DeTrusty 0.7.1

DeTrusty - Decentralized and Trustable Query Engine

DeTrusty

DeTrusty is a federated query engine. At this stage, only SPARQL endpoints are supported. DeTrusty differs from other query engines through its focus on the explainability and trustworthiness of the query result.

Notice: DeTrusty is under active development! The current version is a federated query engine following the SPARQL 1.1 protocol, i.e., you can use the SERVICE clause to specify the endpoint. However, the parts about the explainability and trustworthiness have not been implemented yet.

DeTrusty only supports SELECT queries. The following SPARQL 1.1 operations are not supported (yet):

• Aggregation functions, e.g., COUNT, MAX
• Conditional functions, e.g., IF, BOUND, COALESCE
• GROUP BY clause (including HAVING)
• BIND clause
• Specifying RDF datasets using FROM, FROM NAMED, or GRAPH

This list might not be exhaustive. Please, let us know if we missed something.

Running DeTrusty

DeTrusty can be used in two different ways: (i) as a Service by running it with Docker, and (ii) as a Python library within your own project. Both methods will be explained in the following section.

DeTrusty as a Service

In order to use DeTrusty as a service, you need to execute the following steps:

1. Pull the Docker image docker pull sdmtib/detrusty:latest
2. Start the container docker run --name DeTrusty -p 5000:5000 -v $(pwd)/Config:/DeTrusty/Config -d sdmtib/detrusty:latest
3. Create the file ./Config/endpoints.txt
4. Add the URLs to the SPARQL endpoints you want to include in the federation to that file; one URL per line and accessible from within the container.
5. Now it is time to collect the metadata necessary for DeTrusty to operate as expected. docker exec -it DeTrusty bash -c 'create_rdfmts.py -s /DeTrusty/Config/endpoints.txt'
6. Once the metadata is collected, restart the workers in order to reload the configuration. docker exec -it DeTrusty restart_workers.sh
7. After restarting the workers, you can use DeTrusty to query your federation. When running DeTrusty as a service, you can access it through (i) an API, (ii) a command line interface, and (iii) a graphical user interface.

DeTrusty as a Service: API

You can use DeTrusty by making POST requests to its API. If you started the container as described above, the base URL of the API is localhost:5000. In the following, the different API calls are described.

/version

Returns the version number of DeTrusty.

Example call:

curl -X POST localhost:5000/version

Example output:

DeTrusty v0.7.0

/sparql

This API call is used to send a query to the federation and retrieve the result. The result will be returned as a JSON (see example below).

Example call:

curl -X POST -d "query=SELECT ?s WHERE { ?s a <http://dbpedia.org/ontology/Scientist> } LIMIT 10" localhost:5000/sparql

Example output for the above query (shortened to two results):

{
  "cardinality": 10,
  "execution_time": 0.1437232494354248,
  "output_version": "2.0",
  "head": { "vars": ["s"] },
  "results": {
    "bindings": [
      {
        "__meta__": { "is_verified": True },
        "s": {
          "type": "uri",
          "value": "http://dbpedia.org/resource/A.E._Dick_Howard"
        }
      },
      {
        "__meta__": { "is_verified": True },
        "s": {
          "type": "uri",
          "value": "http://dbpedia.org/resource/A.F.P._Hulsewé"
        }
      },
    ]
  }
}

• 'cardinality' is the number (integer) of results retrieved
• 'execution_time' (float) gives the time in seconds the query engine has spent collecting the results
• 'output_version' (string) indicates the version number of the output format, i.e., to differentiate the current output from possibly changed output in the future
• 'variables' (list) returns a list of the variables found in the query
• 'result' is a list of dictionaries containing the results of the query, using the variables as keys; metadata about the result verification is included in the key '__meta__'. The current version returns all results as verified as can be seen in the key 'is_verified' of the metadata.

When sending a SPARQL 1.1 query with the SERVICE clause, you need to set the sparql1_1 flag to True:

curl -X POST -d "query=SELECT ?s WHERE { SERVICE <https://dbpedia.org/sparql> { ?s a <http://dbpedia.org/ontology/Scientist> }} LIMIT 10" -d "sparql1_1=True" localhost:5000/sparql

DeTrusty as a Service: CLI

You can also run DeTrusty from the command line. The following example call assumes a query stored in a file ./query.sparql.

docker exec -it DeTrusty bash -c 'python runDeTrusty.py -q ./query.sparql'

If you want to execute a SPARQL 1.1 query with the SERVICE clause, add -o True.

