cancel
Showing results for 
Search instead for 
Did you mean: 

Language Issue in Repository Manager and T-REX

Former Member
0 Kudos

Hi Everybody,

we are trying to implement a repository manager for SAP KM. Our systems stores documents language specific in a database. Now we are wondering how we should create a view to the documents, so that every document in each language is indexed by T-REX correctly. Lets say the internal view is something like this:


/
   /Documents
      ->File.pdf
  /Images
      ->Image.jpg

In this case File.pdf and Image.jpg can exist in different languages. So our question is, does T-REX create an index for each language available in the portal? If so, which would be very nice, how can we get the current language T-REX is using for indexing within our RM implementation?

Otherwise we have to create virtual TOP-Level laguage folders for each language in our database that would look like this


/de
   /Documents
      ->File.pdf
  /Images
      ->Image.jpg
/en
   /Documents
      ->File.pdf
  /Images
      ->Image.jpg
...

We try to avoid this, since File.pdf ist a single item with different language versions.

Does anybody know the best way to solve this?

Kind Regards,

Ali

Accepted Solutions (1)

Accepted Solutions (1)

Former Member
0 Kudos

Hello,

I fear you have to implement your approach with the different language folders. To get the documents into the index a crawler ist started, which runs over all documents unsing the getChildren call of the repository. Therefore the crawler has to find all documents and especially all language versions of each document.

Regards,

Achim

Former Member
0 Kudos

Hi,

well i was wondering if we can solve this in another way. Since TREX, or lets better say the crawler service, uses a special user (index_service) to create an index, our first idea was to create the language specific folders only for this user... however this would be a hack and we don't wan't to do this...

Is it maybe possible to configure different crawlers each using a specific user with a configured language? How can this be done and how would it affect the results returned on a search?

So where are those real TREX experts out there givin me some hints?

Regards,

Ali

Former Member
0 Kudos

It is not possible to configure different crawlers with a special user. The indexmangement service is starting all crawlers for indexing and this service starts the crawler with exactly this user.

Creating folder wich are only accessible for the index_service user is also not a good idea. In this case no user wild find these documents, because the access rights for search results are checked.

Regards,

Achim

detlev_beutner
Active Contributor
0 Kudos

Hi Achim, hi Ali,

> Creating folder wich are only accessible for the

> index_service user is also not a good idea

I know of (a certified) repository manager which just uses the trick of "shadowed" structures for TREX. It does not mean (and I think/hope that Ali [please to meet you] had this in mind) really to have some folder structure which is only accessible for the crawler user but at runtime to present such a shadowed structure, however, to provide the link for the "original" resource and it's position at the same time.

When I checked that implementation, I really liked the idea. I think it may be called "hack", but a nice one

Hope it helps

Detlev

D021954
Advisor
Advisor
0 Kudos

Hi Ali,

sorry, I didn't get your scenario?

Is it important for you that documents are indexed at TREX with the 'right' language?

Or do you want to search language specific?

Both?

Regards Matthias

Former Member
0 Kudos

Hi Matthias,

Here the scenario:

We want to create only one view to the repository.

Users sccess the repository should get each document in their user specific language (if not available our system automaticly returns the document in a default language).

When searching we want that all language versions of the documents are indexed.

Does this help you to understand our problem?

Kind Regards,

Ali

Former Member
0 Kudos

Hi Detlev,

very nice to meet you too

I have never heared about the shadowed structures. Would you mind to send me some links where I can find more informations about that? Is it possible to find out more how the guys from the certified RM done all this?

Thank you very much

Ali

detlev_beutner
Active Contributor
0 Kudos

Hi Ali,

> I have never heared about the shadowed structures

It's my own term for more or less this (your) solution:

> > the language specific folders only for this user

> Would you mind to send me some links

They don't exist, for it was the original idea of the guy who implemented this. I only reviewed the solution and debugged some minor problems. And also this is about one year ago.

It may eventually also worked like this, just an idea: present the crawler /de/Documents/File.pdf etc pp, whereas the user only sees /Documents/File.pdf etc pp. Searching for a term in the german File.pdf would lead to the search result pointing to /de/Documents/File.pdf which by the RepManager gets translated to /Documents/File.pdf<german version>.

Or there was a trick that when the crawler crawls /de/Documents/File.pdf, you return that it should remember this link under /Documents/File.pdf. As said, I don't remember exactly, so I don't know if the API offers this possibility.

> how the guys from the certified RM done all this?

That was a small customer/partner project, so I won't be able to go too far into the details, sorry. I can only present my general knowledge which I have gained so far...

Hope it helps nevertheless

Detlev

D021954
Advisor
Advisor
0 Kudos

Hi Ali,

very nice idea.

With some tricks it might be possible to show users only language specific documents. But a fall back to a default language document is IMHO impossible without code extension.

A context sensitve document resolution (variant handling) is planned for a future release. But don't expect it in a short time range.

SOrry and Regards Matthias

Former Member
0 Kudos

Hi Matthias,

I think the language handling and the mapping of default languages is application specific and should be done within the RM. In our case this available allready, since our WCMS system allready take care about that... So for a logged in user, we are able to create a language specific folder structure very nicely.

Our problem is the indexing by TREX. We wan't to make sure that every language version of the docment can be found when people searching. So in our case the perfect solution would be that TREX asks the RM for available languages in the repository and starts a crawler for each language available in the repository.

Something that I am not sure about is the presentation of the search results. Since the URL of the search result might point to a different language version than the logged in user has. So if the URL is

http://server/irj/go/km/repsotory/Documents/File.pdf

this might result to a german document, even if the searched term is in the english document. So the language must be set in the search result. For example like this

http://server/irj/go/km/repsotory/Documents/File.pdf?locale=en

This would be our preferred way to handle this issue. I think we will keep the virtual language folders until the new feature is available.

Thank you guys for this very high quality answers )

Kind Regards,

Ali

Answers (0)