Package 'rpymat'

Title: Easy to Configure an Isolated 'Python' Environment
Description: Aims to create a single isolated 'Miniconda' and 'Python' environment for reproducible pipeline scripts. The package provides utilities to run system command within the 'conda' environment, making it easy to install, launch, manage, and stop 'Jupyter-lab'.
Authors: Zhengjia Wang [cph, aut, cre]
Maintainer: Zhengjia Wang <[email protected]>
License: Apache License (>= 2)
Version: 0.1.7
Built: 2024-11-13 03:43:10 UTC
Source: https://github.com/dipterix/rpymat

Help Index


Choose file or directory to open via 'Python'

Description

Choose a directory, one or multiple files to open, or choose a file to save.

Usage

choose_fileopen(
  initialfile = NULL,
  multiple = FALSE,
  title = ifelse(multiple, "Choose Files", "Choose a File"),
  message = "",
  verbose = FALSE,
  force = FALSE
)

choose_filesave()

choose_directory(
  initialdir = NULL,
  title = "Choose a Directory",
  message = "",
  verbose = FALSE
)

Arguments

initialfile, initialdir

initial selection of file or directory

multiple

whether to open multiple files

title, message

dialogue title and message

verbose

whether to verbose debug information

force

whether to force using 'Python' when native R functions are available, default is false

Details

Base-R has file.choose function to choose files. However, users cannot select multiple files nor directories. These functions fill the gap by using 'Python' 'tkinter' package. Please make sure that one-time setup function configure_conda has executed before running these functions.

The functions must run as interactive mode. If you run the functions on a server, most likely you will get nothing. The functions themselves do not check if you are running under interactive sessions. You must check by yourself.

Value

User-selected paths. If the users select nothing, then NULL will be returned. For multiple file selection, multiple paths will be returned.

Examples

if(interactive()) {
  choose_fileopen(multiple = TRUE)
}

'Miniconda' environment

Description

These functions/variables are used to configure 'Miniconda' environment.

Usage

CONDAENV_NAME(env_name)

conda_path()

conda_bin()

env_path()

list_pkgs(...)

configure_matlab(matlab, python_ver = "auto")

configure_conda(
  python_ver = "auto",
  packages = NULL,
  matlab = NULL,
  update = FALSE,
  force = FALSE,
  standalone = FALSE
)

remove_conda(ask = TRUE)

add_packages(packages = NULL, python_ver = "auto", ...)

ensure_rpymat(verbose = TRUE, cache = TRUE)

matlab_engine()

call_matlab(
  fun,
  ...,
  .options = getOption("rpymat.matlab_opt", "-nodesktop -nojvm"),
  .debug = getOption("rpymat.debug", FALSE)
)

Arguments

env_name

alternative environment name to use; default is "rpymat-conda-env"

...

for add_packages, these are additional parameters passing to conda_install; for call_matlab, ... are the parameters passing to fun

matlab

'Matlab' path to add to the configuration path; see 'Details'

python_ver

python version to use; see 'Configuration'

packages

additional python or conda packages to install

update

whether to update conda; default is false

force

whether to force install the 'Miniconda' even a previous version exists; default is false. Setting false=TRUE rarely works. Please see 'Configuration'.

standalone

whether to install conda regardless of existing conda environment

ask

whether to ask for user's agreement to remove the repository. This parameter should be true if your functions depend on remove_conda (see 'CRAN Repository Policy'). This argument might be removed and force to be interactive in the future.

verbose

whether to print messages

cache

whether to use cached configurations; default is true

fun

'Matlab' function name, character (experimental)

.options

'Matlab' compiler options

.debug

whether to enable debug mode

Value

None

Background & Objectives

Package reticulate provides sophisticated tool-sets that allow us to call python functions within R. However, the installation of 'Miniconda' and python can be tricky on many platforms, for example, the 'M1' chip, or some other 'ARM' machines. The package rpymat provides easier approach to configure on these machines with totally isolated environments. Any modifications to this environment will not affect your other set ups.

