Python interoperability

This chapter describes how to interact with Python code from Hy code and vice versa.

Mangling

Mangling allows variable names to be spelled differently in Hy and Python. For example, Python's str.format_map can be written str.format-map in Hy, and a Hy function named valid? would be called hyx_valid_Xquestion_markX in Python. You can call hy.mangle and hy.unmangle from either language.

Keyword mincing

Another kind of mangling may be necessary in Python to refer to variables with the same name as reserved words. For example, while (setv break 13) is legal Hy, import hy, my_hy_module; print(my_hy_module.break) is syntactically invalid Python. String literals work, as in getattr(my_hy_module, "break"), but to use what is syntactically a Python identifier, you'll have to take advantage of Python's Unicode normalization (via NFKC) and write something like my_hy_module.𝐛reak. Here are all the MATHEMATICAL BOLD SMALL letters (U+1D41A through U+1D433) for convenient copying:

𝐚𝐛𝐜𝐝𝐞𝐟𝐠𝐡𝐢𝐣𝐤𝐥𝐦𝐧𝐨𝐩𝐪𝐫𝐬𝐭𝐮𝐯𝐰𝐱𝐲𝐳

Libraries that expect Python

There are various means by which Hy may interact poorly with a Python library because the library doesn't account for the possibility of Hy. For example, when you run the program hy, sys.executable will be set to this program rather than the original Python binary. This is helpful more often than not, but will lead to trouble if e.g. the library tries to call sys.executable with the -c option. In this case, you can try setting sys.executable back to hy.sys-executable, which is a saved copy of the original value. More generally, you can use hy2py, or you can put a simple Python wrapper script like import hy, my_hy_program in front of your code.

See the wiki for tips on using specific packages.

Packaging a Hy library

Generally, the same infrastructure used for Python packages, such as setup.py files and the Python Package Index (PyPI), is applicable to Hy. Don't write the setup file itself in Hy, since you'll be declaring your package's dependence on Hy there, likely in the install_requires argument of setup. Similarly, at the top level of the package, use __init__.py rather than __init__.hy, and begin it with import hy to set up the import hooks for Hy. You can still import a Hy file from there in order to write the real logic in Hy. If you want allow users to import or require from the top level of your module, as in from my_module import my_function or (require my-module [my-macro]), use an __init__.py such as

import hy
from my_module.hy_init import *
hy.eval(hy.read('(require my-module.hy-init :macros * :readers *)'))

If you want to compile your Hy code into Python bytecode at installation-time (in case e.g. the code is being installed to a directory where the bytecode can't be automatically written later, due to permissions issues), see Hy's own setup.py for an example.

For PyPI packages, use the trove classifier Programming Language :: Hy for libraries meant to be useful for Hy specifically (e.g., a library that provides Hy macros) or other projects that are about Hy somehow (e.g., an instructive example Hy program). Don't use it for a package that just happens to be written in Hy.

Using Python from Hy

To use a Python module from Hy, just import it. In most cases, no additional ceremony is required.

You can embed Python code directly into a Hy program with the macros py and pys, and you can use standard Python tools like eval() or exec() to execute or manipulate Python code in strings.

Using Hy from Python

To use a Hy module from Python, you can just import it, provided that hy has already been imported first, whether in the current module or in some earlier module executed by the current Python process. As mentioned previously, you can put import hy in a package's __init__.py to make this happen automatically.

You can use hy2py to convert a Hy program to Python. The output will still import hy, and thus require Hy to be installed in order to run; see Implicit names for details and workarounds.

To execute Hy code from a string, use hy.read-many to convert it to models and hy.eval to evaluate it:

>>> hy.eval(hy.read_many("(setv x 1) (+ x 1)"))
2

There is no Hy equivalent of exec() because hy.eval works even when the input isn't equivalent to a single Python expression.

You can use hy.REPL.run() to launch the Hy REPL from Python, as in hy.REPL(locals = {**globals(), **locals()}).run().