Title:	Zero-Copy Julia to R Array Bridge via ALTREP
Version:	0.1.0
Description:	Provides zero-copy R views of Julia-owned arrays by implementing ALTREP (Alternative Representations) classes that return pointers directly into Julia's memory. The package integrates with 'JuliaCall' and uses C-level finalizers for safe cross-runtime garbage collection.
License:	MIT + file LICENSE
URL:	https://github.com/tanaylab/jlview
BugReports:	https://github.com/tanaylab/jlview/issues
Depends:	R (≥ 4.0.0)
SystemRequirements:	Julia (>= 1.6, https://julialang.org/downloads/)
Imports:	JuliaCall, Matrix, methods
Suggests:	knitr, rmarkdown, testthat (≥ 3.0.0)
VignetteBuilder:	knitr
Config/testthat/edition:	3
Encoding:	UTF-8
Language:	en-US
RoxygenNote:	7.3.3
NeedsCompilation:	yes
Packaged:	2026-03-19 12:45:24 UTC; aviezerl
Author:	Aviezer Lifshitz [aut, cre], Weizmann Institute of Science [cph]
Maintainer:	Aviezer Lifshitz <aviezer.lifshitz@weizmann.ac.il>
Repository:	CRAN
Date/Publication:	2026-03-23 17:40:08 UTC

Check if an object is a jlview ALTREP vector

Description

Check if an object is a jlview ALTREP vector

Usage

is_jlview(x)

Arguments

Value

TRUE if x is a jlview ALTREP vector, FALSE otherwise

Create a zero-copy R view of a Julia array

Description

Creates an ALTREP vector (or matrix, if 2D+) backed by Julia memory. The resulting R object shares the same memory as the Julia array, avoiding data copying. Modifications to the R object trigger copy-on-write (unless writeable = TRUE).

Usage

jlview(julia_array, writeable = FALSE, names = NULL, dimnames = NULL)

Arguments

Value

An ALTREP vector backed by Julia memory, or a standard R vector if the Julia type is not supported for zero-copy.

Examples

## Not run: 
JuliaCall::julia_setup()
# Create a Julia array and view it in R without copying
JuliaCall::julia_command("x = randn(1000)")
x <- jlview(JuliaCall::julia_eval("x"))
sum(x) # operates directly on Julia memory

## End(Not run)

Get current GC pressure information

Description

Returns the current amount of Julia memory pinned by jlview objects and the threshold at which forced GC is triggered.

Usage

jlview_gc_pressure()

Value

A list with pinned_bytes and threshold

Get information about a jlview object

Description

Returns metadata about a jlview ALTREP vector including the Julia element type, length, writeability, and release status.

Usage

jlview_info(x)

Arguments

Value

A named list with components:

type

Julia element type (e.g., "Float64")

length

Number of elements

writeable

Whether the view allows direct writes

released

Whether the view has been released

materialized

Whether COW materialization has occurred

Create a zero-copy R view of a named Julia matrix

Description

Creates a zero-copy ALTREP view of a Julia NamedArray matrix, preserving row and column names as dimnames.

Usage

jlview_named_matrix(julia_named_matrix)

Arguments

Value

An ALTREP matrix with dimnames set from the Julia NamedArray

Examples

## Not run: 
JuliaCall::julia_setup()
JuliaCall::julia_command("using NamedArrays")
m <- JuliaCall::julia_eval('NamedArray(randn(3,2), (["a","b","c"], ["x","y"]))')
x <- jlview_named_matrix(m)
rownames(x) # returns c("a", "b", "c")
colnames(x) # returns c("x", "y")

## End(Not run)

Create a zero-copy R view of a named Julia vector

Description

Creates a zero-copy ALTREP view of a Julia NamedArray vector, preserving the axis names.

Usage

jlview_named_vector(julia_named_array)

Arguments

Value

An ALTREP vector with names set from the Julia NamedArray

Examples

## Not run: 
JuliaCall::julia_setup()
JuliaCall::julia_command("using NamedArrays")
v <- JuliaCall::julia_eval('NamedArray([1.0, 2.0, 3.0], (["a", "b", "c"],))')
x <- jlview_named_vector(v)
names(x) # returns c("a", "b", "c")

## End(Not run)

Explicitly release a jlview object

Description

Unpins the Julia array immediately, freeing memory without waiting for R's garbage collector. After release, accessing the data will error.

Usage

jlview_release(x)

Arguments

Value

Invisible NULL

Set the GC pressure threshold

Description

When total pinned bytes exceeds this threshold, jlview forces an R garbage collection to reclaim stale ALTREP objects. Default is 10GB.

Usage

jlview_set_gc_threshold(bytes)

Arguments

Value

Invisible NULL

Create a zero-copy R sparse matrix view of a Julia SparseMatrixCSC

Description

Creates a dgCMatrix-class backed by a zero-copy ALTREP vector for the nonzero values (x slot). The row indices (i slot) and column pointers (p slot) are copied and shifted from 1-based (Julia) to 0-based (R) indexing in Julia, then returned as plain R integer vectors.

Usage

jlview_sparse(julia_sparse_matrix, lazy_indices = FALSE)

Arguments

Value

A dgCMatrix-class sparse matrix.

Examples

## Not run: 
JuliaCall::julia_setup()
JuliaCall::julia_command("using SparseArrays")
m <- JuliaCall::julia_eval("sprand(Float64, 100, 50, 0.1)")
s <- jlview_sparse(m)
class(s) # "dgCMatrix"

## End(Not run)

Use a jlview object within a scope, releasing it on exit

Description

Creates a jlview object and ensures it is released when the scope exits, even if an error occurs. This prevents memory leaks from forgotten releases.

Usage

with_jlview(
  julia_array,
  expr,
  writeable = FALSE,
  names = NULL,
  dimnames = NULL
)

Arguments

Value

The result of evaluating expr

Examples

## Not run: 
JuliaCall::julia_setup()
JuliaCall::julia_command("big = randn(100000)")
result <- with_jlview(JuliaCall::julia_eval("big"), {
    c(mean(.x), sd(.x))
})
# .x is automatically released here

## End(Not run)