Since 2014, 'Matlab' has introduced its official compiler for python. The package rpymat provides a simple approach to link the compiler, provided that you have proper versions of 'Matlab' installed. Here is a list of 'Matlab' versions with official compilers and their corresponding python versions.

Configuration

If 'Matlab' compiler is not to be installed, In most of the cases, function configure_conda with default arguments automatically downloads the latest 'Miniconda' and configures the latest python. If any other versions of 'Miniconda' is ought to be installed, please set options "reticulate.miniconda.url" to change the source location.

If 'Matlab' is to be installed, please specify the 'Matlab' path when running configure_conda. If the environment has been setup, configure_matlab can link the 'Matlab' compilers without removing the existing environment. For 'ARM' users, unfortunately, there will be no 'Matlab' support as the compilers are written for the 'Intel' chips.

Initialization

Once conda and python environment has been installed, make sure you run ensure_rpymat() before running any python code. This function will make sure correct compiler is linked to your current R session.

Examples

# The script will interactively install \code{conda} to `R_user_dir`
## Not run: 

# Install conda and python 3.9

configure_conda(python_ver = '3.9')


# Add packages h5py, pandas, jupyter

add_packages(c('h5py', 'pandas', 'jupyter'))

# Add pip packages

add_packages("itk", pip = TRUE)

# Initialize the isolated environment

ensure_rpymat()


# Remove the environment

remove_conda()


## End(Not run)

Install, register, launch 'Jupyter' notebook to the virtual environment

Description

Install, register, launch 'Jupyter' notebook to the virtual environment

Usage

add_jupyter(..., register_R = TRUE)

jupyter_bin()

jupyter_register_R(
  user = NULL,
  name = "ir",
  displayname = "R",
  rprofile = NULL,
  prefix = NULL,
  sys_prefix = NULL,
  verbose = getOption("verbose")
)

jupyter_options(
  root_dir,
  host = "127.0.0.1",
  port = 8888,
  open_browser = FALSE,
  token = rand_string(),
  base_url = "/jupyter/"
)

jupyter_launch(
  host = "127.0.0.1",
  port = 8888,
  open_browser = TRUE,
  workdir = getwd(),
  async = FALSE,
  ...,
  dry_run = FALSE
)

jupyter_check_launch(
  port = 8888,
  host = "127.0.0.1",
  open_browser = TRUE,
  workdir = getwd(),
  async = "auto",
  ...
)

jupyter_server_list()

jupyter_server_stop(port, ...)

jupyter_server_stop_all(...)

Arguments

...

for add_jupyter, these are additional parameters passed to jupyter_register_R; for jupyter_launch, these are additional parameters passed to jupyter_options

register_R

whether to register IRkernel to the notebook

user, name, displayname, rprofile, prefix, sys_prefix, verbose

see installspec

root_dir, workdir

default root directory of the notebook

host, port

'IP' and port of the hosting 'URL'

open_browser

whether to open the browser once launched

token

access token of the notebook

base_url

base address, default is '/jupyter/'

async

whether to open the notebook in the background

dry_run

whether to display the command instead of executing them; used to debug the code

Value

jupyter_bin returns the 'Jupyter' notebook binary path; jupyter_options returns the 'Jupyter' configuration in strings; jupyter_server_list returns a table of existing local 'Jupyter' server hosts, ports, and tokens; jupyter_check_launch returns true if a new server has been created, or false if there has been an existing server at the port; other functions return nothing.