DeTrusty as a Service: GUI

Starting with version 0.4.0, DeTrusty also comes with a Web interface. Hence, you can run queries using your favorite browser. The Web interface uses the JavaScript library YASGUI. If you started the container as described above, the interface is accessible at localhost:5000/sparql.

DeTrusty as a Library

Starting with version 0.5.0, DeTrusty can also be used as a library. You first need to install it via python3 -m pip install DeTrusty. The following example script shows how you can use DeTrusty within your project.

from DeTrusty.Molecule.MTCreation import create_rdfmts
from DeTrusty import run_query

endpoints = ['http://url_to_endpoint_1', 'https://url_to_endpoint_2:port/sparql']
create_rdfmts(endpoints, './Config/rdfmts.json')

query = "SELECT ?s WHERE { ?s a <http://dbpedia.org/ontology/Scientist> } LIMIT 10"
query_result = run_query(query)

print(query_result)

Of course, you can also further process the results according to your needs.

If you want to execute a SPARQL 1.1 query with the SERVICE clause, add the parameter sparql_one_dot_one=True to your call of run_query. It is also not necessary to collect the metadata in that case.

from DeTrusty import run_query

query = "SELECT ?s WHERE { SERVICE <https://dbpedia.org/sparql> { ?s a <http://dbpedia.org/ontology/Scientist> }} LIMIT 10"
query_result = run_query(query, sparql_one_dot_one=True)

print(query_result)

Running Multiple Queries

The aforementioned method is not very performant in a use case where multiple queries will be executed over the same federation of endpoints. This is due to the fact, that the run_query method will read the metadata from file by default. You can keep the metadata in its internal representation and pass it to the run_query method in oder to save the time of reading it from file everytime.

from DeTrusty.Molecule.MTCreation import create_rdfmts
from DeTrusty import run_query

endpoints = ['http://url_to_endpoint_1', 'https://url_to_endpoint_2:port/sparql']
config = create_rdfmts(endpoints, None)  # returns the configuration instead of writing it to a file 

query = "SELECT ?s WHERE { ?s a <http://dbpedia.org/ontology/Scientist> } LIMIT 10"
query_result = run_query(query, config=config)  # pass config file to avoid reading from file again
print(query_result)

query2 = "SELECT ?s WHERE { ?s a <http://dbpedia.org/ontology/City> } LIMIT 10"
query2_result = run_query(query2, config=config)  # pass config file to avoid reading from file again
print(query2_result)

If you want to keep the configuration for later so that you do not need to compute it again, you can save it to a file:

config.saveToFile('./Config/rdfmts.json')

Private Endpoints

Starting with version 0.6.0, DeTrusty can handle private endpoints that require authentication via tokens. DeTrusty assumes that there is a server providing these tokens and that the response includes the token in the access_token field together with the lifespan of the token in seconds (expires_in).

The aforementioned configuration of DeTrusty changes slightly when using private endpoints since additional information is needed.

DeTrusty as a Service

If you run DeTrusty as a service, in step 3, you will use the file ./Config/endpoints.json instead of a plain text version. The file should look like this:

{
  "https://url_to_endpoint_1": {
    "keycloak": "https://url_to_token_server",
    "username": "YOUR_USERNAME",
    "password": "YOUR_PASSWORD"
  }
}

Each endpoint is a single JSON object identified by its URL (as key). The key keycloak is the URL to the token server; username and password represent the credentials for the token server. In step 5, you need to adjust the file name to the JSON file and add the -j switch to tell the script that you are using JSON input.

docker exec -it DeTrusty bash -c 'create_rdfmts.py -s /DeTrusty/Config/endpoints.json -j'

Note: If your SPARQL endpoint uses basic authentication with Base64 encoding instead of a token server, simply omit the entry for keycloak.

