Re: [PATCH] Supporting +-Infinity values by to_timestamp(float8)

On 2016-03-15, David Steele <david(at)pgmasters(dot)net> wrote:
> On 3/4/16 2:56 PM, Vitaly Burovoy wrote:
>> On 3/4/16, Anastasia Lubennikova <a(dot)lubennikova(at)postgrespro(dot)ru> wrote:
>>
>>> I think that you should update documentation. At least description of
>>> epoch on this page:
>>> http://www.postgresql.org/docs/devel/static/functions-datetime.html
>>
>> Thank you very much for pointing where it is located (I saw only
>> "to_timestamp(TEXT, TEXT)").
>> I'll think how to update it.
>
> Vitaly, have you decided how to update this yet?

Yes, there are three versions:
* remove mentioning how to get timestamptz from UNIX stamp;
* leave a note how to get timestamptz and add a note that such
encapsulation existed prior to 9.6;
* replace to the proposed current behavior (without interval).

I decided to implement the third case (but a result there has a time
zone which can be different, so the result can be not exactly same for
a concrete user). If a committer decides to do somehow else, he is
free to choose one from the list above or to do something else.

>>> 3. (nitpicking) I don't sure about "4STAMPS" suffix. "4" is nice
>>> abbreviation, but it seems slightly confusing to me.
>>
>> It doesn't matter for me what it is called, it is short enough and
>> reflects a type on which it is applied.
>> What would the best name be for it?
>
> Anastasia, any suggestions for a better name, or just leave it as is?
>
> I'm not in favor of the "4", either. I think I would prefer
> JULIAN_MAXYEAR_STAMP.

It turns out that Tom has changed almost one third of timestamp.h and
now the constant has a name "TIMESTAMP_END_JULIAN".

I've rebased the patch to the current master (5db5146) and changed it
according to the new timestamp.h.

Now it passes both versions (integer and float timestamps).
I deleted test for the upper boundary for integer timestamps, changed
a little comments.

I decided to delete hints about minimum and maximum allowed values
because no one type has such hint.

Please find attached a new version of the patch.

--
Best regards,
Vitaly Burovoy