You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@harmony.apache.org by Jim Yu <ju...@gmail.com> on 2008/06/03 13:14:08 UTC
[classlib][luni] Complement the Missing TimeZone IDs
Hi all,
If we run a simple comparison of RI's time zone IDs with Harmony's, we will
find there are quite a few in the same category.
I suggest that we need to complement the missing TimeZone IDs. Two solutions
in my mind:
[1] We load the TimeZone IDs into available zones cache via ICU completely
and don't use the available zones initialized by TimeZones
any more. But for this solution, many testcase will fail since they based on
the behavior of those TimeZone instances initialized by
TimeZones. So we need to check each failure testcase and identify which
behavior is more reasonable, instances of TimeZones or
those of ICU?
[2] We just do a small trick. When there is a missing TimeZone ID
which exists in ICU, we add it into available zones cache in TimeZone.
So it will be found from now on. For this solution, almost all the testcase
will succeed since instances of TimeZones are still working.
I think [2] is a simple fix, although [1] is also reasonable since we can
totally delegate to ICU. Personally I'd go with [2].
--
Best Regards,
Jim, Jun Jie Yu
China Software Development Lab, IBM
Re: [classlib][luni] Complement the Missing TimeZone IDs
Posted by Jim Yu <ju...@gmail.com>.
Hello Alexei, all,
Thanks for your helpful comments. I implemented the second solution
and attached my patch to
the JIRA at https://issues.apache.org/jira/browse/HARMONY-5860. Is there
anybody to verify it?
2008/6/3 Alexei Fedotov <al...@gmail.com>:
> Hello Jim, all,
>
> It's nice to see that you are digging the timezones further. I believe
> the second approach is better because we have already noticed on the
> list that different sources may understand timezones differently, and
> establishing compatibility between RI and ICU is just a second
> priority compared to the proper functionality.
>
> The second approach is also nice because it may be evolutionary
> evolved into the first one by getting more and more timezones from
> ICU. This is even more nice, because people usually work with few
> timezones, so lazy initialization approach is better than initializing
> all of them.
>
> Thanks.
>
> On Tue, Jun 3, 2008 at 3:14 PM, Jim Yu <ju...@gmail.com> wrote:
> > Hi all,
> >
> > If we run a simple comparison of RI's time zone IDs with Harmony's, we
> will
> > find there are quite a few in the same category.
> > I suggest that we need to complement the missing TimeZone IDs. Two
> solutions
> > in my mind:
> >
> > [1] We load the TimeZone IDs into available zones cache via ICU
> completely
> > and don't use the available zones initialized by TimeZones
> > any more. But for this solution, many testcase will fail since they based
> on
> > the behavior of those TimeZone instances initialized by
> > TimeZones. So we need to check each failure testcase and identify which
> > behavior is more reasonable, instances of TimeZones or
> > those of ICU?
> >
> > [2] We just do a small trick. When there is a missing TimeZone ID
> > which exists in ICU, we add it into available zones cache in TimeZone.
> > So it will be found from now on. For this solution, almost all the
> testcase
> > will succeed since instances of TimeZones are still working.
> >
> > I think [2] is a simple fix, although [1] is also reasonable since we can
> > totally delegate to ICU. Personally I'd go with [2].
> >
> > --
> > Best Regards,
> > Jim, Jun Jie Yu
> >
> > China Software Development Lab, IBM
> >
>
>
>
> --
> With best regards,
> Alexei
>
--
Best Regards,
Jim, Jun Jie Yu
China Software Development Lab, IBM
Re: [classlib][luni] Complement the Missing TimeZone IDs
Posted by Alexei Fedotov <al...@gmail.com>.
Hello Jim, all,
It's nice to see that you are digging the timezones further. I believe
the second approach is better because we have already noticed on the
list that different sources may understand timezones differently, and
establishing compatibility between RI and ICU is just a second
priority compared to the proper functionality.
The second approach is also nice because it may be evolutionary
evolved into the first one by getting more and more timezones from
ICU. This is even more nice, because people usually work with few
timezones, so lazy initialization approach is better than initializing
all of them.
Thanks.
On Tue, Jun 3, 2008 at 3:14 PM, Jim Yu <ju...@gmail.com> wrote:
> Hi all,
>
> If we run a simple comparison of RI's time zone IDs with Harmony's, we will
> find there are quite a few in the same category.
> I suggest that we need to complement the missing TimeZone IDs. Two solutions
> in my mind:
>
> [1] We load the TimeZone IDs into available zones cache via ICU completely
> and don't use the available zones initialized by TimeZones
> any more. But for this solution, many testcase will fail since they based on
> the behavior of those TimeZone instances initialized by
> TimeZones. So we need to check each failure testcase and identify which
> behavior is more reasonable, instances of TimeZones or
> those of ICU?
