sparql-anything / pysparql-anything Goto Github PK

Currently, the SparqlAnything.construct method returns a rdflib graph object.

While it is not tricky to then convert this to a nx.MultiDiGraph, would it be worth presenting this directly to the user as an option @enridaga ?

Hi Guys, I have tried to run SPARQL anything onmy laptop and I get the following error. What shall I do?

engine = cli.SparqlAnything()
---------------------------------------------------------------------------
JavaException                             Traceback (most recent call last)
Cell In[3], line 1
----> 1 engine = cli.SparqlAnything()

File ~/anaconda3/envs/py38/lib/python3.8/site-packages/pysparql_anything/sparql_anything.py:20, in SparqlAnything.__init__(self)
     18 def __init__(self):
     19     """ Constructor for the class SparqlAnything."""
---> 20     self.receiver = Engine()

File ~/anaconda3/envs/py38/lib/python3.8/site-packages/pysparql_anything/engine.py:35, in Engine.__init__(self)
     33 """ Constructor for the class Engine. """
     34 location = 'com.github.sparqlanything.cli.SPARQLAnything'
---> 35 self.reflection = autoclass(location)

File ~/anaconda3/envs/py38/lib/python3.8/site-packages/jnius/reflect.py:209, in autoclass(clsname, include_protected, include_private)
    206 cls_start_packagename = '.'.join(clsname.split('.')[:-1])
    208 # c = Class.forName(clsname)
--> 209 c = find_javaclass(clsname)
    210 if c is None:
    211     raise Exception('Java class {0} not found'.format(c))

File jnius/jnius_export_func.pxi:22, in jnius.find_javaclass()

File jnius/jnius_utils.pxi:79, in jnius.check_exception()

JavaException: JVM exception occurred: com/github/sparqlanything/cli/SPARQLAnything java.lang.NoClassDefFoundError

We developed a method to get the output from the Java CLI but we need the Java code to implement it!

Python has a more convenient installation method so we should promote it as the primary way of getting the CLI onboard.
However, the Python based CLI does not use the full list of options as the Java version.

Rather than the current single letter arguments, q, f, o, etc., it has been suggested by @enridaga that these should be changed to the more explicative query, format, output, etc.

Version: sparql-anything-v0.8.2.jar

This issue is not an important one. It could be a small cosmetic improvement.

When I tried to sa.run() a query as below, it failed because I mistyped the filename, with the error as below. I think the error message could be more informative, eg FileNotFoundError.

In [4]: sa.run(q='query1.sparql')
[main] INFO com.github.sparqlanything.cli.SPARQLAnything - SPARQL anything
---------------------------------------------------------------------------
JavaException                             Traceback (most recent call last)
Cell In[4], line 1
----> 1 sa.run(q='query1.sparql')

File ~\miniconda3\Lib\site-packages\pysparql_anything\sparql_anything.py:30, in SparqlAnything.run(self, **kwargs)
     23 """
     24 The run method replaces the regular command line execution.
     25 @param **kwargs The keyword arguments are the same as the regular
     26     flags for the Sparql Anything CLI, minus the hyphen.
     27 See the User Guide for an example.
     28 """
     29 command = cmd.RunCommand(kwargs, self.receiver)
---> 30 command.execute()

File ~\miniconda3\Lib\site-packages\pysparql_anything\command.py:159, in RunCommand.execute(self)
    157 """ instructions for a run request """
    158 args = super()._combine()
--> 159 self.receiver.main(args)

File ~\miniconda3\Lib\site-packages\pysparql_anything\engine.py:39, in Engine.main(self, args)
     37 def main(self, args):
     38     """ Wrapper for the SPARQL Anything main method. """
---> 39     self.reflection.main(args)

File jnius\\jnius_export_class.pxi:876, in jnius.JavaMethod.__call__()

File jnius\\jnius_export_class.pxi:1059, in jnius.JavaMethod.call_staticmethod()

File jnius\\jnius_utils.pxi:79, in jnius.check_exception()

JavaException: JVM exception occurred: Lexical error at line 1, column 14.  Encountered: <EOF> after prefix "query1.sparql" org.apache.jena.query.QueryParseException

The current version does not allow the user to have any control over the JVM and its parameters (e.g. max heap space).

I believe this is mostly due to how pyjnius works, as the JVM is launched at the import statement.
Thus, when pysparql_anything is imported the JVM is launched when __init__.py runs and imports the namespaces.

I believe we need:

1. Decouple the JVM configuration and launch from the initialisation process and its current location in engine.py.
2. Create either a specific jvm.py module or create specific function in config.py to make visible to clients (code and users)
   the required functionalities.

Thoughts @enridaga ?

As the plan moving forward is to release a new version of PySPARQL Anything for each new SPARQL Anything release, how should we interact with the user with respect to this @enridaga ?

1. Should we notify the user automatically that a new version is available and then let them proceed manually to the update process, or

2. Should we notify the user and possibly give them an option (Y/N) to proceed with the update process automatically?

The SA user guide states that "the argument can be passed multiple times".

But the Python **kwargs are mapped to a dictionary, which does not allow duplicate keys.

A work around such as say v_i and v_o for input and output variables may be needed for passing arguments to the PySPARQL cli.
These can then be normalised in the implementation before being passed to the SA main method.

@enridaga wdyt?

It would be beneficial for developers/maintainers to add in the README.md a guide specifying how the software is built and uploaded to PyPi.

It would be useful if you could directly get an RDFlib Graph object from it. We've been using SA in that way, so we have some code lying around that we could contribute.

Interesting roundtrip required, ideally without saving to file. I think this is related to supporting STDIN in SPARQL Anything, see this issue SPARQL-Anything/sparql.anything#244

I suggest to specify the dependencies here

https://github.com/SPARQL-Anything/PySPARQL-Anything/blob/main/pyproject.toml#L34

more relaxed, in order to use your library more easy in other projects.

See also -> caret notification: https://python-poetry.org/docs/dependency-specification/#caret-requirements

As the SA jar is not installed with pip but subsequently, it is also not uninstalled when running pip uninstall.

It would be a nice additional feature to return the result of a select query as a pd.DataFrame and a dictionary of the bindings.

The title explains it.

Experimenting with the Python build tools I noticed that we can specify an entry point to the library. This in theory would allow the user to for example be able to call the SparqlAnything.run method directly from the command line without running the Python shell, import statements and create an object.

@enridaga Would you like me to look into this further?

To improve the stability of the release of the module the dependencies need to be better determined in the build proess.

The style that I use to write docstrings has changed over time, so that there are multiple styles at once throughout the code at present.

Ideally, the code documentation should be rendered stylistically homogeneous.

On this note, it might be worth settling on a style for the project moving forward, both for the docstrings and the Python code developed.

The official Python documents are 1 and 2 for docstrings and code respectively.

Lately though I have actually been using the Google Python style guide 3, which covers pretty much everything. The only thing I disagree with it is over the linting tool to be used. I find pylint needlessly opinionated in some things and would be inclined more towards flake8 instead.

@enridaga opinions are welcome!