maria-developers team mailing list archive

[Sergei Golubchik <serg@xxxxxxxxxxxx>]

Hi, Sanja!

few preliminary comments:

On Dec 19, sanja@xxxxxxxxxxxx wrote:
> At file:///home/bell/maria/bzr/work-maria-10.0-cassandra/
> ------------------------------------------------------------
> revno: 3480
> revision-id: sanja@xxxxxxxxxxxx-20121219200839-oyxvfz7im9p7wqpw
> parent: sanja@xxxxxxxxxxxxxxxx-20121119121604-5h5tu0zn11em0sb3
> committer: sanja@xxxxxxxxxxxx
> branch nick: work-maria-10.0-cassandra
> timestamp: Wed 2012-12-19 22:08:39 +0200
> message:
>   Post review changes in the interface part 1.

> === modified file 'mysys/ma_dyncol.c'
> --- a/mysys/ma_dyncol.c	2012-09-28 12:27:16 +0000
> +++ b/mysys/ma_dyncol.c	2012-12-19 20:08:39 +0000
> @@ -26,6 +26,39 @@
>     SUCH DAMAGE.
>  */
>  
> +/*
> + Numeric format:
> + ===============
> +  * Fixed header part
> +    1 byte flags:
> +      0,1 bits - <offset size> - 1
> +      2-7 bits - 0
> +    2 bytes column counter
> +  * Columns directory sorted by column number, each entry contains of:
> +    2 bytes column number
> +    <offset size> bytes (1-4) combined offset from beginning of
> +      the data segment + 3 bit type
> +  * Data of above columns size of data and length depend on type
> +
> + Columns with names:
> + ===================
> +  * Fixed header part
> +    1 byte flags:
> +      0,1 bits - <offset size> - 2
> +      2 bit    - 1 (mens format with names)

s/mens/means/g

> +      3,4 bits - 01 (mens <names offset size> - 1, now 2 is only supported size)

eh. I would simply assume that if the bit 2 is 1, this also means
offset-2. Like, you have "old format" as above, and "new format" with
names, support for recursion (4 bits per type, offset-2), etc.

you only have 6 bits here, let's use them sparingly and keep things
simple.

> +      5-7 bits - 0
> +    2 bytes column counter
> +  * Variable header part
> +    <names offset size> (2) bytes size of stored names pool
> +  * Column directory sorted by names, each consists of
> +    <names offset size> (2) bytes offset of name
> +    <offset size> bytes (1-4)bytes combined offset from beginning of
> +      the data segment + 4 bit type
> +  * Names stored one after another
> +  * Data of above columns size of data and length depend on type
> +*/
> +
>  #include "mysys_priv.h"
>  #include <m_string.h>
>  #include <ma_dyncol.h>
> === modified file 'storage/cassandra/ha_cassandra.cc'
> --- a/storage/cassandra/ha_cassandra.cc	2012-11-19 12:16:04 +0000
> +++ b/storage/cassandra/ha_cassandra.cc	2012-12-19 20:08:39 +0000
> @@ -893,9 +893,37 @@ public:
>  /**
>    Converting dynamic columns types to/from casandra types
>  */
> +
> +
> +/**
> +  Check and initialize (if it is needed) string MEM_ROOT
> +*/
> +static void alloc_strings_memroot(MEM_ROOT *mem_root)
> +{
> +  if (mem_root->block_size == 0)

write

   if (alloc_root_inited(mem_root))

> +  {
> +    /*
> +      The mem_root used to allocate UUID (of length 36 + \0) so make
> +      appropriate allocated size
> +    */
> +    init_alloc_root(mem_root,
> +                    (36 + 1 + ALIGN_SIZE(sizeof(USED_MEM))) * 10 +
> +                    ALLOC_ROOT_MIN_BLOCK_SIZE,
> +                    (36 + 1 + ALIGN_SIZE(sizeof(USED_MEM))) * 10 +
> +                    ALLOC_ROOT_MIN_BLOCK_SIZE);
> +  }
> +}
> +
> +static void free_strings_memroot(MEM_ROOT *mem_root)
> +{
> +  if (mem_root->block_size != 0)

and here

> +    free_root(mem_root, MYF(0));
> +}
> +
> @@ -1730,6 +1759,7 @@ int ha_cassandra::read_cassandra_columns
>    int res= 0;
>    ulong total_name_len= 0;
>  
> +  strings_root.block_size= 0; // indicates uninitialized MEM_ROOT

Use

  clear_alloc_root(&strings_root);

>    /*
>      cassandra_to_mariadb() calls will use field->store(...) methods, which
>      require that the column is in the table->write_set

Regards,
Sergei