---
title: "A quick overview of the _S4Arrays_ package"
author:
- name: Hervé Pagès
  affiliation: Fred Hutchinson Cancer Research Center, Seattle, WA
date: "Compiled `r doc_date()`;  Modified 24 July 2023"
package: S4Arrays
vignette: |
  %\VignetteIndexEntry{A quick overview of the S4Arrays package}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
output:
  BiocStyle::html_document
---


# Introduction

`r Biocpkg("S4Arrays")` is an infrastructure package that provides a
framework intended to facilitate implementation of _array-like_ containers
in other Bioconductor packages. _Array-like_ containers are S4 objects that
mimic the behavior of ordinary matrices or arrays in R. Please note that
the package is not intended to be used directly by the end user.


# Installation

Like any other Bioconductor package, `r Biocpkg("S4Arrays")` should always
be installed with `BiocManager::install()`:
```{r, eval=FALSE}
if (!requireNamespace("BiocManager", quietly = TRUE))
    install.packages("BiocManager")
BiocManager::install("S4Arrays")
```
However, note that `r Biocpkg("S4Arrays")` will usually get automatically
installed as a dependency of other Bioconductor packages, so explicitly
installing the package is usually not needed.


# The Array virtual class

At the center of the framework provided by the `r Biocpkg("S4Arrays")`
package is the Array virtual class whose only purpose is to be extended
by other S4 classes that wish to implement a container with an array-like
semantic. Examples of such classes are:

- The SparseArray class defined in the `r Biocpkg("SparseArray")` package.

- The DelayedArray class defined in the `r Biocpkg("DelayedArray")` package.

- The ArrayGrid and ArrayViewport classes defined in the
  `r Biocpkg("S4Arrays")` package itself.

Note that Array is a virtual class with no slots:

```{r, message=FALSE}
library(S4Arrays)

showClass("Array")
```


# The extract_array() generic function

The `r Biocpkg("S4Arrays")` package also introduces the `extract_array()`
S4 generic function that Arrays subclasses (a.k.a. Arrays extensions)
are expected to support via specific methods. This allows some basic
operations like `type()`, `as.array()` or `as.matrix()` to work
out-of-the-box on instances of these Arrays subclasses (a.k.a.
Arrays derivatives). It also allows them to be used as the seed of a
DelayedArray object.

Please refer to the man page of the `extract_array()` function for more
information: `?extract_array`


# Block processing of array-like objects

The `r Biocpkg("S4Arrays")` package provides a framework that facilitates
block processing of array-like objects. Note that block processing is
typically used on _on-disk_ objects, that is, on objects where the array
data is stored on disk. This framework consists of:

- ArrayGrid and ArrayViewport objects for representing grids and viewports
  on array-like objects. See `?ArrayGrid` for more information.

- The `read_block()` and `write_block()` functions to read and write array
  blocks. See `?read_block` and `?write_block` for more information.

- The `mapToGrid()` and `mapToRef()` functions to map a set of reference
  array positions to grid positions and vice-versa. See `?mapToGrid` for
  more information.


# Other functionalities

In addition to the above, the `r Biocpkg("S4Arrays")` package provides
the following functionalities:

- The `is_sparse()` generic function for determining whether an array-like
  object uses a sparse representation or not. See `?is_sparse` for more
  information.

- Low-level utilities for manipulating array selections. See `?Lindex2Mindex`
  for more information.

- `aperm2()`: an extension of `base::aperm()` that allows dropping and/or
  adding _ineffective dimensions_. See `?aperm2` for more information.

- The `abind()` generic function for binding multidimensional array-like
  objects along any dimension. See `?abind` for more information.


# Session information

```{r}
sessionInfo()
```