Re: GSoC Week 8 - I_S/P_S storage engine

[Sergei Golubchik]

Hi, scut_tang!

On Jun 30, scut_tang wrote:
>  
> KEY ACCOMPLISHMENTS LAST WEEK
> =============================
> Almost finish infoschema_discover function, but meet some problems in debugging.
> It will be described in KEY CONCERNS section.
> 
> KEY TASKS THAT STALLED LAST WEEK
> ==============================
> May be I should perfect discover function. 
> 
> KEY CONCERNS
> ============
> 1.I want to know more about how to debug MySQL internal. Before this
>   project, I always used gdb to debug mysqld and it worked well.
>   But in this project, I meet some problems. For example, the source
>   code like:
...<skip>
>   Does the source code text file have some problems? Sergei, how do
>   you debug MySQL internal? 

I've seen that you solved that already. Great!
Yes, I always compile with -O0 for debugging.

> 2. What kinds of members in TABLE_SHARE must be initialized in
>    open_table? In my discover, I fill TABLE_SHARE members:
>    share->mem_root; share->field; share->fields; share->key_info;
>    share->keys; share->set_table_cache_key;
>    share->version; share->table_space; share->table_map_id;
>    share->LOCK_ha_data; share->null_fields; 
>    share->blob_fields; share->db_plugin; share->reclength;
>    share->rec_buff_length; 

you don't need to initialize manually mem_root and some other fields -
just call init_tmp_table_share(), it'll do it for you.

You need to set fields that describe the table, such as:

  Field **field;
  Field **found_next_number_field; // will be 0 for you
  Field *timestamp_field;          // same here
  KEY  *key_info;
  uint  *blob_field;
  uchar *default_values;
  LEX_STRING comment;
  CHARSET_INFO *table_charset;
  key_map keys_in_use;
  key_map keys_for_keyread;
  ha_rows min_rows, max_rows;
  ulong   avg_row_length;
  ulong   version, mysql_version;
  ulong   timestamp_offset;
  ulong   reclength;
  plugin_ref db_plugin;
  enum row_type row_type;
  enum ha_storage_media default_storage_media;
  char *tablespace;
  enum tmp_table_type tmp_table;
  enum ha_choice transactional;
  enum ha_choice page_checksum;
  uint null_bytes, last_null_bit_pos;
  uint fields;
  uint rec_buff_length;
  uint keys, key_parts;
  uint max_key_length, max_unique_length, total_key_length;
  uint uniques;
  uint null_fields;
  uint blob_fields;
  uint timestamp_field_offset;
  uint varchar_fields;
  uint db_create_options;
  uint db_options_in_use;
  uint primary_key;
  uint column_bitmap_size;
  uchar frm_version;
  bool null_field_first;
  bool db_low_byte_first;

But fear not! :)

Most of these fields can be set to the default values, as they don't
apply to the information_schema tables. Many others can be generated -
that is, the engine in discover() don't need to set them - but after
successful discovery MySQL will need to set these fields. They are
fields like 'name_hash' (hash of field names), 'varchar_fields' (number
of varchar fields), etc. The engine will need to set only 'field',
'key_info', 'default_values' and few others, everything else can be
calculated from them.

>    Because I can not still push my branch, I write these here. 

On your
https://answers.edge.launchpad.net/launchpad-code/+question/75673

there was a comment that you should try bzr 1.16.1, perhaps it'll
help...
 
Regards / Mit vielen Grüßen,
Sergei

-- 
   __  ___     ___ ____  __
  /  |/  /_ __/ __/ __ \/ /   Sergei Golubchik <serg@stripped>
 / /|_/ / // /\ \/ /_/ / /__  Principal Software Engineer/Server Architect
/_/  /_/\_, /___/\___\_\___/  Sun Microsystems GmbH, HRB München 161028
       <___/                  Sonnenallee 1, 85551 Kirchheim-Heimstetten
Geschäftsführer: Thomas Schroeder, Wolfgang Engels, Wolf Frenkel
Vorsitzender des Aufsichtsrates: Martin Häring