Re: Re: Add generate_series(date,date) and generate_series(date,date,integer)

On 3/17/16 4:49 AM, Dean Rasheed wrote:

> On 16 March 2016 at 23:32, David Steele <david(at)pgmasters(dot)net> wrote:
>
>>
>> I think in this case it comes down to a committer's judgement so I have
>> marked this "ready for committer" and passed the buck on to Álvaro.
>
> So I was pretty much "meh" on this patch too, because I'm not
> convinced it actually saves much typing, if any.
>
> However, I now realise that it introduces a backwards-compatibility
> breakage. Today it is possible to type
>
> SELECT * FROM generate_series('01-01-2000'::date, '01-04-2000', '7 days');

It can also be broken as below and this is even scarier to me:

postgres=# SELECT * FROM generate_series('01-01-2000'::date,
'01-04-2000'::date, '7 days');
ERROR: invalid input syntax for integer: "7 days"
LINE 1: ...te_series('01-01-2000'::date, '01-04-2000'::date, '7 days');

And only works when:

postgres=# SELECT * FROM generate_series('01-01-2000'::date,
'01-04-2000'::date, '7 days'::interval);
generate_series
------------------------
2000-01-01 00:00:00+00
(1 row)

One might argue that string constants for dates in actual code might be
rare but I think it's safe to say that string constants for intervals
are pretty common. I also think it unlikely that they are all
explicitly cast.

I marked this "waiting for author" so Corey can respond. I actually
tested with the v3 patch since the v4 patch seems to be broken with git
apply or patch:

$ patch -p1 < ../other/generate_series_date.v4.diff
patching file doc/src/sgml/func.sgml
Hunk #1 succeeded at 14787 (offset 87 lines).
Hunk #2 succeeded at 14871 (offset 87 lines).
patching file src/backend/utils/adt/date.c
patch: **** malformed patch at line 163: diff --git
a/src/include/catalog/pg_proc.h b/src/include/catalog/pg_proc.h

$ git apply -3 ../other/generate_series_date.v4.diff
fatal: corrupt patch at line 163

--
-David
david(at)pgmasters(dot)net