When a document is already in a corpus...

[benel]

... use this corpus instead of the user's "bookmarks".

This applies especially to texts in Cassandre, since all of them are already recorded as being in a corpus

[chao]

I don't understand clearly, what do you mean a document in a corpus?

For example: http://cassandre.hypertopic.org/text/IF14/0df56d1f70eee373461ae249350008f5 is in Corpus "IF14" right?

So for this document I should use Corpus "IF14" instead of using users' "bookmarks", right?

[benel]

Right.

LaSuli should query Cassandre first and then Argos:

http://cassandre.hypertopic.org/item/?resource=http://cassandre.hypertopic.org/text/IF14/0df56d1f70eee373461ae249350008f5

{"rows":[
{"key":["http://cassandre.hypertopic.org/text/IF14/0df56d1f70eee373461ae249350008f5"], "value":{"item":{"corpus":"IF14", "id":"0df56d1f70eee373461ae249350008f5"}}}
]}

• Because Cassandre knows the corpus and it is IF14, items created on Argos will use item_corpus:"IF14".
• If Cassandre does not know any corpus for the resource but Argos does, then the latter is used.
• If Argos doesn't know any corpus either, then the "bookmarks" corpus can be used.

[chao]

Fixed

[benel]

I tested 3 possible implementations:

• with Cassandre defined as the first Hypertopic server (which would mean that it is the first server to ask),
• with Cassandre defined as the second Hypertopic server (the first one being the read-write server),
• with only Argos defined as an Hypertopic server (which means that LaSuli would check automatically if the corpus server is Hypertopic compliant).

None seems to work... Did you implemented it in a fourth way?

[benel]

Please, note that what is said about the reuse of the corpus should be done also with the reuse of the item ID.

[benel]

Hmm, I suppose that my specifications were not clear enough. Now that Steatite is ready for integration it should be easier to explain it and to test it. Here is a real life scenario:

1. Set up LaSuli with the following parameters:
   • http://argos2.youknowwhere/ yourlogin yourpassword (default)
   • http://steatite.youknowwhere/ (empty) (empty)
2. Go to http://steatite.youknowwhere/picture/c8ffb3e9346ac80ea730e512a57fdee79a34aac8 LaSuli should show "Boucle d'oreille" as the name.
3. Add the attribute grave: 3. LaSuli should create a new item on Argos with the same ID (c8ffb3e9346ac80ea730e512a57fdee79a34aac8) and the same corpus (Megara Hyblaea) it found on Steatite at step 2.

Note: When LaSuli searches an item at step 2, it must query the servers in the same order and stop at the first positive answer.

[benel]

Note that to do the same test again, you will have to delete the new item in Argos at the end of the test.

[benel]

The code to fix is there: https://github.com/Hypertopic/LaSuli/commit/e486f621a0252b0e7dc4517c4f6c2527076cf99f#commitcomment-2788165 and https://github.com/Hypertopic/LaSuli/commit/9b0d8d385d9aeceb139976850208c4755fead00a#commitcomment-2788303.

[Slals]

I'm currently working on it, I need some insights to continue!

First I found a way to get the existing item id as id when LaSuli posts the item in the default server (argos as it is in your example).

But for the corpus it's a bit though. Actually, when LaSuli creates a corpus item it calls the method createWithId(id, name). The point is, when this method is called, the same variable is put as parameters : as id and as name, see here. I'm only able to get the corpus name of the resource.

Now, I need some insights. In Steatite, an item and a corpus have both the same id. Regarding to this, it could be a solution to get the corpus name and to get the item id in order to call the method createWithId(itemId, corpusName). This would be correct for the Steatite case, but what about other servers?

[benel]

First I found a way to get the existing item id as id when LaSuli posts the item in the default server (argos as it is in your example).

I created a patch for this yesterday. I will push it on a dev branch (since there are other related issues to be fixed).

In Steatite, an item and a corpus have both the same id.

What do you mean? For example 71ff73e95a7df128d247c01d0892ba5c3c7cbfe0 is an item ID while Vitraux - Bénel is its corpus ID.

[Slals]

For example 71ff73e95a7df128d247c01d0892ba5c3c7cbfe0 is an item ID while Vitraux - Bénel is its corpus ID.

That exactly what I meant. I thought at first a corpus id was the item id. The database structure in Steatite looks like that (id, attribute, value).

So, for an item 71ff73e95a7df128d247c01d0892ba5c3c7cbfe0, in Steatite we have : (71ff73e95a7df128d247c01d0892ba5c3c7cbfe0, name, SNZ 004) (71ff73e95a7df128d247c01d0892ba5c3c7cbfe0, corpus, Vitraux - Bénel)

Here the corpus ID and the item ID are the same. I mistaken corpus ID because of that, but now it's clearer.

Your solution has the same result as mine and it's shorter :+1: .

(since there are other related issues to be fixed).

Are you refering to issues in this ticket? Does your solution resolve all the issues?

[benel]

Here the corpus ID and the item ID are the same. I mistaken corpus ID because of that, but now it's clearer.

(71ff73e95a7df128d247c01d0892ba5c3c7cbfe0, corpus, Vitraux - Bénel) should be read as "item 71ff...'s corpus is Vitraux - Bénel".

This is a triple notation just like in RDF.

Think about is as the edge of a graph: 71ff... ---corpus---> Vitraux - Bénel

[benel]

(since there are other related issues to be fixed).

Are you refering to issues in this ticket? Does your solution resolve all the issues?

There is a related problem with the name attribute: only the name from the read-write server is shown. This prevents to test easily (from the user interface) that the patch worked.

Highlights are also broken, but this seems to be unrelated with my patch.