Package 'sakura'

Title: Extension to R Serialization
Description: Extends the functionality of R serialization by augmenting the built-in reference hook system. This enhanced implementation allows an integrated single-pass operation that combines R serialization with third-party serialization methods. Facilitates the serialization of even complex R objects, which contain non-system reference objects, such as those accessed via external pointers, to enable their use in parallel and distributed computing.
Authors: Charlie Gao [aut, cre] , Hibiki AI Limited [cph]
Maintainer: Charlie Gao <[email protected]>
License: MIT + file LICENSE
Version: 0.1.0.9000
Built: 2025-03-04 13:28:32 UTC
Source: https://github.com/shikokuchuo/sakura

Help Index


sakura: Extension to R Serialization

Description

Extends the functionality of R serialization by augmenting the built-in reference hook system. This enhanced implementation allows an integrated single-pass operation that combines R serialization with third-party serialization methods. Facilitates the serialization of even complex R objects, which contain non-system reference objects, such as those accessed via external pointers, to enable their use in parallel and distributed computing.

Author(s)

Charlie Gao [email protected] (ORCID)

See Also

Useful links:


Create Serialization Configuration

Description

Returns a serialization configuration for custom serialization and unserialization of non-system reference objects, using the 'refhook' system of R native serialization. This allows their use across different R sessions.

Usage

serial_config(class, sfunc, ufunc, vec = FALSE)

Arguments

class

character string of the class of object custom serialization functions are applied to, e.g. ‘ArrowTabular’ or ‘torch_tensor’.

sfunc

a function that accepts a reference object inheriting from ‘class’ (or a list of such objects) and returns a raw vector.

ufunc

a function that accepts a raw vector and returns a reference object (or list of such objects).

vec

[default FALSE] whether or not the serialization functions are vectorized. If FALSE, they should accept and return reference objects individually e.g. arrow::write_to_raw and arrow::read_ipc_stream. If TRUE, they should accept and return a list of reference objects, e.g. torch::torch_serialize and torch::torch_load.

Value

A pairlist comprising the configuration. This may be provided to the 'hook' argument of serialize and unserialize.

Examples

serial_config("test_class", base::serialize, base::unserialize)

Serialize

Description

An extension of R native serialization using the 'refhook' system for custom serialization and unserialization of non-system reference objects.

Usage

serialize(x, hook = NULL)

unserialize(x, hook = NULL)

Arguments

x

an object.

hook

[default NULL] optionally, a configuration returned by serial_config.

Value

For serialize: a raw vector. For unserialize: the unserialized object.

Examples

vec <- serialize(data.frame())
vec
unserialize(vec)


obj <- list(arrow::as_arrow_table(iris), arrow::as_arrow_table(mtcars))
cfg <- serial_config(
  "ArrowTabular",
  arrow::write_to_raw,
  function(x) arrow::read_ipc_stream(x, as_data_frame = FALSE)
)
raw <- serialize(obj, cfg)
unserialize(raw, cfg)