DeTrusty as a Library

If you use DeTrusty as a library, you will need to pass a dictionary to the create_rdfmts() method instead of list. The following example shows how:

from DeTrusty.Molecule.MTCreation import create_rdfmts
from DeTrusty.Molecule.MTManager import ConfigFile

endpoints = {
  'https://url_to_endpoint_1': {
    'keycloak': 'https://url_to_token_server',
    'username': 'YOUR_USERNAME',
    'password': 'YOUR_PASSWORD'
  }
}
rdfmt_file = './Config/rdfmts.json'
create_rdfmts(endpoints, rdfmt_file)
config = ConfigFile(rdfmt_file)

The keys of the endpoint_dict are the URLs of the SPARQL endpoints; just as in the endpoints.json file described above. Each endpoint is represented as a dictionary itself; holding all parameters in the form of (key, value) pairs.

Metadata from RML Mappings

Starting with version 0.6.0, DeTrusty can collect the metadata necessary for source selection and decomposition from RML mappings. Collecting the metadata from RML mappings instead from the SPARQL endpoint considerably increases the performance of the metadata collection process. Of course, this is only feasible for endpoints that were created using RML mappings.

The aforementioned configuration of DeTrusty changes slightly when collecting the metadata from RML mappings.

DeTrusty as a Service

If you run DeTrusty as a service, in step 3, you will use the file ./Config/endpoints.json instead of a plain text version. The file should look like this:

{
  "https://url_to_endpoint_1": {
    "mappings": [
      "path/to/mapping/file1",
      "path/to/mapping/file2"
    ]
  }
}

Each endpoint is a single JSON object identified by its URL (as key). The key mappings is a list of all mapping files associated with the endpoint. In step 5, you need to adjust the file name to the JSON file and add the -j switch to tell the script that you are using JSON input.

docker exec -it DeTrusty bash -c 'create_rdfmts.py -s /DeTrusty/Config/endpoints.json -j'

DeTrusty as a Library

If you use DeTrusty as a library, you will need to pass a dictionary to the create_rdfmts() method instead of list. The following example shows how:

from DeTrusty.Molecule.MTCreation import create_rdfmts
from DeTrusty.Molecule.MTManager import ConfigFile

endpoints = {
  'https://url_to_endpoint_1': {
    'mappings': [
      'path/to/mapping/file1',
      'path/to/mapping/file2'
    ]
  }
}
rdfmt_file = './Config/rdfmts.json'
create_rdfmts(endpoints, rdfmt_file)
config = ConfigFile(rdfmt_file)

The keys of the endpoint_dict are the URLs of the SPARQL endpoints; just as in the endpoints.json file described above. Each endpoint is represented as a dictionary itself; holding a list with all mapping files associated with the endpoint under the key mappings.

Restricting Classes of an Endpoint

Starting with version 0.7.0, DeTrusty can restrict the collection of metadata to specific classes of an endpoint. The mentioned classes need to be specified using their full URI.

The aforementioned configuration of DeTrusty changes slightly when collecting the metadata from RML mappings.

DeTrusty as a Service

If you run DeTrusty as a service, in step 3, you will use the file ./Config/endpoints.json instead of a plain text version. The file should look like this:

{
  "https://url_to_endpoint_1": {
    "types": [
      "http://example.com/ontology/ClassA",
      "http://example.com/ontology/ClassB"
    ]
  }
}

Each endpoint is a single JSON object identified by its URL (as key). The key types is a list of RDF classes for which the metadata will be collected. All other RDF classes in the endpoint will be ignored. In step 5, you need to adjust the file name to the JSON file and add the -j switch to tell the script that you are using JSON input.

docker exec -it DeTrusty bash -c 'create_rdfmts.py -s /DeTrusty/Config/endpoints.json -j'

DeTrusty as a Library

If you use DeTrusty as a library, you will need to pass a dictionary to the create_rdfmts() method instead of list. The following example shows how:

from DeTrusty.Molecule.MTCreation import create_rdfmts
from DeTrusty.Molecule.MTManager import ConfigFile

endpoints = {
  'https://url_to_endpoint_1': {
    'types': [
      'http://example.com/ontology/ClassA',
      'http://example.com/ontology/ClassB'
    ]
  }
}
rdfmt_file = './Config/rdfmts.json'
create_rdfmts(endpoints, rdfmt_file)
config = ConfigFile(rdfmt_file)

The keys of the endpoint_dict are the URLs of the SPARQL endpoints; just as in the endpoints.json file described above. Each endpoint is represented as a dictionary itself; holding a list with the RDF classes that should be considered during the metadata extraction under the key types.

License

DeTrusty is licensed under GPL-3.0.