Re: alternative to PG_CATCH

On 12/13/18 7:26 AM, Kyotaro HORIGUCHI wrote:
> At Thu, 13 Dec 2018 11:33:39 +0100, Peter Eisentraut <peter.eisentraut@2ndquadrant.com> wrote in
<c170919d-c78b-3dac-5ff6-9bd12f7a38bc@2ndquadrant.com>
>>
>> I've played with a way to express this more compactly:
>>
>>     PG_TRY();
>>     {
>>         ... code that might throw ereport(ERROR) ...
>>     }
>>     PG_FINALLY({
>>         cleanup();
>>     });
>>
>> See attached patch for how this works out in practice.

+1

> Though I didn't look into individual change, this kind of
> refactoring looks good to me. But the syntax looks
> somewhat.. uh..
> 
> I'm not sure it is actually workable, but I suspect that what we
> need here is just a shortcut of 'PG_CATCH();{PG_RE_THROW();}'.
> Something like this:
> 
> #define PG_FINALLY()    \
>         } \
>         else \
>         { \
>             PG_exception_stack = save_exception_stack; \
>             error_context_stack = save_context_stack; \
>             PG_RE_THROW();        \
>         } \
>         PG_exception_stack = save_exception_stack;    \
>         error_context_stack = save_context_stack; \
>         {
> 
> Which can be used as:
> 
> PG_TRY();
> {
>     ... code that might throw ereport(ERROR) ...
> }
> PG_FINALLY();
> {
>     cleanup();
> }
> PG_TRY_END();

I like this syntax better.  We use something very similar in the
pgBackRest project:

TRY_BEGIN()
{
    <Do something that might throw an error>
}
CATCH(Error1)
{
    <Handle Error1>
}
CATCH(Error2)
{
    <Handle Error2>
}
CATCH_ANY()
{
    <Handle errors that are not Error1 or Error2>
}
FINALLY()
{
    <Cleanup code that runs in all cases>
}
TRY_END();

The syntax works out simpler if the FINALLY is part of the TRY block.

See attached for the implementation.

Regards,
-- 
-David
david@pgmasters.net