Re: [xwiki-devs] [Discussion] Committing date changes on document XML files

[Sergiu Dumitriu]

git add --interactive

On 07/27/2016 03:59 AM, Alexandru Cotiuga wrote:
> I'm +1 for not committing date changes and it would be helpful to ignore
> them at some point, since I need all the time to edit files to commit in
> order to revert date changes.
> 
> Thanks,
> Alex
> 
> On Mon, Jul 25, 2016 at 10:40 PM, Thomas Mortagne > wrote:
> 
>> Le 25 juil. 2016 18:31, "Vincent Massol"  a écrit :
>>>
>>>
 On 25 Jul 2016, at 18:20, Eduard Moraru  wrote:

 Hi,

 On Mon, Jul 25, 2016 at 6:46 PM, Vincent Massol 
>> wrote:

> I think I’d be in favor of:
> * Have our xar:format remove the dates
> * Have xar:verify fail if the dates are in the XML (thus our quality
>> build
> will fail if that’s the case)
> * Have the import set the current date if no dates are defined (that’s
> probably the case already, would need to be checked)
>

 A side-effect of this is that, when you upgrade and extension, all the
 pages of the extension will be changed and set to the update date as
>> their
 last modification dates, right? (i.e. it affects both fresh installs
>> and
 upgrades)
>>>
>>> Isn’t this what happens now already, i.e. when an existing page is
>> imported the current date is set (unless it’s a backup pack)?
>>
>> Yes EM does not take into account plumbing stuff like date and version.
>>
>>>
>>> If the issue is about the diff, I guess we could have a diff that doesn’t
>> take into account the dates (or a better algorithm could be to not update a
>> page that only has the date metadata modified).
>>
>> It's already the case...
>>
>>>
 Thinking more about it, it could be problematic for all the pages of an
 extension that you upgrade to appear as being modified, even if nothing
 changed in them in that particular version.
>>>
>>> We should definitely not update pages with no changes.
>>>
 Another minor negative side-effect would also be searching or listing
 documents and sorting them by the last update time. Of course, this
>> would
 mostly affect admins or users with "show hidden documents" enabled.
>>>
>>> I we don’t update pages that haven’t been changed we won’t have this
>> problem, right?
>>>
 However, if you happen to also manage some content pages in your build
 (that are not supposed to be hidden), this becomes a nuissance.

 WDYT about the 2 problems? I guess we could always accept them and say
>> that
 installs/upgrades are relatively rare and that the impact is minimal
>> (and
 similar to an empty save in a document - something that can already be
 observed in practice in a document's history - so we don`t introduce
 anything new).

 Thanks,
 Eduard

 P.S.: Here`s an existing issue more or less related to this topic
 http://jira.xwiki.org/browse/XWIKI-7058. Caty reminded me about it.
>>>
>>> And http://jira.xwiki.org/browse/XWIKI-11764
>>>
>>> Thanks
>>> -Vincent
>>>
 * Have AS and Watchlist exclude import / new wiki (already the case)
>
> Thanks
> -Vincent
>
>> On 25 Jul 2016, at 14:08, Eduard Moraru 
>> wrote:
>>
>> Hi, devs,
>>
>> This interesting discussion [1] came up recently on a github commit
>> that
>> lead us to realise that a practice which we have been doing since
>> forever
>> is not documented in our best practices guides and that we also seem
>> to
>> lack consensus on it.
>>
>> It`s about the practice of skipping date field changes from document
>> XML
>> pages when committing them to source control. This includes doc date
>> and contentUpdateDate
>> fields, but also attachment dates.
>>
>> You can see some arguments on the discussion[1], but I also wanted to
>> mention that this practice goes in line with what we do for document
>> versions (which is handled by the xar:format maven plugin goal which
>> we
>> execute every time, before committing XML pages). If we are to update
>> doc
>> dates, then we should also increment doc versions, otherwise it does
>> not
>> make any sense.
>>
>> The idea was, AFAIR, that XWIki`s code pages should not generate any
>> updates in the user`s wiki content, in any way, and that and update
>> of
> the
>> code of a "system"/XWiki page should not show up as an update of *the
>> user's content*, since it would otherwise confuse him.
>>
>> What we are currently missing from xar:format is exactly this: the
>> reset
> of
>> XML page dates to have a clearer and more consistent date for XWiki`s
> code
>> pages.
>>
>> Your input is appreciated and the result of this discussion would be
>> the
>> update of our Development Practices [2] and Application Development
>> Best
>> Practices [3] pages.
>>
>> Thanks,
>> Eduard
>>
>> 

Re: [xwiki-devs] [Discussion] Committing date changes on document XML files

[Alexandru Cotiuga]

I'm +1 for not committing date changes and it would be helpful to ignore
them at some point, since I need all the time to edit files to commit in
order to revert date changes.

