Re: Virtual generated columns

On 12.11.24 09:49, jian he wrote:
>> On Wed, Nov 6, 2024 at 12:17 AM Peter Eisentraut <peter@eisentraut.org> wrote:
> RelationBuildPartitionKey
> if (!isnull)
> {
>          char       *exprString;
>          Node       *expr;
>          exprString = TextDatumGetCString(datum);
>          expr = stringToNode(exprString);
>          pfree(exprString);
>          expr = expand_generated_columns_in_expr(expr, relation);
> }
> no need expand_generated_columns_in_expr?
> in ComputePartitionAttrs, we already forbidden generated columns to be
> part of the partition key.

True.  I have removed this extra code in v9.

> check_modified_virtual_generated, we can replace fastgetattr to
> heap_attisnull? like:
>              // bool        isnull;
>              // fastgetattr(tuple, i + 1, tupdesc, &isnull);
>              // if (!isnull)
>              //     ereport(ERROR,
>              //             (errcode(ERRCODE_E_R_I_E_TRIGGER_PROTOCOL_VIOLATED),
>              //              errmsg("trigger modified virtual generated
> column value")));
>              if (!heap_attisnull(tuple, i+1, tupdesc))
>                  ereport(ERROR,
>                          (errcode(ERRCODE_E_R_I_E_TRIGGER_PROTOCOL_VIOLATED),
>                           errmsg("trigger modified virtual generated
> column value")));

I don't know.  fastgetattr() is supposed to be "fast". ;-)  It's all 
inline functions, so maybe that is actually correct.  I don't have a 
strong opinion either way.