Thanks for your interest in contributing to spaCy 🎉 This page will give you a quick overview of how things are organized and most importantly, how to get involved.
- Issues and bug reports
- Contributing to the code base
- Code conventions
- Adding tests
- Updating the website
- Publishing extensions and plugins
- Code of conduct
First, do a quick search to see if the issue has already been reported. If so, it's often better to just leave a comment on an existing issue, rather than creating a new one. Old issues also often include helpful tips and solutions to common problems. You should also check the troubleshooting guide to see if your problem is already listed there.
If you're looking for help with your code, consider posting a question on the GitHub Discussions board or Stack Overflow. Please understand that we won't be able to provide individual support via email. We also believe that help is much more valuable if it's shared publicly, so that more people can benefit from it.
When opening an issue, use a descriptive title and include your environment (operating system, Python version, spaCy version). Our issue templates help you remember the most important details to include. If you've discovered a bug, you can also submit a regression test straight away. When you're opening an issue to report the bug, simply refer to your pull request in the issue body. A few more tips:
-
Describing your issue: Try to provide as many details as possible. What exactly goes wrong? How is it failing? Is there an error? "XY doesn't work" usually isn't that helpful for tracking down problems. Always remember to include the code you ran and if possible, extract only the relevant parts and don't just dump your entire script. This will make it easier for us to reproduce the error.
-
Getting info about your spaCy installation and environment: You can use the command line interface to print details and even format them as Markdown to copy-paste into GitHub issues:
python -m spacy info --markdown. -
Checking the model compatibility: If you're having problems with a statistical model, it may be because the model is incompatible with your spaCy installation. In spaCy v2.0+, you can check this on the command line by running
python -m spacy validate. -
Sharing a model's output, like dependencies and entities: spaCy comes with built-in visualizers that you can run from within your script or a Jupyter notebook. For some issues, it's helpful to include a screenshot of the visualization. You can simply drag and drop the image into GitHub's editor and it will be uploaded and included.
-
Sharing long blocks of code or logs: If you need to include long code, logs or tracebacks, you can wrap them in
<details>and</details>. This collapses the content so it only becomes visible on click, making the issue easier to read and follow.
See this page for an overview of the system we use to tag our issues and pull requests.
You don't have to be an NLP expert or Python pro to contribute, and we're happy
to help you get started. If you're new to spaCy, a good place to start is the
spaCy 101 guide and the
help wanted (easy)
label, which we use to tag bugs and feature requests that are easy and
self-contained. If you've decided to take on one of these problems and you're
making good progress, don't forget to add a quick comment to the issue. You can
also use the issue to ask questions, or share your work in progress.
Every library has a different inclusion philosophy — a policy of what should be shipped in the core library, and what could be provided in other packages. Our philosophy is to prefer a smaller core library. We generally ask the following questions:
-
What would this feature look like if implemented in a separate package? Some features would be very difficult to implement externally – for example, changes to spaCy's built-in methods. In contrast, a library of word alignment functions could easily live as a separate package that depended on spaCy — there's little difference between writing
import word_alignerandimport spacy.word_aligner. spaCy makes it easy to implement custom pipeline components, and add your own attributes, properties and methods to theDoc,TokenandSpan. If you're looking to implement a new spaCy feature, starting with a custom component package is usually the best strategy. You won't have to worry about spaCy's internals and you can test your module in an isolated environment. And if it works well, we can always integrate it into the core library later. -
Would the feature be easier to implement if it relied on "heavy" dependencies spaCy doesn't currently require? Python has a very rich ecosystem. Libraries like PyTorch, TensorFlow, scikit-learn, SciPy or Gensim do lots of useful things — but we don't want to have them as default dependencies. If the feature requires functionality in one of these libraries, it's probably better to break it out into a different package.
-
Is the feature orthogonal to the current spaCy functionality, or overlapping? spaCy strongly prefers to avoid having 6 different ways of doing the same thing. As better techniques are developed, we prefer to drop support for "the old way". However, it's rare that one approach entirely dominates another. It's very common that there's still a use-case for the "obsolete" approach. For instance, WordNet is still very useful — but word vectors are better for most use-cases, and the two approaches to lexical semantics do a lot of the same things. spaCy therefore only supports word vectors, and support for WordNet is currently left for other packages.
