Skip to contents

Introduction

Bradley Buchsbaum

2023-04-26

Introduction.Rmd

Introduction

The deflist package provides a read-only list-like object that retrieves elements with a function call. This is particularly useful for handling large datasets where elements are computed on-demand and not stored in memory.

In this vignette, we will provide an overview of the main features of the deflist package and demonstrate its usage with examples.

Creating a deferred list

To create a deferred list, use the deflist function. The key input is a function that defines how to access elements in the list. The deferred list can also be configured to memoise element access and cache the results for faster retrieval.

Let’s create a simple deferred list for square numbers:

library(deflist)

square_fun <- function(i) i^2
square_deflist <- deflist(square_fun, len = 5)
square_deflist
#> deflist:  5  elements. 
#> memoised:  FALSE

Accessing elements

Elements in a deferred list can be accessed using standard list indexing, either by position or by name (if names are provided):

# Access by index
square_deflist[[1]]
#> [1] 1
square_deflist[[2]]
#> [1] 4

# Access multiple elements
square_deflist[c(1, 3, 5)]
#> [[1]]
#> [1] 1
#> 
#> [[2]]
#> [1] 9
#> 
#> [[3]]
#> [1] 25

Memoisation

Deferred lists can be configured to use memoisation, which caches the results of function calls to speed up repeated access to the same elements. To enable memoisation, set the memoise argument to TRUE when creating the deferred list. You can also specify a cache type (“memory” or “file”) and a cache directory for file-based caching.

Here’s an example with memoisation enabled:

memoised_square_fun <- deflist(square_fun, len = 5, memoise = TRUE, cache = "memory")
memoised_square_fun
#> deflist:  5  elements. 
#> memoised:  TRUE

# Access an element multiple times
system.time({ for (i in 1:1000) memoised_square_fun[[1]] })
#>    user  system elapsed 
#>   0.097   0.000   0.098

Converting to a list

A deferred list can be converted to a standard R list using the as.list method:


square_list <- as.list(square_deflist)
square_list
#> [[1]]
#> [1] 1
#> 
#> [[2]]
#> [1] 4
#> 
#> [[3]]
#> [1] 9
#> 
#> [[4]]
#> [1] 16
#> 
#> [[5]]
#> [1] 25

Read-only protection

Deferred lists are read-only, meaning that you cannot modify their elements. Attempting to do so will result in an error:

try(square_deflist[[1]] <- 0)
#> Error in `[[<-`(`*tmp*`, 1, value = 0) : 
#>   read only list, cannot set elements

Use Cases

Some potential use cases for deflist include:

Large datasets: When working with very large datasets, loading the entire dataset into memory might not be feasible. By using deflist, you can compute and retrieve elements as needed, saving memory and potentially speeding up processing.

Expensive computations: When elements of a list are expensive to compute, deflist can be used to cache the results, reducing the time required to recompute the same element multiple times.

Dynamic content: In cases where the content of a list may change over time, deflist can be used to ensure that the most up-to-date information is always retrieved when an element is accessed.

API calls: When working with APIs that have rate limits or require significant processing time, deflist can be used to make API calls on-demand, avoiding unnecessary calls and reducing the chance of hitting rate limits.

Lazy evaluation: In situations where you only need a subset of the data or elements from a list, deflist allows you to retrieve only the required elements, potentially speeding up processing and reducing memory usage.

Caveats

While deflist offers several advantages in certain scenarios, there are some potential dangers or caveats that you should be aware of when using this data structure:

Memory management: Since deflist retrieves elements on-demand, it can help reduce memory usage in certain scenarios. However, if the retrieved elements are not discarded after usage, it could lead to increased memory consumption over time. It is essential to manage memory carefully when working with deflist.

Performance: While deflist can improve performance in cases where the entire dataset is not required or when caching is beneficial, it may introduce overhead when accessing elements. This overhead could lead to slower performance compared to using a standard list if elements are accessed frequently or sequentially.

Complexity: Using deflist introduces an additional level of complexity to your code, which can make it harder to understand, maintain, and debug. You should weigh the benefits of using deflist against the potential increase in complexity when deciding whether to use it.

Error handling: If the function used to retrieve elements from the deflist encounters an error or fails, you need to handle such cases gracefully to avoid breaking your application. This might require additional error handling and exception handling in your code.

Caching and memoization: If deflist is used with memoization, it is crucial to consider the implications of caching results. In some cases, caching might not be desired, as it can lead to stale or outdated results. Additionally, when using file-based caching, you need to ensure that the cache directory is managed correctly, which could involve handling file I/O errors, managing disk space, and implementing cache eviction policies.

Conclusion

The deflist package provides a flexible way to work with large datasets that require on-demand computation. It can be easily integrated into existing R workflows, making it a valuable tool for a wide range of applications.