Hello Roy,
On 05/27/2011 06:21 PM, Roy Lyseng wrote:
> Hi Guilhem, Gleb,
>
> A couple of comments on Guilhem's comments first.
>
>>> + This method:
>>> + a) allocates a select_send object (if no one pre-allocated available),
>>> + b) calculates and sends whole EXPLAIN data.
>>> +
>>> + @returns
>>> + @retval FALSE Ok
>>> + @retval TRUE Error
>>
>> GB39 nowadays the coding style allows true/false too, it's up to you.
>
> My preferred Doxygen documentation for return values is
>
> @return false if successful, true if error
>
> This is so common, and saves two lines without loss of information.
Done.
>
>>
>> GB75 this code above resembles what is done in execute_sqlcom_select() (if
>> lex->describe is true), which is:
>>
>> thd->send_explain_fields(result);
>> res= mysql_explain_union(thd, &thd->lex->unit, result);
>> ...
>> if (res)
>> result->abort_result_set();
>> else
>> result->send_eof();
>> Can we make execute_sqlcom_select() call explain_data_modification(), to
>> avoid duplication?
>> The creation of the warning (of EXPLAIN EXTENDED) can also move into
>> explain_data_modification().
>
> If this is done, explain_data_modification() must be renamed as a SELECT is not a
> data modification. My suggestion in that case would be explain_query_specification()
> (Query specification is the term used in the SQL standard for a SELECT block).
Done. (OTOH query printing is not done yet -- most likely will be in the some small future
WL. I've added a check inside explain_query_specification()).
>
> Thank,
> Roy
>
Thank you,
Gleb.
