PostgreSQL tsrange: is it correct for lower_inf('(-infinity,today)'::tsrange) to be false?

问题

In the course of writing a program that accepts tsrange literals from the user, which are then plugged into various SQL queries, today I was testing some tsranges to see how they are interpreted by PostgreSQL 9.3.5.

This one, in particular, behaves strangely: '(-infinity,today)'::tsrange

The lower_inf function says the lower bound is not infinite (!)

test=> SELECT lower_inf('(-infinity,today)'::tsrange);
 lower_inf
-----------
 f
(1 row)

Yet PostgreSQL reports that this tsrange contains a timestamp like '1000-01-01 BC' . . .

test=> SELECT '(-infinity,today)'::tsrange @> '1000-01-01 BC'::timestamp;
 ?column? 
----------
 t
(1 row)

Can anyone shed light on this?

回答1:

The confusion stems from two different meanings of "infinity" here.

1. timestamp types accept special values for infinity and -infinity.
2. Range types have a general concept for ranges without lower / upper bound. The functions to test for it are called lower_inf() and upper_inf(), but they are really testing for "no bound" in the range. Ranges with no upper / lower bound include the value infinity / -infinity for timestamp respectively.

The manual:

Also, some element types have a notion of "infinity", but that is just another value so far as the range type mechanisms are concerned. For example, in timestamp ranges, [today,] means the same thing as [today,). But [today,infinity] means something different from [today,infinity) — the latter excludes the special timestamp value infinity.

SQL Fiddle.

Maybe those functions should really be called something like lower_nobound() and upper_nobound() to avoid confusion ...

来源：https://stackoverflow.com/questions/27103608/postgresql-tsrange-is-it-correct-for-lower-inf-infinity-todaytsrange-to