Lucene not working correctly with "Cluster:xx" query

[a-marcel]

OrientDB Version, operating system, or hardware.

• v2.2.7[x]

Operating System

• Linux
• MacOSX
• Windows
• Other Unix
• Other, name?

Expected behavior and actual behavior

i've created an lucene index with following java:

final ODocument metadata = new ODocument();
    metadata.field("default", "org.apache.lucene.analysis.de.GermanAnalyzer");
regionType.createIndex(CLASS_NAME + ".name_de", "FULLTEXT", null, metadata, "LUCENE",
        new String[] { "name_de" });

i insert this test data:

[{"name_de": "nord"},{"name_de":"nördlich"}]

Steps to reproduce the problem

My select query:

select * from Region where [name_de] LUCENE 'nor*'

returns the two rows and this is what i expect.

For the future, i'd like to query two or more classes with one query. i figure out, this is possible via "cluster:". my query is now:

select * from cluster:Region where [name_de] LUCENE 'nor*'

and this returns just one row -> "nord".

Is that a valid behaviour ?

Thanks
Marcel Alburg

[wolf4ood]

@marcelalburg

What do you mean by

For the future, i'd like to query two or more classes with one query.

[a-marcel]

Hey, sorry maybe i use the wrong expression to describe that.

in the future, i'll use something like this:

select * from cluster:[Region,Abc,Cde] where [name_de] LUCENE 'nor*'

Thanks
Marcel

[wolf4ood]

@marcelalburg

I think it will not work.

Indexes works at class level not cluster level.

[a-marcel]

Mhh ok, then i will use a manual index for this.

Maybe you could answer me one more question. If i add multiple times the same value with the same vertex (#nn:n) to a manual index. is there any problem about that ?

Thanks for your answer
Marcel

[wolf4ood]

@marcelalburg

no problem. You will get multiple times the same vertex in the results set.

Can you explain a little bit you use case maybe i can help more

Thanks

[StarpTech]

Hi @maggiolo00 indexes arent supported for Clusters? This would mean that the concept of clustering and distribution is uncomplete?!

When I understand correctly then clustering is a mechanism of grouping but the set are still classes so mechanism of look up is missing.

[StarpTech]

The same issue is that edges cannot be deleted by a cluster selection.

#6590

@maggiolo00 whats the roadmap for the clustering strategy ?

[a-marcel]

Hello,

i try to explain my usecase:

I've following classes:

Continent
Country
Region
Area
aso.

Every Class has different properties but min. a "name" property.

and i like to create an AutoComplete/Suggest Lucene index on this field and all classes (this works). But i need to query them now.

I hope this explained it a little bit.

Thanks
Marcel

Code Example for one "class":

OrientVertexType areaType = noTxGraph.createVertexType(CLASS_NAME);
    areaType.createProperty(KEY_NAME, OType.STRING);
    areaType.createProperty("name", OType.STRING);
    areaType.createProperty("latitude", OType.DOUBLE);
    areaType.createProperty("longitude", OType.DOUBLE);
    areaType.createProperty("route_count", OType.INTEGER);
    areaType.createProperty(PROPERTY_SUGGEST_NAME_DE, OType.STRING);
    areaType.createProperty(PROPERTY_SUGGEST_TAGS_DE, OType.STRING);

    final ODocument metadata = new ODocument();
    metadata.field("default", "org.apache.lucene.analysis.de.GermanAnalyzer");

    areaType.createIndex(CLASS_NAME + ".suggest_de", "FULLTEXT", null, metadata, "LUCENE",
        new String[] { PROPERTY_SUGGEST_NAME_DE, PROPERTY_SUGGEST_TAGS_DE });

[a-marcel]

Hello, i found an other solution for my problem.

i create a base Class "Searchable" wich one contains all fields for searching. On this class, i create my fulltext lucene index.

all other classes (see above) extends this "Searchable" class.

Now, it's possible to query just the "Searchable" class with the LUCENE Query.

Thanks
Marcel

[StarpTech]

Hi @marcelalburg your solution is no answer to your issue. Its still an issue. Please reopen or I will open a new one because it will address me soon. Thanks.

[StarpTech]

@maggiolo00 can you answer my questions? Thanks.

[robfrank]

Hi, on the next 2.2.16 version Lucene indexes will contain additional metadata fields _CLUSTER and _CLASS. So a query could be partitioned by cluster:

SELECT FROM Person WHERE name LUCENE '+_CLUSTER:26  OTHER QUERY ' "

[StarpTech]

You closed the issue before it released. Awesome!

[robfrank]

What do you mean?

[StarpTech]

@robfrank the current release is 2.2.15 you closed the issue before 2.2.16 was released.

[robfrank]

Can you test with a snapshot? I mean 2.2.16-SNAPSHOT.

[robfrank]

this feature is included in 2.2.16. I close it, feel free to open a new issue relative to the feature as it is implemented.