Need help with libpython-clj?
Click the “chat” button below for chat support from the developer who created it, or find similar developers for support.

About the developer

750 Stars 48 Forks Eclipse Public License 2.0 576 Commits 14 Opened issues


Python bindings for Clojure

Services available


Need anything else?

Contributors list

Deep Clojure/Python Integration

Version Info

Clojars Project travis integration

## New Versions Are HOT! Huge New Features! You Can't Afford To Miss Out!

  • JDK-17 and Mac M-1 support. To use libpython-clj2 with jdk-17 you need to enable the foreign module - see deps.edn for a working alias.
  • You can now Embed Clojure in Python - you can launch a Clojure REPL from a Python host process.
  • Main namespace is now
    -- docs have not been updated.
  • 32 bit support!!
  • Partial support for JDK-16 - The tests pass for core Python functionality but numpy failed to load - the JDK-16 dynamic library loading path is different than JNA and of course this causes issues.
  • 20-30% better performance.
  • - Note - Please avoid deprecated versions such as
    [cnuernber/libpython-clj "1.36"]
    (note name change).
  • API Documentation
  • This library, which has received the efforts of many excellent people, is built mainly upon cnuernber/dtype-next and the JNA library.
  • Static code generation - generate clojure namespaces wrapping python modules that are safe to use with AOT and load much faster than analogous
    calls. These namespace will not automatically initialize the python subsystem -- initialize! must be called first (or a nice exception is throw).

libpython-clj features

  • Bridge between JVM objects and Python objects easily; use Python in your Java and use some Java in your Python.
  • Python objects are linked to the JVM GC such that when they are no longer reachable from the JVM their references are released. Scope based resource contexts are also available.
  • Finding the python libraries is done dynamically allowing one system to run on multiple versions of python.
  • REPL oriented design means fast, smooth, iterative development.
  • Carin Meier has written excellent posts on plotting and advanced text generation. She also has some great examples.


We aim to integrate Python into Clojure at a deep level. This means that we want to be able to load/use python modules almost as if they were Clojure namespaces. We also want to be able to use Clojure to extend Python objects. I gave a talk at Clojure Conj 2019 that outlines more of what is going on.

This code is a concrete example that generates an embedding for faces:

(ns facial-rec.face-feature
  (:require [libpython-clj2.require :refer [require-python]]
            [libpython-clj2.python :refer [py. py.. py.-] :as py]
            [tech.v3.datatype :as dtype]))

(require-python 'mxnet '(mxnet ndarray module io model)) (require-python 'cv2) (require-python '[numpy :as np])

(defn load-model [& {:keys [model-path checkpoint] :or {model-path "models/recognition/model" checkpoint 0}}] (let [[sym arg-params aux-params] (mxnet.model/load_checkpoint model-path checkpoint) all-layers (py. sym get_internals) target-layer (py/get-item all-layers "fc1_output") model (mxnet.module/Module :symbol target-layer :context (mxnet/cpu) :label_names nil)] (py. model bind :data_shapes [["data" [1 3 112 112]]]) (py. model set_params arg-params aux-params) model))

(defonce model (load-model))

(defn face->feature [img-path] (py/with-gil-stack-rc-context (if-let [new-img (cv2/imread img-path)] (let [new-img (cv2/cvtColor new-img cv2/COLOR_BGR2RGB) new-img (np/transpose new-img [2 0 1]) input-blob (np/expand_dims new-img :axis 0) data (mxnet.ndarray/array input-blob) batch ( :data [data])] (py. model forward batch :is_train false) (-> (py. model get_outputs) first (py. asnumpy) (#(dtype/make-container :java-array :float32 %)))) (throw (Exception. (format "Failed to load img: %s" img-path))))))


Config namespace

(ns my-py-clj.config
  (:require [libpython-clj2.python :as py]))

;; When you use conda, it should look like this. (py/initialize! :python-executable "/opt/anaconda3/envs/my_env/bin/python3.7" :library-path "/opt/anaconda3/envs/my_env/lib/libpython3.7m.dylib")

Update project.clj

  ;; This namespace going to run when the REPL is up.
  :repl-options {:init-ns my-py-clj.config}
user> (require '[libpython-clj2.require :refer [require-python]])
...logging info....
user> (require-python '[numpy :as np])
user> (def test-ary (np/array [[1 2][3 4]]))
user> test-ary
[[1 2]
 [3 4]]

We have a document on all the features but beginning usage is pretty simple. Import your modules, use the things from Clojure. We have put effort into making sure things like sequences and ranges transfer between the two languages.


One very complimentary aspect of Python with respect to Clojure is it's integration with cutting edge native libraries. Our support isn't perfect so some understanding of the mechanism is important to diagnose errors and issues.

Current, we launch the python3 executable and print out various different bits of configuration as json. We parse the json and use the output to attempt to find the
shared library so for example if we are loading python 3.6 we look for
on Linux or
on the Mac.

This pathway has allowed us support Conda albeit with some work. For examples using Conda, check out the facial rec repository a)bove or look into how we build our test docker containers.


We like to talk about libpython-clj on Zulip as the conversations are persistent and searchable.

Further Information

New To Clojure

New to Clojure or the JVM? Try remixing the nextjournal entry and playing around there. For more resources on learning and getting more comfortable with Clojure, we have an introductory document.


To install jar to local .m2 :

$ lein install

Deploy to clojars

$ lein deploy clojars

This command will sign jar before deploy, using your gpg key. (see dev/src/build.clj for signing options)


Copyright © 2019 Chris Nuernberger

This program and the accompanying materials are made available under the terms of the Eclipse Public License 2.0 which is available at

We use cookies. If you continue to browse the site, you agree to the use of cookies. For more information on our use of cookies please see our Privacy Policy.