Hi!
If anyone is up for doing the full solution it should work like this:
All key/pair params for CREATE TABLE should be collected up and pushed
into the engine during create_table. Stewart should, I believe, have
already implemented these in the proto for Table.
Engines should verify the definition of the params, and if they don't
support one should toss an error.
Verification in the parser is not required (nor is it a good idea...).
Cheers,
-Brian
On Mar 14, 2009, at 2:54 PM, Padraig O'Sullivan wrote:
On Sat, Mar 14, 2009 at 4:41 PM, Jay Pipes <[email protected]> wrote:
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Padraig O'Sullivan wrote:
Based on advice from Stewart, Jay and Monty on IRC, I added a new
flag
to the new storage engine API for resolving this bug. The behavior
is
still the same as what I outlined in earlier comments in this
thread but
the fix is a little different. I added a new flag to
drizzler/handlerton.h named HTON_BIT_DATA_INDEX_DIR and only 1
storage
engine adds this flag to its flags - MyISAM. Then, the check I
have in
drizzled/sql_table.cc is as follows:
/*
If the DATA_DIRECTORY or INDEX_DIRECTORY options are specified
and the
storage engine is not MyISAM, then we should return an error.
*/
if (create_info->data_file_name &&
!ha_check_storage_engine_flag(create_info->db_type,
HTON_BIT_DATA_INDEX_DIR))
{
my_error(ER_ILLEGAL_HA_CREATE_OPTION, MYF(0),
ha_resolve_storage_engine_name(create_info->db_type),
"DATA DIRECTORY");
goto err;
}
if (create_info->index_file_name &&
!ha_check_storage_engine_flag(create_info->db_type,
HTON_BIT_DATA_INDEX_DIR))
{
my_error(ER_ILLEGAL_HA_CREATE_OPTION, MYF(0),
ha_resolve_storage_engine_name(create_info->db_type),
"INDEX DIRECTORY");
goto err;
}
So I'm utilizing the storage engine API call
ha_check_storage_engine_flag(). I also modified the symlink test
case
based on the new behavior after implementing this fix. Does the
above
solution look like the right way to go about fixing this?
The branch with the fix is now linked to this bug so if you want
to have
a look at it feel free. If the above solution sounds reasonable,
then
I'll propose the branch for merging.
-Padraig
I like the ideas so far, but would recommend the following interface
change. Instead of ha_check_storage_engine_flag(), can we instead
just
do something like this on the handlerton API:
/**
* Returns whether or not the storage engine support a specific
* capability. Capability is represented by a hton_flag_bit.
*/
bool supports(enum hton_flag_bits capability);
That way, the interface would simply be:
if (! hton->supports(HTON_BIT_DATA_DIR)) {//do something..}
Isn't that less verbose and more natural for the caller?
Yep, I agree its less verbose. The reason I used the
ha_check_storage_engine_flag() call was because that was already
present and it is used in a few places. I implemented what you
suggested and it looks good. But now it seems like the
ha_check_storage_engine_flag() method has become obsolete since we
could just call hton->support() to check any of the storage engine
flags?
Also, I added 2 flags instead of 1 now because the ARCHIVE engine
supports the DATA DIRECTORY option as Brian stated but I'm assuming it
doesn't support the INDEX DIRECTORY option from what I've read about
the ARCHIVE storage engine. Thus, one flag indicates whether an engine
supports the DATA DIRECTORY option and the other flag indicates
whether an engine supports the INDEX DIRECTORY option.
-Padraig
Cheers,
Jay
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org
iEYEARECAAYFAkm8FoQACgkQ2upbWsB4UtGJggCeKZ3HOPJkMf+o2e/vZjNztG53
vzYAnRXEVb2qxwXDFDAj3gkuRVbOHdSG
=01h8
-----END PGP SIGNATURE-----
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
--
_______________________________________________________
Brian "Krow" Aker, brian at tangent.org
Seattle, Washington
http://krow.net/ <-- Me
http://tangent.org/ <-- Software
_______________________________________________________
You can't grep a dead tree.
_______________________________________________
Mailing list: https://launchpad.net/~drizzle-discuss
Post to : [email protected]
Unsubscribe : https://launchpad.net/~drizzle-discuss
More help : https://help.launchpad.net/ListHelp
