Use Python with R with reticulate : : CHEAT SHEET

The reticulate package lets you use Python and R together seamlessly in R code, in R Markdown documents, and in the RStudio IDE.

Python in R Markdown Python in R code

Call Python from R in three ways:
(Optional) Build Python env to use.
IMPORT PYTHON MODULES
Add knitr::knit_engines$set(python =
reticulate::eng_python) to the setup Use import() to import any Python module.
chunk to set up the reticulate Python Access the attributes of a module with $.
engine (not required for knitr >= 1.18).
• import(module, as = NULL, convert =
Suggest the Python environment TRUE, delay_load = FALSE) Import a
to use, in your setup chunk. Python module. If convert = TRUE,
Python objects are converted to
their equivalent R types. Also
Begin Python chunks with ```{python}. import_from_path. import("pandas")
Chunk options like echo, include, etc. all
work as expected. • import_main(convert = TRUE)
Import the main module, where Python
Use the py object to access objects created executes code by default. import_main()
in Python chunks from R chunks.
• import_builtins(convert = TRUE)
Import Python's built-in functions.
Python chunks all execute within a import_builtins()
single Python session so you have access
to all objects created in previous chunks.
SOURCE PYTHON FILES
Use the r object to access objects created
in R chunks from Python chunks. Use source_python() to source a Python script
and make the Python functions and objects it
Output displays below chunk, creates available in the calling R environment.
including matplotlib plots.
• source_python(file, envir = parent.frame(),
convert = TRUE) Run a Python script,
assigning objects to a specified R
Object Conversion Tip: To index Python objects begin at 0, use integers, e.g. 0L Helpers environment. source_python("file.py")

Reticulate provides automatic built-in conversion dict(..., convert = FALSE) Create a Python dictionary py_capture_output(expr, type = c("stdout", RUN PYTHON CODE
between Python and R for many Python types. object. Also py_dict to make a dictionary that uses "stderr")) Capture and return Python output. Also
Python objects as keys. dict(foo = "bar", index = 42L) py_suppress_warnings. py_capture_output("x") Execute Python code into the main Python
R Python module with py_run_file() or py_run_string().
Single-element vector Scalar np_array(data, dtype = NULL, order = "C") Create py_get_attr(x, name, silent = FALSE) Get an
Multi-element vector List NumPy arrays. np_array(c(1:8), dtype = "float16") attribute of a Python object. Also py_set_attr, • py_run_string(code, local = FALSE,
py_has_attr, and py_list_attributes. py_get_attr(x) convert = TRUE) Run Python code
List of multiple types Tuple array_reshape(x, dim, order = c("C", "F")) Reshape (passed as a string) in the main
Named list Dict a Python array. x <- 1:4; array_reshape(x, c(2, 2)) module. py_run_string("x = 10"); py$x
py_help(object) Open the documentation
Matrix/Array NumPy ndarray page for a Python object. py_help(sns)
py_func(object) Wrap an R function in a Python • py_run_file(file, local = FALSE, convert =
Data Frame Pandas DataFrame
function with the same signature. py_func(xor) TRUE) Run Python file in the main
Function Python function py_last_error() Get the last Python module. py_run_file("script.py")
NULL, TRUE, FALSE None, True, False py_main_thread_func(object) Create a function error encountered. Also py_clear_last_error
that will always be called on the main thread. to clear the last error. py_last_error() • py_eval(code, convert = TRUE) Run
Or, if you like, you can convert manually with a Python expression, return the result.
py_to_r(x) Convert a Python object to iterate(..., convert = FALSE) Apply an R function to py_save_object(object, filename, pickle = "pickle") Also py_call. py_eval("1 + 1")
an R object. Also r_to_py. py_to_r(x) each value of a Python iterator or return the values Save and load Python objects with pickle. Also
as an R vector, draining the iterator as you go. Also py_load_object. py_save_object(x, "x.pickle") Access the results, and anything else in Python's
tuple(..., convert = FALSE) Create a iter_next and as_iterator. iterate(iter, print) main module, with py.
Python tuple. tuple("a", "b", "c") with(data, expr, as = NULL, ...) Evaluate an
py_iterator(fn, completed = NULL) Create a Python expression within a Python context manager. • py An R object that contains
iterator from an R function. seq_gen <- function(x){n py <- import_builtins(); with(py$open("output.txt", the Python main module and
<- x; function() {n <<- n + 1; n}};py_iterator(seq_gen(9)) "w") %as% file, { file$write("Hello, there!")}) the results stored there. py$x

