Usage of Qlever via C++ API

[adel-ayara]

Hello,

Is it possible to control Qlever directly via C++ API instead of the provided script?

Best regards,
Adel

[hannahbast]

@adel-ayara If you enter a query on https://qlever.cs.uni-freiburg.de and click on the "Share" button, you see various API calls. You can use these to get results from the command line (via curl) or from any programming language (you need a library to send a HTTPS request). Note that the API is fully standardized, so the same API calls that work for https://query.wikidata.org also work for https://qlever.cs.uni-freiburg.de/wikidata

[adel-ayara]

@hannahbast Thanks for your answer1
But is it possible to directly control Qlever via C++ programming API (function call) without the usage of HTTPS? I don't want to query an existing store like wikidata but our proper data

[hannahbast]

@adel-ayara It's certainly possible, but why would you want that? Communicating via HTTPS is very comfortable, has very low overhead when your program and QLever run on the same machine, and you compile the programs separately.

[adel-ayara]

We would like to deploy and test Qlever in embedded system like car where A HTTPS communication may not be available. This is why we want to call its different functions via a C++ API (direct function call)
But if you say that it's possible can you send some references on how to do it?

[hannahbast]

@adel-ayara Can you provide a bit more information on the program from which you want to make function calls to QLever:

1. Should the program only ask (SPARQL) queries for data that has already been loaded (preprocessed) before? Or should the program also load the data (this takes some time, depending on the size of the data).
2. How large is your dataset? (how many triples)
3. Is your program also written in C++ or in another language?

[adel-ayara]

The background of my question is as following:
we have map (geo spatial) data extracted from an NDS binary BLOBs and sent to the car via the air as compressed RDF triples (ttl) files. Then we would like to load the data in a kind of triple store locally in one of the car ECUs and execute some SPARQL queries.

Only C++ is supported and the same C++ program that will load the data should also query the store. The data represents the map information located in the car region of interest and it's around 100K triples. But it changes with the car movement.

In this kind of embedded environment, accessing the disk is usually very slow and not recommended. So, the optimal solution should support loading the data in memory once at start up phase and then do all operations (add, delete and query) on memory

[hannahbast]

@adel-ayara Thanks for the additional information. I think I understand what you need now.

A meaningful first step would be to write a small C++ program with a main function that creates an index from a string with a set of triples and then asks some queries on it. There are already such functions in our unit tests, for example, https://github.com/ad-freiburg/qlever/blob/master/test/IndexTestHelpers.h#L36-L47.

The code that is there builds the index, writes it to disk, and then reads it from disk again. There are several ways to avoid reading from disk. One is to not use real files but memory-mapped files. They behave just like files but reside in RAM. Another possibility would be to create a RAM disk with a program like tmpfs, which creates a file system that resides in RAM.

How proficient are you in writing C++ code?