Examples

## Not run: 

# Requires installation of conda
library(rpymat)

# Install conda, if you have done so, skip
configure_conda()


# Install Jupyter notebook
add_jupyter(register_R = TRUE)


# Utility functions
jupyter_bin()

# Please install `dipsaus` package to enable `async=TRUE` with
# better experience
jupyter_launch(async = FALSE, open_browser = TRUE)



## End(Not run)

Get 'Python' built-in object

Description

Get 'Python' built-in object

Usage

py_builtin(name, convert = FALSE)

Arguments

name

object name

convert

see import_builtins

Value

A python built-in object specified by name

Examples

if(interactive() && dir.exists(env_path())) {


# ------ Basic case: use python `int` as an R function ---------
py_int <- py_builtin("int", convert = TRUE)

# a is an R object now
a <- py_int(9)
print(a)
class(a)

# ------ Use python `int` as a Python function -----------------
py_int2 <- py_builtin("int", convert = FALSE)

# b in a python object
b <- py_int2(9)

# There is no '[1] ' when printing
print(b)
class(b)

# convert to R object
py_to_r(b)



}

List in 'Python'

Description

List in 'Python'

Usage

py_list(..., convert = FALSE)

Arguments

...

passing to list ('Python')

convert

whether to convert the results back into R; default is no

Value

List instance, or an R vector if converted

Examples

if(interactive() && dir.exists(env_path())) {

  py_list(list(1,2,3))
  py_list(c(1,2,3))

  py_list(array(1:9, c(3,3)))
  py_list(list(list(1:3), letters[1:3]))

}

Slice index in 'Python' arrays

Description

Slice index in 'Python' arrays

Usage

py_slice(...)

Arguments

...

passing to slice ('Python')

Value

Index slice instance

Examples

if(interactive() && dir.exists(env_path())) {

  x <- np_array(array(seq(20), c(4, 5)))

  # equivalent to x[::2]
  x[py_slice(NULL, NULL, 2L)]

}

Read data frame from a 'xlsx' file

Description

Tries to use 'readxl' package or 'pandas' to read data frame.

Usage

read_xlsx(
  path,
  sheet = NULL,
  method = c("auto", "pandas", "readxl"),
  n_max = Inf,
  ...
)

Arguments

path

'xlsx' file path

sheet

either a character or an integer of which spread-sheet to read; the number starts from 1

method

which method to use for reading the 'xlsx' file; choices are 'auto' (automatically find proper method), 'pandas' (use pandas.read_xlsx), or 'readxl' (use the corresponding R package)

n_max

maximum number of rows (excluding headers) to read

...

passed to 'Python' function pandas.read_xlsx or readxl::read_excel, depending on method

Value

A data.frame table

Examples

## Not run: 

rpymat::read_xlsx("Book1.xlsx", sheet = 1)

rpymat::read_xlsx("Book1.xlsx", sheet = "sheet1")


## End(Not run)

Enable interactive 'python' from R

Description

Allows users to type 'python' command from R console for quick code evaluation or debugging.

Usage

repl_python(...)

Arguments

...

passed to repl_python in 'reticulate' package

Value

See repl_python


Wrappers around 'reticulate' package

Description

Almost the same with 'reticulate' functions, with rpymat enabled by default and some minor changes (see parameter convert and local)

Usage

import_main(convert = FALSE)

tuple(..., convert = FALSE)

py_tuple(..., convert = FALSE)

py_help(object)

np_array(data, ...)

import(module, as = NULL, convert = FALSE, delay_load = FALSE)

r_to_py(x, convert = FALSE)

py_to_r(x)

py_to_r_wrapper(x)

py_str(object, ...)

py_run_string(code, local = TRUE, convert = FALSE)

py_bool(x)

py_dict(keys, values, convert = FALSE)

py_call(x, ...)

py_del_attr(x, name)

py_del_item(x, name)

py_eval(code, convert = FALSE)

py_get_attr(x, name, silent = FALSE)

py_set_attr(x, name, value)

py_get_item(x, key, silent = FALSE)

py_set_item(x, name, value)

py_len(x, default = NULL)

py_none()

Arguments

convert

whether to convert 'Python' objects to R; default is FALSE. This is different to 'reticulate', but less error prone: users must explicitly convert 'Python' objects to R.

object, data, x, code, keys, values, ...

