The other day I needed to generate UUID's (universally unique identifier) for some database inserts. I couldn't think of a package off the top of my head. Off to Google I went.
There were several promising results.
Below I discuss the how I ultimately decided to use the {uuid} package to generate UUIDs.
uuid
The {uuid} package is small. There is 1 function UUIDgenerate. It is important to check the package descriptions for information (see the code above for a truncated version). This package doesn't import any other packages, which is nice. However it uses a low version number (0.1-2) and was last published in 2015. That isn't very actively maintained. But, a UUID generate probably doesn't need to be updated very frequently. The author as a r-project.org email address. This is the organization that builds R. So he knows R well.
I even went so far as to check the mirrored CRAN GitHub repo for this package. The actual code that generates the UUIDs is written in C then called from R. Cool, it'll be fast.
But how is the interface? The only parameter is use.time and it can be NA, TRUE, or FALSE. Pretty easy. Depending on the input provided to this parameter it will generate a v4 or v1 UUID. This isn't super important for my applications. I do, however, like to generate v4 UUIDs for consistency between all the different languages, databases I use.
uuid::UUIDgenerate(use.time = NA)
#> [1] "da84f022-2ee1-4221-898c-c0f580a12c7a"
uuid::UUIDgenerate(use.time=F)
# > [1] "9781816b-d5b1-45a3-b90d-a8136fe23392"
uuid::UUIDgenerate(use.time=T)
#> [1] "79d194fc-4510-11ea-bbdb-109add52c272"Ulitmately I decided to use this package because of it's simplicity, C source code, and the author's affiliation with the R Project.
The contenders
{dplR} is a dendrochronolgy package that has many, many functions. I didn't want that much bloat in the package I was writing.
{ids} has functions that generate many different types of IDs. This might be useful in the future. But right now I only needed a UUID. The code that generates the UUID in this package directly calls the {uuid} package. Might as well go to the source.
The R-Bloggers post is short and instructive. However, UUIDs rely on very good random number generators and I don't trust R to do that very well. Another important note is that this implementation uses sample which may cause problems if using set.seed on the randomness of the UUID.
Encapsulate UUID generation
I need to generate UUIDs for each row in a data.frame. So I will need to loop over each row and generate a UUID. I wanted to make this process easier than uuid::UUIDgenerate(...) in a loop. So I wrote the following function.
gen_uuid <- function (n) {
v4 <- function (i) uuid::UUIDgenerate(use.time = F)
vapply(seq_len(n), v4, character(1))
}
# usage example
df <- data.frame(
x = sample(LETTERS, 5),
y = runif(5)
)
df$uuid <- gen_uuid(nrow(df))
#> x y uuid
#> 1 O 0.03186449 25febfb8-c175-46cd-8faf-6803089538c7
#> 2 N 0.97447114 879ec786-c53c-4cec-ba1b-8bab5faf4a77
#> 3 C 0.27902098 8e0ee5e2-545e-4d0e-b3a7-c22fc6a93db1
#> 4 G 0.76055082 a1ac592d-551c-4fdf-bcbb-bc86c424108c
#> 5 W 0.44379492 f88d8a9b-8716-4ee8-88ea-9ce4ff89d538Yes. It might be overkill. But programming is all about making your life easier.
Wrap up
So there you have it. My process of looking for a package. I don't always spend this much time trying to figure out which packages to use. In cases like this it was worth it. I don't want the package I am writing to rely on giant packages. Avoid the bloat!
One last tip. Try different search engines when looking for R packages. My default search engine is DuckDuckGo. It isn't the best at surfacing tech related results. When I initially searched for UUIDs with R it returned the {dplR} package. When I switched to Google the first result was the {uuid} package.