On Tue, Feb 14, 2017 at 8:59 AM, Jim Nasby <Jim.Nasby@bluetreble.com> wrote:
>> From my point of your it is strange and wrong expectation.
>> I am choosing "float4" type for a column just because it is enough to
>> represent range of data I have and I need to minimize size of record.
>
> In other words, you've decided to trade accuracy for performance...
>
>> But when I am calculating sum, I expect to receive more or less precise
>> result. Certainly I realize that even in case of using double it is
>
> ... but now you want to trade performance for accuracy? Why would you expect
> the database to magically come to that conclusion?
Well put. Although it's worth noting that we aren't 100% consistent
about this stuff: sum(smallint), sum(integer), and sum(bigint) all use
an output data type different from the input data type, but other
versions of sum() don't. To some extent all of these decisions are
just guesses about what users will find useful, and as this thread
shows, not everybody's going to agree. But I don't think our guesses
are flagrantly unreasonable or anything.
There's also nothing to prevent Konstantin or anybody else who doesn't
like the default behavior to create their own version of sum(float4)
and put it in a schema that's listed before pg_catalog in search_path.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company
