Re: [PATCH] pgbench: improve \sleep meta command

On 2021/03/09 0:54, Fujii Masao wrote:
> 
> 
> On 2021/03/08 23:10, Alvaro Herrera wrote:
>> On 2021-Mar-08, kuroda.hayato@fujitsu.com wrote:
>>
>>> Dear Fujii-san, Miyake-san
>>>
>>>> Isn't it better to accept even negative sleep time like currently pgbench does?
>>>> Otherwise we always need to check the variable is a positive integer
>>>> (for example, using \if command) when using it as the sleep time in \sleep
>>>> command. That seems inconvenient.
>>>
>>> Both of them are acceptable for me.
>>> But we should write down how it works when the negative value is input if we adopt.
>>
>> Not sleeping at all seems a good reaction (same as for zero, I guess.)
> 
> +1. BTW, IIUC currently \sleep works in that way.

Attached is the updated version of the patch.

Currently a variable whose value is a negative number is allowed to be
specified as a sleep time as follows. In this case \sleep command doesn't
sleep. The patch doesn't change this behavior at all.

     \set hoge -1
     \sleep :hoge s

Currently a variable whose value is a double is allowed to be specified as
a sleep time as follows. In this case the integer value that atoi() converts
the double number into is used as a sleep time. The patch also doesn't
change this behavior at all because I've not found any strong reason to
ban that usage.

     \set hoge 10 / 4.0
     \sleep :hoge s

Regards,

-- 
Fujii Masao
Advanced Computing Technology Center
Research and Development Headquarters
NTT DATA CORPORATION