[
https://issues.apache.org/jira/browse/DERBY-4752?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Knut Anders Hatlen updated DERBY-4752:
--------------------------------------
Attachment: CheapFormatterPerfTest.java
Thanks for providing the historical background for the CheapDateFormatter
class, Kathey.
I wrote a small performance test (see the attached CheapFormatterPerfTest.java)
to see the impact of the fix. On my laptop, running Java 6u21 with default
options, the old implementation managed to format about 280000 timestamps per
second. The new implementation only manages about 190000 timestamps per second,
which is slower, but I cannot imagine that it would be noticed in any realistic
scenario.
> CheapDateFormatter returns incorrect and invalid date strings
> -------------------------------------------------------------
>
> Key: DERBY-4752
> URL: https://issues.apache.org/jira/browse/DERBY-4752
> Project: Derby
> Issue Type: Bug
> Components: Services
> Affects Versions: 10.7.0.0
> Reporter: Knut Anders Hatlen
> Assignee: Knut Anders Hatlen
> Priority: Minor
> Attachments: CheapFormatterPerfTest.java, derby-4752-1a.diff,
> derby-4752-1b.diff, derby-4752-1c.diff
>
>
> CheapDateFormatter has multiple problems. These are the ones I'm aware of:
> 1) On the boundary between non-leap years and leap years it will return first
> day of thirteenth month in previous year (for instance, 2011-13-01 instead of
> 2012-01-01)
> 2) It treats all years divisible by four as leap years. Those divisible by
> 100 and not by 400 are not leap years. It attempts to adjust for that (see
> the snippet below) but it always ends up setting leapYear=true if (year%4)==0.
> // It's a leap year if divisible by 4, unless divisible by 100,
> // unless divisible by 400.
> if ((year % 4L) == 0) {
> if ((year % 100L) == 0) {
> if ((year % 400L) == 0) {
> leapYear = true;
> }
> }
> leapYear = true;
> }
> 3) More leap year trouble. To find out which year it is, it calculates the
> number of four year periods that have elapsed since 1970-01-01. A four year
> period is considered 365*3+366 days. Although most four year periods are of
> that length, some are shorter, so we'll get one day off starting from year
> 2100, two days off from year 2200, and so on.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