-
Do you need the feature to get basic things done? We do want spaCy to be at least somewhat self-contained. If we keep needing some feature in our recipes, that does provide some argument for bringing it "in house".
To make changes to spaCy's code base, you need to fork then clone the GitHub repository and build spaCy from source. You'll need to make sure that you have a development environment consisting of a Python distribution including header files, a compiler, pip, virtualenv and git installed. The compiler is usually the trickiest part.
If you've made changes to .pyx files, you need to recompile spaCy before you
can test your changes by re-running python setup.py build_ext --inplace.
Changes to .py files will be effective immediately.
📖 For more details and instructions, see the documentation on compiling spaCy from source and the quickstart widget to get the right commands for your platform and Python version.
When fixing a bug, first create an issue if one does not already exist. The description text can be very short – we don't want to make this too bureaucratic.
Next, add a test to the relevant file in the
spacy/testsfolder. Then add a pytest
mark,
@pytest.mark.issue(NUMBER), to reference the issue number.
# Assume you're fixing Issue #1234
@pytest.mark.issue(1234)
def test_issue1234():
...Test for the bug you're fixing, and make sure the test fails. Next, add and commit your test file. Finally, fix the bug, make sure your test passes and reference the issue number in your pull request description.
📖 For more information on how to add tests, check out the tests README.
Code should loosely follow pep8.
spaCy uses black for code
formatting and flake8 for linting its
Python modules. If you've built spaCy from source, you'll already have both
tools installed.
As a general rule of thumb, we use f-strings for any formatting of strings.
One exception are calls to Python's logging functionality.
To avoid unnecessary string conversions in these cases, we use string formatting
templates with %s and %d etc.
.py files, not Cython modules in .pyx and .pxd files.
After cloning the repo, after installing the packages from requirements.txt, enter the repo folder and run pre-commit install.
Each time a git commit is initiated, black and flake8 will run automatically on the modified files only.
In case of error, or when black modified a file, the modified file needs to be git add once again and a new
git commit has to be issued.
black is an opinionated Python code
formatter, optimized to produce readable code and small diffs. You can run
black from the command-line, or via your code editor. For example, if you're
using Visual Studio Code, you can add the
following to your settings.json to use black for formatting and auto-format
your files on save:
{
"python.formatting.provider": "black",
"[python]": {
"editor.formatOnSave": true
}
}See here for the full list of available editor integrations.
There are a few cases where auto-formatting doesn't improve readability – for
example, in some of the language data files or in the tests that construct Doc objects from lists of words and other labels.
Wrapping a block in # fmt: off and # fmt: on lets you disable formatting
for that particular code. Here's an example:
# fmt: off
text = "I look forward to using Thingamajig. I've been told it will make my life easier..."
heads = [1, 1, 1, 1, 3, 4, 1, 6, 11, 11, 11, 11, 14, 14, 11, 16, 17, 14, 11]
deps = ["nsubj", "ROOT", "advmod", "prep", "pcomp", "dobj", "punct", "",
"nsubjpass", "aux", "auxpass", "ROOT", "nsubj", "aux", "ccomp",
"poss", "nsubj", "ccomp", "punct"]
# fmt: onflake8 is a tool for enforcing code
style. It scans one or more files and outputs errors and warnings. This feedback
can help you stick to general standards and conventions, and can be very useful
for spotting potential mistakes and inconsistencies in your code. The most
important things to watch out for are syntax errors and undefined names, but you
also want to keep an eye on unused declared variables or repeated
(i.e. overwritten) dictionary keys. If your code was formatted with black
(see above), you shouldn't see any formatting-related warnings.
The flake8 section in setup.cfg defines the configuration we use for this
codebase. For example, we're not super strict about the line length, and we're
excluding very large files like lemmatization and tokenizer exception tables.
