21 Feb
2012
21 Feb
'12
7:09 p.m.
Hi devs,
I`ve updated the pull request and pushed a migration script [1], but I`d
like a bit of help on 2 issues that I am facing regarding it:
1. Since I have already updated XWiki.getCommentsClass() [2] to add the new
fields to the XWikiComments class, I am wondering if I need to do that too
in the migration script. If you look at the script that I have pushed, you
can see that I have done it [3], but, while testing, it seems that the
XWikiComments class is already updated, probably due to
XWiki.getCommentsClass() that gets called before the migration. So the
question is whether I should keep the updating of XWikiComments in the
script or just leave it to XWiki.getCommentsClass()?
2. For some reason, the changes done in the migration script are not saved
to the database. I use session.update(...) and session.delete(...) to store
my changes, but they don`t seem to be committed to the DB. For instance,
the query in my
sortCommentsAndReassignObjectNumbers(...) method [4] does not see the
objects modified by my previous
updateExistingAnnotations(...) method [5]. I must be missing some
Hibernate-specific detail, or maybe I am not managing correctly the
transaction. Can someone please have a look at the script [1] and point me
in the right direction?
Thanks,
Eduard
[1] https://github.com/xwiki/xwiki-platform/pull/34/files#diff-6
[2] https://github.com/xwiki/xwiki-platform/pull/34/files#diff-5
[3] https://github.com/xwiki/xwiki-platform/pull/34/files#L6R109
[4] https://github.com/xwiki/xwiki-platform/pull/34/files#L6R202
[5] https://github.com/xwiki/xwiki-platform/pull/34/files#L6R156
On Fri, Feb 17, 2012 at 5:19 PM, Eduard Moraru <enygma2002(a)gmail.com> wrote:
Hi Anca,
Thanks for your feedback.
Please read below...
On Fri, Feb 17, 2012 at 1:54 PM, Anca Luca <lucaa(a)xwiki.com> wrote:
Yes, sounds good. So I`m left with polishing the comment display and
seeing how I should make the migrator. We should be good to go very soon.
Thanks,
Eduard
Actually I was replying to Guillaume and I was
thinking about the same
set of issues that you mentioned for 2/.
In short, it feels to me that the merging with comments is a
particularization of the annotation system, which we could make by default
in XE, but still allow people that want to change it to change it, with a
bit of code / knowledge about what they're doing. I think very little
people used annotations let alone customized the annotations class, in the
default XE, so, using the 80/20 rule, we can do it.
On 02/17/2012 12:31 PM, Eduard Moraru wrote:
Yes, you are right. I was mislead by the fact that the annotations count
is decided based on the objects on the page. Seems to have been a little
hack that is easily fixable though.
The problems that I am facing now are related to
2 things:
= Approach 1 =
If you use only XWiki.XWikiComments (extended with annotation fileds as
the
pull request is now), you lose the Annotation feature's capability to
specify a different AnnotationClass. This means that, if you want to
customize the annotation, you need to extend the XWikiComments class
directly, and nothing else.
Why would you? I mean, the way I see it, this should be just a change in
the annotations config (from annotationsclass to xwikicomments class) with
a bit of hairstyling for the XWikiComments class to contain the things
needed for annotations (note that we already have an unused field in
XWikiComments designed for that -- highlighttext or something).
What customizers should lose only (in my view) is the possibility to see
annotations in the comments tab, that's it. We can even leave the
annotations tab (who made it use AnnotationClass objects? because if I
remember correctly, when I wrote it, it was using the annotations service
to get annotations, regardless of where / how they're stored ), and we hide it by default. So displaying it
for customizers should
only be a matter of setting in the preferences.
It makes a lot of sense when you look at it that way.
This might break existing clients that have
provided their own annotation
class and will not display their annotations.
As per my comment above, they will just have to enable the annotations
tab. Or not even, since annotations tab is enabled by default in the
current version, if they don't update their preferences, annotations tab
will still stay visible.
Writing a migrator for that
Yes, I forgot about that. Will have 2 options here when the migration will
be run:
1. Rearrange only once the comment IDs, sorting the comments by date
2. Do the sorting at display time (either in velocity in commentsinline.vm
or in the API in XWikiDocument.getComments()), every time, in the
likelihood that the comments ID order might be disturbed by such a
migration (or anything else, for the matter). Most likely the API would be
best since getComments() should order by date anyway instead of id. Even if
it's an in-memory operation and the comments are usually already naturally
sorted, I wonder if would be a considerable performance penalty.
I think option 1 here is most likely the best fit for the problem at hand.
would imply looking at the configured annotation
class that they have
provided, update XWikiComments class with the new fields that might be
present in that annotation class and convert the annotation objects to
use
the updated XWikiComments class instead. Would that be enough?
I think we can write this on request (the migrator), if anyone wants it.
If they don't, their annotation class stays the same as they had, and they
see annotations in the annotations tab and comments in the comments tab. If
they want to benefit from seeing them in the same tab, we'll explain what
code to write to merge their data -- you'll have an ordering issue as well
in there, since newly added comments (as a result of migrated annotations)
will be at the end of the comments list, not "in between" according to when
they were added :).
