This project provides a small python wrapper for sempr. Be aware that is not feature-complete, and its usage is in many aspects different from the C++ version. What did you expect? This is python! ๐
This project uses pybind11
to create the python extension module:
pip install pybind11
Although a CMakeLists.txt
is present (and you are free to use it,
if you want -- personally, I'm using it during development with my PYTHONPATH
pointed to the build directory), the compilation and installation is handled by
the setup.py
:
python setup.py build
python setup.py install
If it fails to find pybind11 in the process, you'll need to specify its location. In my case, like this:
pybind11_DIR=~/pythonenv/lib/python3.6/site-packages/pybind11/share/cmake/pybind11 python setup.py build
python setup.py install
Now you are good to go!
But wait, there is more: Out of the box, many python IDEs do not support code completion and the like for binary modules. I'm using vscode with pylance and struggled a bit to get it all working, so here is what worked for me in the end:
After having installed semprpy, you will need to create stubs for semprpy
anywhere on your system (does not need to happen inside the semprpy repository).
There are several different methods out there to do this, and what worked in the
end for me was using either mypy
or pybind11-stubgen
.
Note: I struggled a lot to get it all working, but as usual, it is quite simple if you know how. I first got it working with mypy, but noticed that it did not export any documentation, only the signatures. Hence, I gave pybind11-stubgen another try, and that works just fine now! (The only remaining problem being that I did not include much documentation in the python bindings, yet... but that is something else.)
The preferred way right now is to use pybind11-stubgen
, as it was specifically
made to generate stubs from C-extension-modules created with pybind11. It
properly includes documentation and signatures in the stubs.
Note: pybind11-stubgen is being moved into mypy, but at the time of writing this, it still outperforms mypy.
pip install pybind11-stubgen
pybind11-stubgen semprpy
This creates a folder ./stubs
which you can add to your IDE as a location for
stub files. In vscode with pylance, this looks like this for me:
As already mentioned, pybind11-stubgen should give better results. But if you have any trouble with that, you can also use mypy:
pip install mypy
This includes a script called stubgen
which we can use to create stubs for
any module, package or source file. In this case, we use it to create stubs for
the whole semprpy-package:
stubgen -p semprpy
This creates a folder out/semprpy
with your stub files. You can use these in
vscode with pylance by adding the path to the out
folder to the
Python > Analysis: Extra Paths
. For me, this looks like this:
The created python extension is called semprpy
.
import semprpy as sempr
core = sempr.Core('my_database_folder') # or without parameter to skip persistence
core.loadPlugins()
core.addRules('[true() -> (<ex:foo> <ex:bar> <ex:baz>)]')
core.addRules('[EC<Component>(?e ?c) -> (?e <ex:bar> <ex:blub>)]')
for r in core.rules():
print(f'{r.id:2d} : {r.name if r.name else "<unnamed rule>"}')
entity = sempr.Entity()
core.addEntity(entity)
component = sempr.Component() # base class without data, only for testing
entity.addComponent(component)
# as always, inference before querying
core.performInference()
# sparql queries are available, too - and simplified.
# No need to get the RDFPlugin first.
# But: This will throw if the plugin is not loaded!
results = core.query('SELECT * WHERE { ?a <ex:bar> ?b . }')
for r in results:
# r is a dict, with variables used in the query as keys, and a tuple of
# (ValueType, value) as value.
print(f'a: {r["a"][1]:20s} b: {r["b"][1]:10s}')
Outputs something along the lines:
1 : extractTriples
2 : updateSoprano
4 : inferRules
5 : <unnamed rule>
6 : <unnamed rule>
a: sempr:Entity_0 b: ex:blub
a: ex:foo b: ex:baz
Also, take a look at test/test.py
, where I test stuff during development.