List:Summer of Code« Previous MessageNext Message »
From:scut_tang Date:June 23 2009 1:50pm
Subject:GSoC Week 7 - I_S/P_S storage engine
View as plain text  
Hi, I am late to post my week report, because I want to push my branch before post my
report.
I am sorry it failed after I tried many times and I am keeping in touch with launchpad. Or
can I post the code
to Sergei directly?

KEY ACCOMPLISHMENTS LAST WEEK
=============================
Because before your reply last night, I have no idea to do with dicover, and I take some
attention to another place, like how to
get tables' data. 
Know how INFORMATION_SCHEMA.ENGINES table to get data. I think it is easy to migrate to
new storage engine.

KEY TASKS THAT STALLED LAST WEEK
==============================
Funcition discover is not finished. It is hard to generate frm binary blob for me,
although I think it is best implementation.
In fact, I have the idea that ha_discover return TABLE_SHARE filled structure before, but
it will change ha_discover interface and
need to change NDB. So I drop the idea. And now it seems, that return TABLE_SHARE by
ha_discover is the fast way.

KEY CONCERNS
============


1. Now I realize the way P_S storage engine open table by pre-created frm files is not
friendly for user. If user does not compile P_S
  storage engine into MySQL server, the scripts mysql_install_db will be failed. 
2. I think to change discover() to return a TABLE_SHARE structure directly is a good idea.
I see ha_discover gets frm contents and writes
  the contents into frm file, and init_tmp_table_share initializes share for temporary
tables, which is different with alloc_table_share is init_tmp_tables does not put the
share into table cache, and it means threads do not share the TABLE_SHARE. Do tables
discoveried by
NDB are treated as temporary tables? And what about I_S tables? I think I_S tables should
be inserted into table_def_cache and be treated as normal tables.  Another question is
what does parameters char *key and uint key_length in function get_table_share? 
3. Do you think I am behind the progress? In my mind, the project goes like this:  finish
ha_discover and open -> how to get tables' data
-> index implementation -> other issues, like privileges.




TASKS IN THE UPCOMING WEEK
==========================
Finish discover with Sergei's suggestion.

Regards,
Robin.





Thread
GSoC Week 7 - I_S/P_S storage enginescut_tang23 Jun