passed to corresponding 'reticulate' functions as data inputs

module, as, delay_load

import 'Python' module as alias

local

whether to execute code locally so the memory sets free when the function ends; default is true

name, silent, key, value, default

other parameters passing to the 'reticulate' functions

Value

'Python' built-in objects

Examples

library(rpymat)
if(interactive() && dir.exists(env_path())) {

  # tuple
  x <- tuple(1, 2, "a")
  print(x)

  # convert to R object
  py_to_r(x)

  # convert R object to python
  y <- r_to_py(list(a = 1, b = "s"))

  # get element
  py_get_item(y, "a")

  # get missing element
  py_get_item(y, "c", silent = TRUE)

}

Get 'Python' main process environment

Description

py automatically converts 'Python' objects to R objects. import_main does not convert by default; see 'Examples' for details.

Usage

py

Format

An object of class NULL of length 0.

Value

The 'Python' main process as a module

Examples

if(interactive() && dir.exists(env_path())) {

py_no_convert <- rpymat::import_main(convert = FALSE)

py$a <- matrix(seq_len(16), 4)

py_no_convert$a

py$a

}

Execute command with additional environments

Description

Enables 'conda' environment

Usage

cmd_create(command, shell, use_glue = TRUE)

cmd_set_env(command, key, value, quote = TRUE, quote_type = "cmd")

cmd_set_workdir(command, workdir)

cmd_set_conda(command, conda_path, env_path)

cmd_build(command, .env = parent.frame(), ...)

detect_shell(suggest = NULL)

run_command(
  command,
  shell = detect_shell(),
  use_glue = FALSE,
  enable_conda = TRUE,
  stdout = "",
  stderr = "",
  stdin = "",
  input = NULL,
  env_list = list(),
  wait = TRUE,
  timeout = 0,
  ...,
  workdir = getwd(),
  dry_run = FALSE,
  print_cmd = dry_run,
  glue_env = parent.frame()
)

Arguments

command

system command

shell

shell type

use_glue

whether to glue the command; default is false

key, value

environment variable key and value

quote, quote_type

whether to quote the environment variables and what quote type should use; see shQuote

workdir

the working directory

conda_path

'conda' path; default is conda_path

env_path

'conda' environment path; default is env_path

suggest

suggested shell type; default is 'cmd' on windows, or 'bash' on others

enable_conda

whether to activate 'conda'

stdout, stderr, stdin, input, wait, timeout, ...

passed to system2

env_list

a key-value pairs of environment variables

dry_run

whether to dry-run the command (do not execute, simply returns the command), useful to debug

print_cmd

whether to print the command out

glue_env, .env

the environment to evaluate variables when use_glue is true

Value

All the functions return a list with class rpymat_system_command except for run_command, which returns the exit code by system2.

Examples

run_command("conda install -y numpy", dry_run = TRUE)


a <- "This is a message"
run_command('echo "{a}"', dry_run = TRUE,
            enable_conda = FALSE, use_glue = TRUE)


## Not run: 

# Use `jupyter_launch()` instead. This is just a demonstration
run_command('"{jupyter_bin()}" server list', use_glue = TRUE)


## End(Not run)

Run 'Python' script

Description

A wrapper of py_run_file, but with rpymat enabled

Usage

run_script(
  x,
  work_dir = NULL,
  local = FALSE,
  convert = FALSE,
  globals = list()
)

run_pyscript(
  x,
  work_dir = NULL,
  local = FALSE,
  convert = FALSE,
  globals = list()
)

run_pystring(
  code,
  work_dir = NULL,
  local = FALSE,
  convert = FALSE,
  globals = list()
)

Arguments

x

'Python' script path

work_dir

working directory of the script

local, convert

passed to py_run_file

globals

named list of global R variables used by 'Python' script

code

'Python' code

Value

The values returned by py_run_file

Examples

## Not run: 

# Please configure conda environment first

x <- tempfile()
writeLines(c(
  "import re",
  "zipcode = re.findall(r'[0-9]{5,6}', r.address)"
), con = x)

address <- '2341 Main St., 72381'
rpymat::run_script(x)

py$zipcode


## End(Not run)