Ideally, running the following command from within the repo directory should not return any errors or warnings:
flake8 spacySometimes, you explicitly want to write code that's not compatible with our
rules. For example, a module's __init__.py might import a function so other
modules can import it from there, but flake8 will complain about an unused
import. And although it's generally discouraged, there might be cases where it
makes sense to use a bare except.
To ignore a given line, you can add a comment like # noqa: F401, specifying
the code of the error or warning we want to ignore. It's also possible to
ignore several comma-separated codes at once, e.g. # noqa: E731,E123. Here
are some examples:
# The imported class isn't used in this file, but imported here, so it can be
# imported *from* here by another module.
from .submodule import SomeClass # noqa: F401
try:
do_something()
except: # noqa: E722
# This bare except is justified, for some specific reason
do_something_else()All Python code must be written compatible with Python 3.6+. More detailed code conventions can be found in the developer docs.
Code that interacts with the file-system should accept objects that follow the
pathlib.Path API, without assuming that the object inherits from pathlib.Path.
If the function is user-facing and takes a path as an argument, it should check
whether the path is provided as a string. Strings should be converted to
pathlib.Path objects. Serialization and deserialization functions should always
accept file-like objects, as it makes the library IO-agnostic. Working on
buffers makes the code more general, easier to test, and compatible with Python
3's asynchronous IO.
Although spaCy uses a lot of classes, inheritance is viewed with some suspicion — it's seen as a mechanism of last resort. You should discuss plans to extend the class hierarchy before implementing.
We have a number of conventions around variable naming that are still being
documented, and aren't 100% strict. A general policy is that instances of the
class Doc should by default be called doc, Token → token, Lexeme → lex,
Vocab → vocab and Language → nlp. You should avoid naming variables that are
of other types these names. For instance, don't name a text string doc — you
should usually call this text. Two general code style preferences further help
with naming. First, lean away from introducing temporary variables, as these
clutter your namespace. This is one reason why comprehension expressions are
often preferred. Second, keep your functions shortish, so they can work in a
smaller scope. Of course, this is a question of trade-offs.
spaCy's core data structures are implemented as Cython cdef
classes. Memory is managed through the cymem.cymem.Pool class, which allows
you to allocate memory which will be freed when the Pool object is garbage
collected. This means you usually don't have to worry about freeing memory. You
just have to decide which Python object owns the memory, and make it own the
Pool. When that object goes out of scope, the memory will be freed. You do
have to take care that no pointers outlive the object that owns them — but this
is generally quite easy.
All Cython modules should have the # cython: infer_types=True compiler
directive at the top of the file. This makes the code much cleaner, as it avoids
the need for many type declarations. If possible, you should prefer to declare
your functions nogil, even if you don't especially care about multi-threading.
The reason is that nogil functions help the Cython compiler reason about your
code quite a lot — you're telling the compiler that no Python dynamics are
possible. This lets many errors be raised, and ensures your function will run
at C speed.
Cython gives you many choices of sequences: you could have a Python list, a numpy array, a memory view, a C++ vector, or a pointer. Pointers are preferred, because they are fastest, have the most explicit semantics, and let the compiler check your code more strictly. C++ vectors are also great — but you should only use them internally in functions. It's less friendly to accept a vector as an argument, because that asks the user to do much more work.
Here's how to get a pointer from a numpy array, memory view or vector:
cdef void get_pointers(np.ndarray[int, mode='c'] numpy_array, vector[int] cpp_vector, int[::1] memory_view) nogil:
pointer1 = <int*>numpy_array.data
pointer2 = cpp_vector.data()
pointer3 = &memory_view[0]Both C arrays and C++ vectors reassure the compiler that no Python operations are possible on your variable. This is a big advantage: it lets the Cython compiler raise many more errors for you.
When getting a pointer from a numpy array or memoryview, take care that the data is actually stored in C-contiguous order — otherwise you'll get a pointer to nonsense. The type-declarations in the code above should generate runtime errors if buffers with incorrect memory layouts are passed in.
