Hello Tom and Horiguchi-san,
On Tue, Nov 5, 2019 at 1:59 PM Kyotaro Horiguchi
<horikyota.ntt@gmail.com> wrote:
> At Mon, 04 Nov 2019 12:53:48 -0500, Tom Lane <tgl@sss.pgh.pa.us> wrote in
> > I do concur with creating a macro that encapsulates a correct version
> > of this test, maybe like
> >
> > #define DOUBLE_FITS_IN_INT64(num) \
> > ((num) >= (double) PG_INT64_MIN && \
> > (num) < -((double) PG_INT64_MIN))
Thank you for your comments. The proposed macro "DOUBLE_FITS_IN_INT64"
is a good and simple way to check the overflow. According to that, I
revised the patch, which includes regression tests.
In the patch, I additionally modified other occurrences as follows.
=========
+#define FLOAT8_FITS_IN_INT32(num) \
+ ((num) >= (float8) PG_INT32_MIN && (num) < -((float8) PG_INT32_MIN))
=========
- if (unlikely(num < (float8) PG_INT32_MIN ||
- num >= -((float8) PG_INT32_MIN) ||
- isnan(num)))
+ /* Range check */
+ if (unlikely(!FLOAT8_FITS_IN_INT32(num)))
=========
The added macro FLOAT8_FITS_IN_INT32() does not check NaN explicitly,
but it sufficiently handles the case.
Best regards,
Yuya Watari
NTT Software Innovation Center
watari.yuya@gmail.com
