small bug on 3-digit years in 9.2-dev

Started by Marc Cousinalmost 14 years ago2 messageshackers
Jump to latest
#1Marc Cousin
cousinmarc@gmail.com

Hi,

While working on the "What's new in 9.2", I think I found a small bug:

SELECT to_date('519-07-02','YYY-MM-DD');
to_date
------------
0519-07-02
(1 row)

It comes, I think, from the year 519 case not being handled in the following
code. Patch attached

+	if (year < 70)
+		return year + 2000;
+	/* Force 70-99 into the 1900's */
+	else if (year >= 70 && year < 100)
+		return year + 1900;
+	/* Force 100-519 into the 2000's */
+	else if (year >= 100 && year < 519)
+		return year + 2000;
+	/* Force 520-999 into the 1000's */
+	else if (year >= 520 && year < 1000)
+		return year + 1000;
+	else
+		return year;

Regards

Attachments:

correct_dates.patchtext/x-patch; charset=UTF-8; name=correct_dates.patchDownload+1-1
#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Marc Cousin (#1)
Re: small bug on 3-digit years in 9.2-dev

Marc Cousin <cousinmarc@gmail.com> writes:

While working on the "What's new in 9.2", I think I found a small bug:

Yeah, that code certainly looks wrong, thanks for the report!

/* Force 100-519 into the 2000's */
-	else if (year >= 100 && year < 519)
+	else if (year >= 100 && year <= 519)

I think "< 520" would be more in keeping with the adjacent coding.

regards, tom lane