Query running forever in QLever

[divral]

Hello,

I have a query that runs in Wikidata Query Service.
Its purpose is to re-create this Wikipedia List.
This is surprisingly tricky and even though it means getting only 11/35 hits, I resorted to using the "largest city" class instead of determining the largest city of a country manually by getting the (most recent) population countf or each of the cities in country and then taking the MAX.
This is the same query in QLever. There, I keep getting the "no reply from backend" error. I also tried yesterday evening and got the same problem.
Is there a problem with QLever? Or is the problem my query (somehow)?

Another thing I wonder about that isn't really related to QLever. I tried slowly improving the above query by manually determining the largest city for each country as a first step. Now, this query attempt does not run in WDQS either as there is a timeout.
I am just a bit baffled by that. I realize that this query text is huge and that the subquery calculating the date of the most recent population count is likely the offender. But I know no way around this and it seems like this query shouldn't be too difficult to run on any SPARQL engine. Am I missing something? It seems I have to get the most recent population counts (of all cities) before I can determine the largest city in a country.

I would like to get some feedback on my first question (paragraph). I am not expecting any on my second, but I would appreciate it very much as I've spent quite a while with this query now...

Thanks in advance!
Christina

[joka921]

Hi Christina,
This query crashes QLever's query planner because it has too many triples (we currently don't have a good algorithm for planning such queries). If you manually make it simpler by adding {} braces around parts of the query, it works. This forces the query planner to first evaluate the parts of the query that are inside the braces. The following variant of the query works:
click here .
When you click on the Analysis button after executing the query you can see that most of the time is spent planning the query and not executing it.

For your other query:
Blazegraph (the engine that is currently running WDQS struggles with queries that have large intermediate results. I will try to let it run on QLever by using similar techniques as described above.

Best regards
Johannes

[joka921]

As for your second query:
This works on QLever:
click me!

Again I had to manually specify some braces and remove the ROUND function which is currently not supported by QLever.
The result seems surprising,
but probably you need to add some P279* for the cities relation (most "cities" are not "cities", but some other subclass of "human settlement" in Wikidata.

[divral]

Hi Johannes,

thank you very much for your help!

Yes, I noticed the little problems with the ROUND function and the "city" subclass issue (Prof. Bast pointed it out in the past), but handling all these query attempts got a little messy, so I did not correct everything in this post.
There was also the issue that I forgot that the capitals should not have an "end time" statement which, when left out, resulted in Bayeux as the capital of France. There are always very many things to keep in mind when writing queries.

Kind regards
Christina

[divral]

Hey Johannes,

I am still working on that query (as I would like to put it into a small benchmark I am creating and therefore it should be as correct as possible). My question has nothing to do with QLever, but I think it still would be most useful to post it here (with the above context):

This is my best version of the query right now. I cannot figure out how to include the following three countries in the output:

India - capital: New Delhi - largest city: Mumbai
Palau - capital: Ngerulmud - largest city: Koror
Morocco - capital: Rabat - largest city: Casablanca

I am pretty sure it is because India's capital statement does not have any start time and in fact, if I make all "start time" containing parts OPTIONAL, India is in the output.
Palau has a capital statement with a start time, but one of the two cities, Ngerulmud, does not have a "point in time" qualifier for its (only) population statement.

So it is about missing values, but I cannot just use OPTIONAL because this will affect all countries (also ones including all information).
I have thought about maybe using UNION to also include, e.g., countries which only have statements without any qualifier "start time" for the capital property. But I don't know how (in general, not only concerning QLever).

The thought just occurred to me that maybe this is the "correct" output already - with the info available on Wikidata?
At least it could be considered correct.

I would really appreciate your help if you have time!

Thanks in advance!
Christina