>
> [2] We just do a small trick. When there is a missing TimeZone ID
> which exists in ICU, we add it into available zones cache in TimeZone.
> So it will be found from now on. For this solution, almost all the testcase
> will succeed since instances of TimeZones are still working.
>
> I think [2] is a simple fix, although [1] is also reasonable since we can
> totally delegate to ICU. Personally I'd go with [2].
>
> --
> Best Regards,
> Jim, Jun Jie Yu
>
> China Software Development Lab, IBM
>
--
With best regards,
Alexei
Re: [classlib][luni] Complement the Missing TimeZone IDs
Posted by Sean Qiu <se...@gmail.com>.
+1 lazy initialization.
2008/6/3 Jim Yu <ju...@gmail.com>:
> Hi all,
>
> If we run a simple comparison of RI's time zone IDs with Harmony's, we will
> find there are quite a few in the same category.
> I suggest that we need to complement the missing TimeZone IDs. Two solutions
> in my mind:
>
> [1] We load the TimeZone IDs into available zones cache via ICU completely
> and don't use the available zones initialized by TimeZones
> any more. But for this solution, many testcase will fail since they based on
> the behavior of those TimeZone instances initialized by
> TimeZones. So we need to check each failure testcase and identify which
> behavior is more reasonable, instances of TimeZones or
> those of ICU?
>
> [2] We just do a small trick. When there is a missing TimeZone ID
> which exists in ICU, we add it into available zones cache in TimeZone.
> So it will be found from now on. For this solution, almost all the testcase
> will succeed since instances of TimeZones are still working.
>
> I think [2] is a simple fix, although [1] is also reasonable since we can
> totally delegate to ICU. Personally I'd go with [2].
>
> --
> Best Regards,
> Jim, Jun Jie Yu
>
> China Software Development Lab, IBM
>
--
Best Regards
Sean, Xiao Xia Qiu
China Software Development Lab, IBM
Re: [classlib][luni] Complement the Missing TimeZone IDs
Posted by Alexei Zakharov <al...@gmail.com>.
If the search for the missing TimeZone is not very expensive I vote for [2] too.
Thanks,
Alexei
2008/6/3 Jim Yu <ju...@gmail.com>:
> Hi all,
>
> If we run a simple comparison of RI's time zone IDs with Harmony's, we will
> find there are quite a few in the same category.
> I suggest that we need to complement the missing TimeZone IDs. Two solutions
> in my mind:
>
> [1] We load the TimeZone IDs into available zones cache via ICU completely
> and don't use the available zones initialized by TimeZones
> any more. But for this solution, many testcase will fail since they based on
> the behavior of those TimeZone instances initialized by
> TimeZones. So we need to check each failure testcase and identify which
> behavior is more reasonable, instances of TimeZones or
> those of ICU?
>
> [2] We just do a small trick. When there is a missing TimeZone ID
> which exists in ICU, we add it into available zones cache in TimeZone.
> So it will be found from now on. For this solution, almost all the testcase
> will succeed since instances of TimeZones are still working.
>
> I think [2] is a simple fix, although [1] is also reasonable since we can
> totally delegate to ICU. Personally I'd go with [2].
>
> --
> Best Regards,
> Jim, Jun Jie Yu
>
> China Software Development Lab, IBM
>
Re: [classlib][luni] Complement the Missing TimeZone IDs
Posted by Alexey Petrenko <al...@gmail.com>.
+1 for 2
2008/6/3, Jim Yu <ju...@gmail.com>:
> Hi all,
>
> If we run a simple comparison of RI's time zone IDs with Harmony's, we will
> find there are quite a few in the same category.
> I suggest that we need to complement the missing TimeZone IDs. Two solutions
> in my mind:
>
> [1] We load the TimeZone IDs into available zones cache via ICU completely
> and don't use the available zones initialized by TimeZones
> any more. But for this solution, many testcase will fail since they based on
> the behavior of those TimeZone instances initialized by
> TimeZones. So we need to check each failure testcase and identify which
> behavior is more reasonable, instances of TimeZones or
> those of ICU?
>
> [2] We just do a small trick. When there is a missing TimeZone ID
> which exists in ICU, we add it into available zones cache in TimeZone.
> So it will be found from now on. For this solution, almost all the testcase
> will succeed since instances of TimeZones are still working.
>
> I think [2] is a simple fix, although [1] is also reasonable since we can
> totally delegate to ICU. Personally I'd go with [2].
>
>
> --
> Best Regards,
> Jim, Jun Jie Yu
>
> China Software Development Lab, IBM
>