RStudio® is a trademark of RStudio, Inc. • CC BY SA RStudio • [email protected] • 844-448-1212 • rstudio.com • Learn more at rstudio.github.io/reticulate/ • reticulate 1.12.0 • Updated: 2019-04
Python in the IDE Requires reticulate plus RStudio v1.2 or higher. Configure Python
Syntax Tab completion for Python Source Execute Python Press F1 over a matplotlib Reticulate binds to a local instance of Python when you first call import() directly
highlighting for functions and objects (and Python code line by line Python symbol plots display or implicitly from an R session. To control the process, find or build your desired
Python scripts Python modules imported scripts. with Cmd + Enter to display the in plots pane. Python instance. Then suggest your instance to reticulate. Restart R to unbind.
and chunks in R scripts) (Ctrl + Enter) help topic for

Find Python
that symbol.

• py_discover_config() Return all detected • virtualenv_list() List all available virtualenvs.

versions of Python. Use py_config to check Also virtualenv_root(). virtualenv_list()
which version has been loaded. py_config()
• conda_list(conda = "auto") List all
• py_available(initialize = FALSE) Check if available conda envs. Also conda_binary()
Python is available on your system. Also and conda_version(). conda_list()
py_module_available, py_numpy_module.
py_available()

Create a Python env Suggest an env to use

• virtualenv_create(envname) Create a new To choose an instance of Python to bind to,
virtualenv. virtualenv_create("r-pandas") reticulate scans the instances on your computer in
the following order, stopping at the first instance
• conda_create(envname, packages = NULL, that contains the module called by import().
conda = "auto") Create a new Conda env.
conda_create("r-pandas", packages = 1. The instance referenced by the environment
"pandas") variable RETICULATE_PYTHON (if specified).
Tip: set in .Renviron file.
• Sys.setenv(RETICULATE_PYTHON = PATH)
Install Packages Set default Python binary. Persists across
sessions! Undo with Sys.unsetenv.
Sys.setenv(RETICULATE_PYTHON =
Install Python packages with R (below) or the shell: "/usr/local/bin/python")
pip install SciPy
conda install SciPy 2. The instances referenced by use_ functions
if called before import(). Will fail silently if
• py_install(packages, envname = "r- called after import unless required = TRUE.
A Python REPL opens in the console when you run Python code with a keyboard shortcut. Type exit to close. reticulate", method = c("auto", "virtualenv",
"conda"), conda = "auto", ...) Installs Python • use_python(python, required = FALSE)
packages into a Python env named Suggest a Python binary to use by path.
Python REPL "r-reticulate". py_install("pandas") use_python("/usr/local/bin/python")

A REPL (Read, Eval, Print Loop) is a command • virtualenv_install(envname, packages, • use_virtualenv(virtualenv = NULL,
line where you can run Python code and view ignore_installed = FALSE) Install a package required = FALSE) Suggest a Python
the results. within a virtualenv. virtualenv_install( virtualenv. use_virtualenv("~/myenv")
"r-pandas", packages = "pandas")
1. Open in the console with repl_python(), • use_condaenv(condaenv = NULL,
or by running code in a Python script with • virtualenv_remove(envname, packages = conda = "auto", required = FALSE)
Cmd + Enter (Ctrl + Enter) . NULL, confirm = interactive()) Remove Suggest a Conda env to use.
individual packages or an entire virtualenv. use_condaenv(condaenv = "r-nlp",
• repl_python(module = NULL, quiet = virtualenv_remove("r-pandas", packages = conda = "/opt/anaconda3/bin/conda")
getOption("reticulate.repl.quiet", "pandas")
default = FALSE)) Launch a Python 3. Within virtualenvs and conda envs that carry
REPL. Run exit to close. repl_python() • conda_install(envname, packages, forge = the same name as the imported module.
TRUE, pip = FALSE, pip_ignore_installed = e.g. ~/anaconda/envs/nltk for import("nltk")
2. Type commands at >>> prompt TRUE, conda = "auto") Install a package
within a Conda env. conda_install(
3. Press Enter to run code "r-pandas", packages = "plotly") 4. At the location of the Python binary
discovered on the system PATH
4. Type exit to close and return to R console • conda_remove(envname, packages = NULL, (i.e. Sys.which("python"))
conda = "auto") Remove individual packages
or an entire Conda env. conda_remove( 5. At customary locations for Python, e.g./usr/
"r-pandas", packages = "plotly") local/bin/python, /opt/local/bin/python...

RStudio® is a trademark of RStudio, Inc. • CC BY SA RStudio • [email protected] • 844-448-1212 • rstudio.com • Learn more at rstudio.github.io/reticulate/ • reticulate 1.12.0 • Updated: 2019-04