To iterate over the array, the following style is preferred:
cdef int c_total(const int* int_array, int length) nogil:
total = 0
for item in int_array[:length]:
total += item
return totalIf this is confusing, consider that the compiler couldn't deal with
for item in int_array: — there's no length attached to a raw pointer, so how
could we figure out where to stop? The length is provided in the slice notation
as a solution to this. Note that we don't have to declare the type of item in
the code above — the compiler can easily infer it. This gives us tidy code that
looks quite like Python, but is exactly as fast as C — because we've made sure
the compilation to C is trivial.
Your functions cannot be declared nogil if they need to create Python objects
or call Python functions. This is perfectly okay — you shouldn't torture your
code just to get nogil functions. However, if your function isn't nogil, you
should compile your module with cython -a --cplus my_module.pyx and open the
resulting my_module.html file in a browser. This will let you see how Cython
is compiling your code. Calls into the Python run-time will be in bright yellow.
This lets you easily see whether Cython is able to correctly type your code, or
whether there are unexpected problems.
Finally, if you're new to Cython, you should expect to find the first steps a
bit frustrating. It's a very large language, since it's essentially a superset
of Python and C++, with additional complexity and syntax from numpy. The
documentation isn't great, and there are
many "traps for new players". Working in Cython is very rewarding once you're
over the initial learning curve. As with C and C++, the first way you write
something in Cython will often be the performance-optimal approach. In contrast,
Python optimization generally requires a lot of experimentation. Is it faster to
have an if item in my_dict check, or to use .get()? What about try/except?
Does this numpy operation create a copy? There's no way to guess the answers to
these questions, and you'll usually be dissatisfied with your results — so
there's no way to know when to stop this process. In the worst case, you'll make
a mess that invites the next reader to try their luck too. This is like one of
those volcanic gas-traps,
where the rescuers keep passing out from low oxygen, causing another rescuer to
follow — only to succumb themselves. In short, just say no to optimizing your
Python. If it's not fast enough the first time, just switch to Cython.
- PEP 8 Style Guide for Python Code (python.org)
- Official Cython documentation (cython.org)
- Writing C in Cython (explosion.ai)
- Multi-threading spaCy’s parser and named entity recognizer (explosion.ai)
spaCy uses the pytest framework for testing. For more
info on this, see the pytest documentation.
Tests for spaCy modules and classes live in their own directories of the same
name. For example, tests for the Tokenizer can be found in
/spacy/tests/tokenizer. To be interpreted and run,
all test files and test functions need to be prefixed with test_.
When adding tests, make sure to use descriptive names, keep the code short and
concise and only test for one behavior at a time. Try to parametrize test
cases wherever possible, use our pre-defined fixtures for spaCy components and
avoid unnecessary imports. Extensive tests that take a long time should be marked with @pytest.mark.slow.
📖 For more guidelines and information on how to add tests, check out the tests README.
For instructions on how to build and run the website locally see Setup and installation in the website directory's README.
The docs can always use another example or more detail, and they should always be up to date and not misleading. To quickly find the correct file to edit, simply click on the "Suggest edits" button at the bottom of a page.
📖 For more info and troubleshooting guides, check out the website README.
We're very excited about all the new possibilities for community extensions and plugins in spaCy v3.0, and we can't wait to see what you build with it!
-
An extension or plugin should add substantial functionality, be well-documented and open-source. It should be available for users to download and install as a Python package – for example via PyPi.
-
Extensions that write to
Doc,TokenorSpanattributes should be wrapped as pipeline components that users can add to their processing pipeline usingnlp.add_pipe(). -
When publishing your extension on GitHub, tag it with the topics
spacyandspacy-extensionsto make it easier to find. Those are also the topics we're linking to from the spaCy website. If you're sharing your project on Twitter, feel free to tag @spacy_io so we can check it out. -
Once your extension is published, you can open a PR to suggest it for the Universe page.
📖 For more tips and best practices, see the checklist for developing spaCy extensions.
spaCy adheres to the Contributor Covenant Code of Conduct. By participating, you are expected to uphold this code.