This solution will ensure that both
"annotation" and "comment" IDs are in
the same namespace and are unique. This also ensures that:
- the replyTo field points to a proper object ID of class XWikiComments
- the html IDs are unique in the page so that the permalinks and anchors
for jumping to the annotation and back to the comments tab work
- the comments are naturally sorted by ID
General downside of solution 1: you lose the ability to change the
annotation backend, since you just look at XWiki objects in the page.
why? as I mentioned, this would be just one use case (using comments
class for annotations), if anyone wants to do differently, they can, and
they need to accept the fact that they won't see annotations in the
comments tab.
= Approach 2 =
The merging solution has a few problems of it's own:
a) The annotation service returns java object Annotations, while the
commentsinline.vm works with XWiki API objects.
To avoid this problem, we can just do what the old AnnotationsTab did,
and
that is to directly get the XWiki objects stored in the page using the
class configured in the annotations config.
Downside: you lose the ability to change the annotation backend, since
you
just look at XWiki objects in the page.
b) Merging things means that we keep both the comments and annotations as
separate entities. This means that each have their own ID namespace,
causing collisions for the replyTo and permalink features. To avoid this,
we would have to do some not-so-nice tricks and mark annotations ID (like
"annotation1") so that they are separated from regular comments IDs (when
used in the replyTo field).
b.1) One solution would be to change the replyTo field's type from Number
to String so that we can store this separation.
Downside1: This still requires a migration.
Downside2: because this means to use the solution for a), we still loose
the ability to change the annotation backend.
b.2)The other solution would be to use the annotation service API when
getting the list of annotations and make a convention when an annotation
is
created so that it's numeric ID starts from a very large number (like
1.000.000 or more).
Downside: Same as problem a): we have to work with 2 types and lots of
"IF
statements" when processing the comments. Ex: One type has comment.number
(for comments), while the other one has comment.id (for annotations).
Same
thing for dates: comment.getProperty('date').**value vs comment.date.
General downside of solution2: Since they are merged and no longer
naturally sorted by object ID, the comments and annotations also need to
be
sorted by date.
I almost read all this, but I had started this thread of thought before
and I agree it's highly complicated compared to solution 1.
= Conclusion=
As I look more into the problem, it does seem that the first approach is
the best, having the least number of compromises and doing a better job
at
semantically merging the two concepts into one.
Do you agree?
+1 for solution 1, with the mentions I made above.
In short, I see solution 1 as the following steps:
1/ enhance default XWiki.XWikiComments class with annotations fields
(selected text, context, annotation state)
2/ change _default_ annotations class to this XWiki.XWikiComments
3/ disable annotations tab by default
3.1/ polish a bit the comments tab to display new fields of annotations,
etc
potentially with a migrator to transform all
AnnotationCode.AnnotationClass annotations to XWiki.XWikiComments
annotations, but not mandatory since not doing so doesn't loose data, it
just doesn't display it in the comments tab.
This way we keep all the benefits of the annotations system, we just
implement a particularization of it in the default XE.
Feels quite easy-ish and elegant like this, to me, I wonder if I'm
missing a detail (I must be missing a detail).
Thanks,
Anca
Thanks for your feedback,
Eduard
On Fri, Feb 17, 2012 at 11:43 AM, Jerome Velociter<jerome(a)winesquare.**
net <jerome(a)winesquare.net>>wrote>wrote:
Le 16 févr. 2012 18:38, "Eduard Moraru"<enygma2002(a)gmail.com> a écrit
______________________________**_________________
devs mailing list
devs(a)xwiki.org
http://lists.xwiki.org/**mailman/listinfo/devs<http://lists.xwiki.org/ma…
:
> Hi devs,
>
> Based on the work done by Anca and Sorin doring the XWiki 2011 Seminar
> Hackaton, I`ve made the following pull request [1] to integrate their
>
work
> with minor changes.
>
> A summary of the changes contained by the pull request are described in
>
the
> jira issue [2].
>
> The problem at the current stage, as Jerome also hinted, is that we
> need
>
to
> do a migration script to make the existing annotations (in an upgrade
> scenario) use the XWikiComments class instead so that they can be
> picked
>
up
> by commentsinline.vm. However, this might lose the possibility to
> provide
> custom annotations.
>
> An alternative would be to make commentsinline.vm use the annotation
> service and handle and retrieve both Annotation and XWikiComment
> objects.
> This way, the current annotations should need no migration script since
> they are using a class configured in the AnnotationConfig page that the
> annotation service knows how to handle.
>
> WDYT?
>
I'm tending to prefer a proper migration. I've played in the past with
aggregation of both classes on a small "recent reviews" macro, and I
recall
some (non-trivial) HQL queries were a pain to write, when doable in a
single query at all. I can look for specific examples if needed.
I fear that if we let both classes live together, some potential future
UI
or service or whatever that exposes boths in a unified manner might be
harder to implement.
Jérôme
Thanks,
> Eduard
>
> ----------
> [1]
https://github.com/xwiki/**xwiki-platform/pull/34<https://github.com/xwi…
> [2]
http://jira.xwiki.org/browse/**XWIKI-7540<http://jira.xwiki.org/browse/X…
> ______________________________**_________________
> devs mailing list
> devs(a)xwiki.org
>
http://lists.xwiki.org/**mailman/listinfo/devs<http://lists.xwiki.org/ma…
>
______________________________**_________________
devs mailing list
devs(a)xwiki.org
http://lists.xwiki.org/**mailman/listinfo/devs<http://lists.xwiki.org/ma…
______________________________**_________________
devs mailing list
devs(a)xwiki.org
http://lists.xwiki.org/**mailman/listinfo/devs<http://lists.xwiki.org/ma…