Semvecpy is a repository for Semantic Vectors work in Python (version 3 and above).
It is a research project. It includes some well-tested production-grade work, but it's up to users to be aware of which parts this is.
Clone the repository, or install using pip install git+https://github.com/semanticvectors/semvecpy
.
TODO: It would be nice to make this just pip install semvecpy
. See https://packaging.python.org/tutorials/packaging-projects/.
Import paths are designed to start with semvecpy
in the hope that this will avoid name clashes from common terms such
as vectors
. So after installation, the following should work:
>>> from semvecpy.vectors import vector_utils as vu
>>> vu.normalize([3, 4])
array([0.6, 0.8])
- Use Python 3+ throughout.
- Please don't try to get
print "Hello World"
working instead ofprint("Hello World")
.
- Please don't try to get
- Imports should work relative to the project ./semvecpy directory.
- E.g.,
from semvecpy.vectors import vector_utils
, rather than justfrom vectors import vector_utils
. - Please don't add deeper dependencies to your PYTHONPATH or set directories as sources roots in an IDE to make import statements shorter, this may break other internal and external developers.
- It's fine to use
from . import my_module
in tests, though fully qualified dependencies are encouraged for readability in module source files.
- E.g.,
- Module names are preferred with underscore_separators, but there's no firm rule in place.
- Tests for module
.../dir/foo.py
are in.../dir/foo_test.y
.- This is one of the standard patterns, and it makes it particularly easy to see which modules already have dedicated tests, and whether these files should be moved / renamed if the modules they're testing are renamed.
pytest .
should run and pass all tests from thesemvecpy
root directory.
Feel free to clone, fork, and play with this repository. When it comes to merging code, feel free to submit merge requests, and for frequent contributors, ask the Admins for Maintain permissions. We tend to say yes.
Before sending merge requests, please do check that pytest .
runs and passes all tests.
Feel free to ask for help if it doesn't.
Since it's a research project, we encourage work-in-progress and experimental code. This does sometimes lead to duplicated functionality, some variation in naming and code styles, etc.
To make things work together, code may be refactored. It's nice to discuss this with original authors, but for small changes, developers with suitable permissions are encouraged to just go ahead. It follows that some changes may be unexpected to some authors. If it's a bother just sync your repository wherever you left off, feel free to fork, and we can discuss how to merge things back together later. Basically DON'T WORRY.