Re: Enums again

Alvaro Herrera wrote:

>Andrew Dunstan wrote:
>
>  
>
>>In the catalog, pg_type would 
>>get a new column of type text[] that would hold the list of values, and 
>>typtype would have a new possible value of 'e' for enumeration. There 
>>might be other consequential changes too, but I think that would be most 
>>of it.
>>    
>>
>
>Huh, why not have the actual values in a separate catalog like
>pg_enumvalues or some such?
>  
>

Sure, could do that. I don't have strong feelings either way.

>  
>
>>The only functions that actually need to have any knowledge of 
>>the enumeration strings are the input/output functions and the to/from 
>>text casts. These would get the relevant info from fcinfo.flinfo ... and 
>>then looking up the type cache - not sure yet if an extra cache 
>>operation is needed.
>>    
>>
>
>It'd be interesting to measure the difference of having the cache vs.
>not having it.
>  
>

Possibly. I would expect it to make a noticeable difference.

>Thinking on how to pg_dump the whole thing is important too.
>
>  
>

Yes, that would certainly be part of the work. I should have mentioned 
that. It's not a showstopper, though - I see no reason in principal for 
it to be a difficulty.

cheers

andrew