Re: proposal - assign result of query to psql variable

Hello

2012/10/14 Tom Lane <tgl@sss.pgh.pa.us>:
> Pavel Stehule <pavel.stehule@gmail.com> writes:
>> [ gset_08.diff ]
>
> In the course of adding a new backslash command, this patch manages to
> touch:
>
> * the main loop's concept of what types of backslash commands exist
>   (PSQL_CMD_NOSEND ... what's the point of that, rather than making
>   this work the same as \g?)

This is necessary, because there is a new possible state - "query is
complete, but command is wrong" - so I cannot use \g implementation,
because there is no possible error in \g or ';'

> * SendQuery's concept of how to process command results (again, why
>   isn't this more like \g?)

it is similar - difference is only in work with result - \gset uses
StoreQueryResult instead PrintQueryResults, but other is same


> * ExecQueryUsingCursor's concept of how to process command results
>   (why? surely we don't need \gset to use a cursor)

It is my mistake - simply and correct way is not using cursor in this use case

> * the psql lexer (adding a whole bunch of stuff that probably doesn't
>   belong there)

I had to modify lexer - current lexer supports symbols separated by
space, but not symbols separated by comma. We don't would to use
variable evaluation in target variable list.

> * the core psql settings construct (to store something that is in
>   no way a persistent setting)

sorry, I don't understand to this issue

>
> Surely there is a less ugly and invasive way to do this.  The fact
> that the reviewer keeps finding bizarre bugs like "another backslash
> command on the same line doesn't work" seems to me to be a good
> indication that this is touching things it shouldn't.
>

I had too strong in checking and raising errors. Is relative simple to
modify patch to enable more backslash statements on same line

I'll send updated patch early

Regards

Pavel

>                         regards, tom lane