10 Apr
2012
10 Apr
'12
9:19 a.m.
Hi Paul,
Thanks a lot for your comments.
I took a bit of time to check your patch.
Here's a few comments, not really in order but more as a recipient of the
work.
The readme is ok... I am not sure it follows the xwiki practices. I guess
we should discuss how this search should be advertised one day, either as
an extension (the readme would be in the wiki then) or as a default
component.
The Readme is not complete. I didn't follow the XWiki standard
practice as I
created one to find the necessary steps to set it
up. I prefer having a complete documentation in dev wiki or extension
wiki.
Shouldn't the versions be 3.5.1-SNAPSHOT?
This is little confusing, dunno how the community follows it. As said
by Fabio, if we are going to use XWiki Contrib repository its
good to start with 1.0.0-SNAPSHOT (using Continuos Integration ) or
1.0.0. If this is going into platform it should be
4.x-SNAPSHOT. I see the master repository is on 4.x in github.
The values in xwiki.cfg all carry the prefix solr, which is good, but I
think the prefix search must be kept (so it's be search.solr or something
so).
I agree with search.solr.*. But am doubtful about xwiki.cfg .
Recently, I found
out that there is one more file called
xwiki.properties, looks like its a better place to set up the
configuration.
Currently, solr-home is set by an environment
variable. There must be a
way to define the solr home without it... not sure how or when this should
be the case but this is a hard issue I think since many deploy their xwiki
in a place where the startup variable are not even touchable.
May be we can include the solr configuration inside WEB-INF of xwiki
enterprise and pick it up from there.
The readme says to copy two jars, but they have the same name (one of them
should be with search-solr).
Sorry, its my mistake. Am going to rectify it.
It is excellent that your readme works with an identified version of
Lucene and one of SOLR, it has often been the case that identified versions
of solr worked with a privately crafted lucene version.
I am not sure all the packages you include are needed,
e.g. why is
lucene-spatial useful? why is spellchecker useful? is solr-velocity really
useful? (isn't that a completely different velocity run method?) why is
solr-cell (the document attachments' parser I think) not useful?
I haven't spent much time here. The developed component can be
considered as Proof Of Concept with SOLR. It needs to be
refined and standardized.
You have omitted the expansion of solr_config.tar.gz in your readme at the
right place. I note that it'd be highly wishable to separate solr-home/conf
(typically inside WEB-INF/solr) and solr-home/index (typically inside
work/solr) but it may be something hard to achieve.
Yes, I have that in mind. I need to spend more time on how XE is build,
how the artifacts and configuration files are picked
up.
I had a slightly difficult time to make it all work at the start with the
view code indicated in the readme. Too many things to consider for a first
time (sensible defaults should just work, the fields should be prefilled,
also, I think the form should get and not post so that search results are
bookmarkable).
Also, I think you did not include the highlighting in the current search
view code, but that's ok for now.
Also, you haven't indicated in the readme that the wiki should be
multilingual for all the tests.
Sorry for missing that out .
But all in all it is quite good and it worked for me.
So I am happy with this first contribution which shows that it's possible
for you to get something workable in short time.
There remains an amount of issues which should all be addressed when (and
if) the project starts.
I have a rather good feeling.
paul
Thanks once again for your feedback. As the implementation was done in
haste, I couldn't follow all the XWiki standards. I was rather
concentrating on getting a basic solr search working. As we have something
in place now, I would like to refactor the existing code with your comments
here and start with it this summer.
Thanks,
Savitha.
Le 6 avr. 2012 à 18:47, savitha sundaramurthy a écrit
:
--
best regards,
Savitha.s
Thanks Ludovic and Guillaume.
As Fabio pointed out changing the instantiation of SolrServer(), new
EmbeddedServer() ,
or CommonHttpServer() will do. If any case, there is something else to
offer I'll make it into two different components.
Am really looking forward to work with XWiki community.
Thanks ,
Savitha
On Thu, Apr 5, 2012 at 5:43 AM, Ludovic Dubost <ludovic(a)xwiki.com>
wrote:
> Hi Savitha,
>
> This is quite cool. I'm quite impressed at the speed you learned to do
> components and manage to do a first basic implementation.
> It indeed seems to show that SOLR is a good choice.
>
> I've one remark which is the usage of a SOLR Embedded server. I think
one
> of the interesting features of SOLR is also
that we can split the search
> engine which the XWiki engine.
> It would be great to have the choice between both solutions (embedded
and
> not embedded). By default we would bundle the
embedded version and for
high
> performance installation it should be
possible to separate the SOLR
engine
> and still share most of the logic.
>
> Keep up the good work
>
> Ludovic
>
> 2012/4/5 savitha sundaramurthy <savitha.s131(a)gmail.com>
>
>> Hi Paul,
>>
>> Here is the link to the document
>>
>>
>>
>
https://docs.google.com/document/pub?id=1-2GqvjtzS_a8TRKDNsom9egDZHTT5xIzUf…
>>
>> Thanks a lot,
>> Savitha
>>
>>
>>> On Wed, Apr 4, 2012 at 12:03 PM, Paul Libbrecht <paul(a)hoplahup.net>
>> wrote:
>>>
>>>> Hi Savitha,
>>>>
>>>> things are moving. This is good.
>>>>
>>>> Do you have a (temporary) server address where this could be tested?
>>>> (maybe best over IRC?).
>>>> Or, how could I build your code and inject it on a fresh XWiki?
>>>>
>>>> Le 4 avr. 2012 à 18:46, savitha sundaramurthy a écrit :
>>>>> I went through Fabio Mancinelli's code and with the help
of
>> few
>>>>> inputs from there , I have implemented a basic search functionality
>>>> using
>>>>> solr. It also fixes the issue
> http://jira.xwiki.org/browse/XWIKI-6226.
>> I
>>>>> have indexed the page in English, french and Spanish. I have
>> implemented
>>>>> these main functionality:
>>>>>
>>>>> 1) Simple text search using solr.
>>>>> 2) Used *Extended Dismax Parser *to customize search relevancy using
>>>> boost
>>>>> index.
>>>>
>>>> which fixes well the apostrophe case for sure (and a zillion other
>>>> ambiguities, and another zillion power user tricks).
>>>>
>>>>> 3) Hit highlighting for English.
>>>>
>>>> This is a good usage of solr.
>>>>
>>>>> Below is the link to the source code
>>>>> https://github.com/savis/xwiki-platform-search
>>>>> XWiki front end , HTML and Velocity code
>>>>>
>>>>> https://gist.github.com/2295648
>>>>>
>>>>> I wanted to send a pull request but I was playing around when
> renaming
>>>> my
>>>>> username and deleted the forked repository and I'm having
difficulty
>>>>> forking it again. So I have created a new repository and shared the
>> code
>>>>> above.
>>>>
>>>> I think forking is really useful. If not today, it will be needed any
>> day
>>>> soon.
>>>> We need to see the diffs.
>>>> Best would be to create the fork again, then copy your files and let
> git
>>>> give us the diff.
>>>>
>>>>> I have taken few screen shots and have attached the document.
>>>>
>>>> where?
>>>>
>>>>> I'm working on my application and it should be done today.
>>>>
>>>> What is the planned schedule?
>>>>
_______________________________________________
devs mailing list
devs(a)xwiki.org
http://lists.xwiki.org/mailman/listinfo/devs