# CrediblyCurious

Nick Tierney's (mostly) rstats blog

## Glue magic Part I

Categories: rstats glue strings

Lately I’ve found myself using Jim Hester’s glue package instead of paste0 or sprintf. This post marks the start of an ongoing series of little magic spells using the glue package.

# The back story

I’ve been through a few stages of discovery for combining strings of text together.

First, it was just the very idea that this was possible - was AMAZING.

## paste

When I learnt about the paste function, it meant I could do things like this:

paste("path/to/file_", 1:5, ".csv", sep = "")

## [1] "path/to/file_1.csv" "path/to/file_2.csv" "path/to/file_3.csv"
## [4] "path/to/file_4.csv" "path/to/file_5.csv"


Here, this reads as, roughly, “take this string, insert the numbers 1:5 in the middle of it, and separate those strings with no space "".

## paste0

And then there was paste0, which meant that I didn’t have to write sep = "" all the time.

So, now I could write:

paste0("path/to/file_", 1:5, ".csv")

## [1] "path/to/file_1.csv" "path/to/file_2.csv" "path/to/file_3.csv"
## [4] "path/to/file_4.csv" "path/to/file_5.csv"


## sprintf

And then the sprintf function, which means you can do this:

sprintf("path/to/file_%s.csv", 1:5)

## [1] "path/to/file_1.csv" "path/to/file_2.csv" "path/to/file_3.csv"
## [4] "path/to/file_4.csv" "path/to/file_5.csv"


Here, %s is substituted in for the R code you write afterwards. This is nice because it also means you are able to drop the R code into the middle of the string without having to open and close it again. I feel like I can better express what I want to say, and don’t have to spend time remembering other book keeping things.

## And now for glue magic

I am now always turning to glue, because it makes the intent of what I want to do clearer. For example, we can take our sprintf use earlier and instead do the following with glue.

library(glue)
glue("path/to/file_{1:5}.csv")

## path/to/file_1.csv
## path/to/file_2.csv
## path/to/file_3.csv
## path/to/file_4.csv
## path/to/file_5.csv


What is going on here? You are now able to refer to R objects inside the string, which are captured in the {}.

I really like this, because it means that I don’t need to worry about ending the string, inserting the R object, and handling the other bits and pieces. My intent here feels super clear: “Insert the R code in the bit with {}”.

Don’t want to use {}? That’s also fine, you can control that with .open and .close:

glue("path/to/file_[1:5].csv", .open = "[", .close = "]")

## path/to/file_1.csv
## path/to/file_2.csv
## path/to/file_3.csv
## path/to/file_4.csv
## path/to/file_5.csv


## Combining many strings

Or if you want to collapse, or smush together many strings, you use glue_collapse, because you want to collapse together many pieces.

Say, for example, that you want to write out a sentence where you state all of the variables in a dataset, like the french_fries dataset from reshape2:

# get the french fries data
library(reshape2)

time treatment subject rep potato buttery grassy rancid painty
61 1 1 3 1 2.9 0.0 0.0 0.0 5.5
25 1 1 3 2 14.0 0.0 0.0 1.1 0.0
62 1 1 10 1 11.0 6.4 0.0 0.0 0.0
26 1 1 10 2 9.9 5.9 2.9 2.2 0.0
63 1 1 15 1 1.2 0.1 0.0 1.1 5.1
27 1 1 15 2 8.8 3.0 3.6 1.5 2.3

Here, we tell it what we want our separations be - in this case, since we have a list, we want everything to be separate by a comma and a space.

fries_names <- names(french_fries)

fries_inline <- glue::glue_collapse(fries_names,
sep = ", ")

fries_inline

## time, treatment, subject, rep, potato, buttery, grassy, rancid, painty


And now you can include this in your rmarkdown text, so now I can dynamically generate the sentence:

The variables in our dataset are r fries_inline


The variables in our dataset are time, treatment, subject, rep, potato, buttery, grassy, rancid, painty.

(PS, You can include a verbatim inline expression with knitr::inline_expr()).

But, what if you want to add an “and” at the end of the sentence?

You can use the last argument:

fries_inline <- glue::glue_collapse(fries_names,
sep = ", ",
last = ", and ")

The variables in our dataset are r fries_inline


The variables in our dataset are time, treatment, subject, rep, potato, buttery, grassy, rancid, and painty.

# End

paste, paste0, and sprintf are awesome, but I use glue because I find it means I can write code that more clearly captures my intent, and means I don’t need to worry about other book keeping. I also get really nice features, like being able to construct sentences, and modify them to do things at the end.

Massive praise to Jim Hester for his work on glue - you should check out his great talk at UseR!2018 below. Jim has also been putting out some really great videos on #rstats on youtube that are well worth yout time