Thanks,
Alex

On Mon, Jul 25, 2016 at 10:40 PM, Thomas Mortagne  wrote:

> Le 25 juil. 2016 18:31, "Vincent Massol"  a écrit :
> >
> >
> > > On 25 Jul 2016, at 18:20, Eduard Moraru  wrote:
> > >
> > > Hi,
> > >
> > > On Mon, Jul 25, 2016 at 6:46 PM, Vincent Massol 
> wrote:
> > >
> > >> I think I’d be in favor of:
> > >> * Have our xar:format remove the dates
> > >> * Have xar:verify fail if the dates are in the XML (thus our quality
> build
> > >> will fail if that’s the case)
> > >> * Have the import set the current date if no dates are defined (that’s
> > >> probably the case already, would need to be checked)
> > >>
> > >
> > > A side-effect of this is that, when you upgrade and extension, all the
> > > pages of the extension will be changed and set to the update date as
> their
> > > last modification dates, right? (i.e. it affects both fresh installs
> and
> > > upgrades)
> >
> > Isn’t this what happens now already, i.e. when an existing page is
> imported the current date is set (unless it’s a backup pack)?
>
> Yes EM does not take into account plumbing stuff like date and version.
>
> >
> > If the issue is about the diff, I guess we could have a diff that doesn’t
> take into account the dates (or a better algorithm could be to not update a
> page that only has the date metadata modified).
>
> It's already the case...
>
> >
> > > Thinking more about it, it could be problematic for all the pages of an
> > > extension that you upgrade to appear as being modified, even if nothing
> > > changed in them in that particular version.
> >
> > We should definitely not update pages with no changes.
> >
> > > Another minor negative side-effect would also be searching or listing
> > > documents and sorting them by the last update time. Of course, this
> would
> > > mostly affect admins or users with "show hidden documents" enabled.
> >
> > I we don’t update pages that haven’t been changed we won’t have this
> problem, right?
> >
> > > However, if you happen to also manage some content pages in your build
> > > (that are not supposed to be hidden), this becomes a nuissance.
> > >
> > > WDYT about the 2 problems? I guess we could always accept them and say
> that
> > > installs/upgrades are relatively rare and that the impact is minimal
> (and
> > > similar to an empty save in a document - something that can already be
> > > observed in practice in a document's history - so we don`t introduce
> > > anything new).
> > >
> > > Thanks,
> > > Eduard
> > >
> > > P.S.: Here`s an existing issue more or less related to this topic
> > > http://jira.xwiki.org/browse/XWIKI-7058. Caty reminded me about it.
> >
> > And http://jira.xwiki.org/browse/XWIKI-11764
> >
> > Thanks
> > -Vincent
> >
> > > * Have AS and Watchlist exclude import / new wiki (already the case)
> > >>
> > >> Thanks
> > >> -Vincent
> > >>
> > >>> On 25 Jul 2016, at 14:08, Eduard Moraru 
> wrote:
> > >>>
> > >>> Hi, devs,
> > >>>
> > >>> This interesting discussion [1] came up recently on a github commit
> that
> > >>> lead us to realise that a practice which we have been doing since
> forever
> > >>> is not documented in our best practices guides and that we also seem
> to
> > >>> lack consensus on it.
> > >>>
> > >>> It`s about the practice of skipping date field changes from document
> XML
> > >>> pages when committing them to source control. This includes doc date
> > >>> and contentUpdateDate
> > >>> fields, but also attachment dates.
> > >>>
> > >>> You can see some arguments on the discussion[1], but I also wanted to
> > >>> mention that this practice goes in line with what we do for document
> > >>> versions (which is handled by the xar:format maven plugin goal which
> we
> > >>> execute every time, before committing XML pages). If we are to update
> doc
> > >>> dates, then we should also increment doc versions, otherwise it does
> not
> > >>> make any sense.
> > >>>
> > >>> The idea was, AFAIR, that XWIki`s code pages should not generate any
> > >>> updates in the user`s wiki content, in any way, and that and update
> of
> > >> the
> > >>> code of a "system"/XWiki page should not show up as an update of *the
> > >>> user's content*, since it would otherwise confuse him.
> > >>>
> > >>> What we are currently missing from xar:format is exactly this: the
> reset
> > >> of
> > >>> XML page dates to have a clearer and more consistent date for XWiki`s
> > >> code
> > >>> pages.
> > >>>
> > >>> Your input is appreciated and the result of this discussion would be
> the
> > >>> update of our Development Practices [2] and Application Development
> Best
> > >>> Practices [3] pages.
> > >>>
> > >>> Thanks,
> > >>> Eduard
> > >>>
> > >>> --
> > >>> [1]
> > >>>
> > >>
>
>