Datetime issues
Locked
 
|
Hi,
I'm having a disagreement with the Date-Time library in Gemstone. When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. For example, if you had set a week's startday to sunday, the week that is returned will start on monday. If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... I traced the issue and found out that in Gemstone, the following is not true: DateAndTime now midnight weekDay = DateAndTime now weekDay Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. Or should the issue with #weekDay be fixed? cheers Johan |
|
|
Johan,
I'll have to look into this ... I did some Date/Time work over the summer, but I don't think I did much looking at DateAndTime. There are some fundamental implementation differences that make getting things working the same a bit difficult ... with that said, I assume that it is possible to fix... Dale On 11/17/2010 01:34 PM, Johan Brichau wrote: > Hi, > > I'm having a disagreement with the Date-Time library in Gemstone. > > When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. > For example, if you had set a week's startday to sunday, the week that is returned will start on monday. > If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... > > I traced the issue and found out that in Gemstone, the following is not true: > > DateAndTime now midnight weekDay = DateAndTime now weekDay > > Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. > > The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. > > Or should the issue with #weekDay be fixed? > > cheers > Johan |
|
|
In reply to this post by Johan Brichau-2
Johan,
Well, I'm finally taking a look at this (I usually claim that I am nine months behind, so maybe I am finally catching up?). I'm tracking this with issue 198 (http://code.google.com/p/glassdb/issues/detail?id=198)..and I am not able to reproduce your problem on a couple of fronts ... the method weekCay doesn't appear to be implemented in DateAndTime (in Pharo or GemStone) and if I replace weekDay with dayOfWeek, I don't have the issue with midnight ... I assume that there is a bug in here somewhere, so I will look a little deeper... I guess I'm curious if you are still seeing the problem and if so what version of GLASS you're using (I think the work over last summer did involve monkeying with #midnight) and then some help to reproduce the bug in my environment ... Dale On Nov 17, 2010, at 1:34 PM, Johan Brichau wrote: > Hi, > > I'm having a disagreement with the Date-Time library in Gemstone. > > When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. > For example, if you had set a week's startday to sunday, the week that is returned will start on monday. > If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... > > I traced the issue and found out that in Gemstone, the following is not true: > > DateAndTime now midnight weekDay = DateAndTime now weekDay > > Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. > > The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. > > Or should the issue with #weekDay be fixed? > > cheers > Johan |
|
|
Johan,
While I am unable to reproduce the problem it _is_ worth noting that Issue148 (which I just fixed) does involve an error in the base image, where DateAndTime>>dayOfWeek was not calculated using local time ... I haven't seen a difference in testing in my image, but I would imagine that at the right time of day relative to your offset from GMT that at least the dayOfWeek calculation would be wrong ... So, for the moment, I am going to assume that Issue 198 is the same as Issue 148 ... Dale On Mar 24, 2011, at 9:41 AM, Dale Henrichs wrote: > Johan, > > Well, I'm finally taking a look at this (I usually claim that I am nine months behind, so maybe I am finally catching up?). > > I'm tracking this with issue 198 (http://code.google.com/p/glassdb/issues/detail?id=198)..and I am not able to reproduce your problem on a couple of fronts ... > > the method weekCay doesn't appear to be implemented in DateAndTime (in Pharo or GemStone) and if I replace weekDay with dayOfWeek, I don't have the issue with midnight ... > > I assume that there is a bug in here somewhere, so I will look a little deeper... > > I guess I'm curious if you are still seeing the problem and if so what version of GLASS you're using (I think the work over last summer did involve monkeying with #midnight) and then some help to reproduce the bug in my environment ... > > Dale > > On Nov 17, 2010, at 1:34 PM, Johan Brichau wrote: > >> Hi, >> >> I'm having a disagreement with the Date-Time library in Gemstone. >> >> When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. >> For example, if you had set a week's startday to sunday, the week that is returned will start on monday. >> If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... >> >> I traced the issue and found out that in Gemstone, the following is not true: >> >> DateAndTime now midnight weekDay = DateAndTime now weekDay >> >> Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. >> >> The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. >> >> Or should the issue with #weekDay be fixed? >> >> cheers >> Johan > |
|
|
In reply to this post by Dale Henrichs
Hi Dale,
The #weekDay message was indeed wrong in my email and should be #dayOfWeek. So it becomes: DateAndTime now midnight dayOfWeek = DateAndTime now dayOfWeek If I'm evaluating the above expression in Pharo1.1.1, I'm getting true (as it should be). In GLASS 1.0-beta8.6, I'm getting false. The left-hand side of the comparison is 5 and the right-hand side is 6. The right-hand side is wrong (it's still friday morning and not saturday ;-) However, coming back to the originating problem (i.e. the fact that I saw that "Week current" was one day off in GS): I'm indeed not seeing that anymore in a fresh GLASS 1.0-beta8.6 I will investigate this a little deeper today because I just noticed that DateAndTime is now disregarding the timezone setting (for some reason). Thanks for looking into this, I did not have the chance to dive it into it any further yet. I'm getting back to you today. Johan On 24 Mar 2011, at 17:41, Dale Henrichs wrote: > Johan, > > Well, I'm finally taking a look at this (I usually claim that I am nine months behind, so maybe I am finally catching up?). > > I'm tracking this with issue 198 (http://code.google.com/p/glassdb/issues/detail?id=198)..and I am not able to reproduce your problem on a couple of fronts ... > > the method weekCay doesn't appear to be implemented in DateAndTime (in Pharo or GemStone) and if I replace weekDay with dayOfWeek, I don't have the issue with midnight ... > > I assume that there is a bug in here somewhere, so I will look a little deeper... > > I guess I'm curious if you are still seeing the problem and if so what version of GLASS you're using (I think the work over last summer did involve monkeying with #midnight) and then some help to reproduce the bug in my environment ... > > Dale > > On Nov 17, 2010, at 1:34 PM, Johan Brichau wrote: > >> Hi, >> >> I'm having a disagreement with the Date-Time library in Gemstone. >> >> When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. >> For example, if you had set a week's startday to sunday, the week that is returned will start on monday. >> If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... >> >> I traced the issue and found out that in Gemstone, the following is not true: >> >> DateAndTime now midnight weekDay = DateAndTime now weekDay >> >> Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. >> >> The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. >> >> Or should the issue with #weekDay be fixed? >> >> cheers >> Johan > |
|
|
In reply to this post by Dale Henrichs
Dale,
I confirm: issues 198 and 148 are identical. I tested the fix for #198 and it solves the problem. (I should have read all my emails before answering the previous one;-) thanks! Johan On 24 Mar 2011, at 18:10, Dale Henrichs wrote: > Johan, > > While I am unable to reproduce the problem it _is_ worth noting that Issue148 (which I just fixed) does involve an error in the base image, where DateAndTime>>dayOfWeek was not calculated using local time ... I haven't seen a difference in testing in my image, but I would imagine that at the right time of day relative to your offset from GMT that at least the dayOfWeek calculation would be wrong ... > > So, for the moment, I am going to assume that Issue 198 is the same as Issue 148 ... > > Dale > > On Mar 24, 2011, at 9:41 AM, Dale Henrichs wrote: > >> Johan, >> >> Well, I'm finally taking a look at this (I usually claim that I am nine months behind, so maybe I am finally catching up?). >> >> I'm tracking this with issue 198 (http://code.google.com/p/glassdb/issues/detail?id=198)..and I am not able to reproduce your problem on a couple of fronts ... >> >> the method weekCay doesn't appear to be implemented in DateAndTime (in Pharo or GemStone) and if I replace weekDay with dayOfWeek, I don't have the issue with midnight ... >> >> I assume that there is a bug in here somewhere, so I will look a little deeper... >> >> I guess I'm curious if you are still seeing the problem and if so what version of GLASS you're using (I think the work over last summer did involve monkeying with #midnight) and then some help to reproduce the bug in my environment ... >> >> Dale >> >> On Nov 17, 2010, at 1:34 PM, Johan Brichau wrote: >> >>> Hi, >>> >>> I'm having a disagreement with the Date-Time library in Gemstone. >>> >>> When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. >>> For example, if you had set a week's startday to sunday, the week that is returned will start on monday. >>> If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... >>> >>> I traced the issue and found out that in Gemstone, the following is not true: >>> >>> DateAndTime now midnight weekDay = DateAndTime now weekDay >>> >>> Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. >>> >>> The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. >>> >>> Or should the issue with #weekDay be fixed? >>> >>> cheers >>> Johan >> > |
|
|
haha...and I should have dug into the problem sooner:)
Dale On Mar 25, 2011, at 12:36 AM, Johan Brichau wrote: > Dale, > > I confirm: issues 198 and 148 are identical. > > I tested the fix for #198 and it solves the problem. > > (I should have read all my emails before answering the previous one;-) > > thanks! > Johan > > On 24 Mar 2011, at 18:10, Dale Henrichs wrote: > >> Johan, >> >> While I am unable to reproduce the problem it _is_ worth noting that Issue148 (which I just fixed) does involve an error in the base image, where DateAndTime>>dayOfWeek was not calculated using local time ... I haven't seen a difference in testing in my image, but I would imagine that at the right time of day relative to your offset from GMT that at least the dayOfWeek calculation would be wrong ... >> >> So, for the moment, I am going to assume that Issue 198 is the same as Issue 148 ... >> >> Dale >> >> On Mar 24, 2011, at 9:41 AM, Dale Henrichs wrote: >> >>> Johan, >>> >>> Well, I'm finally taking a look at this (I usually claim that I am nine months behind, so maybe I am finally catching up?). >>> >>> I'm tracking this with issue 198 (http://code.google.com/p/glassdb/issues/detail?id=198)..and I am not able to reproduce your problem on a couple of fronts ... >>> >>> the method weekCay doesn't appear to be implemented in DateAndTime (in Pharo or GemStone) and if I replace weekDay with dayOfWeek, I don't have the issue with midnight ... >>> >>> I assume that there is a bug in here somewhere, so I will look a little deeper... >>> >>> I guess I'm curious if you are still seeing the problem and if so what version of GLASS you're using (I think the work over last summer did involve monkeying with #midnight) and then some help to reproduce the bug in my environment ... >>> >>> Dale >>> >>> On Nov 17, 2010, at 1:34 PM, Johan Brichau wrote: >>> >>>> Hi, >>>> >>>> I'm having a disagreement with the Date-Time library in Gemstone. >>>> >>>> When you evaluate 'Week current', the returned timespan is off by one day, i.e. the week starts a day later than the Week>>startDay that you had set. >>>> For example, if you had set a week's startday to sunday, the week that is returned will start on monday. >>>> If you had set a week's startday to monday, the week that is returned will start on tuesday, etc... >>>> >>>> I traced the issue and found out that in Gemstone, the following is not true: >>>> >>>> DateAndTime now midnight weekDay = DateAndTime now weekDay >>>> >>>> Conversely, in Pharo, this is true and because Week>>current has the same implementation in both Pharo and Gemstone, the result is different in Gemstone. >>>> >>>> The easiest fix would be to leave out the use of #midnight in the implementation of Week>>current, but that would mean Gemstone needs a different version for the Chronos package than Pharo. >>>> >>>> Or should the issue with #weekDay be fixed? >>>> >>>> cheers >>>> Johan >>> >> > |
«
Return to GLASS
|
1 view|%1 views
| Free forum by Nabble | Edit this page |