A quosure is a quoted expression that keeps track of the environment it’s in. For example it can be used in code to access parts of objects that exist within the environment used to evaluate that code, i.e. selecting columns from a data frame which has been piped into select.
We have a data frame called “transport_data” and it has 13 columns shown below. We want to select the column “number.of.items”, using quosure.
names(transport_data)## [1] "sender.country" "sender.city" "sender.state"
## [4] "sender.latitude" "sender.longitude" "receiver.country"
## [7] "receiver.city" "receiver.state" "receiver.latitude"
## [10] "receiver.longitude" "date" "number.of.items"
## [13] "percent.of.all.items"measure_col_naked <- quo(number.of.items)We need to tell the select statement to treat the object measure_col_naked as a quosure. To do this we need to do what’s called quasi quotation of an expression. This un-quotes a quosure allowing it to be evaluated within the context that the code appears. The operator of unquoting your quosure is a double quotation mark.
transport_data %>%
select(!!measure_col_naked)## # A tibble: 112 x 1
## number.of.items
## <dbl>
## 1 381
## 2 372
## 3 255
## 4 351
## 5 492
## 6 347
## 7 462
## 8 500
## 9 414
## 10 426
## # ... with 102 more rowsAlternatively, we can use select(UQ) for unquoting a quosure and then the quosure, so measure_col_naked.
transport_data %>%
select(UQ(measure_col_naked))## # A tibble: 112 x 1
## number.of.items
## <dbl>
## 1 381
## 2 372
## 3 255
## 4 351
## 5 492
## 6 347
## 7 462
## 8 500
## 9 414
## 10 426
## # ... with 102 more rowswe have a variable called useful_columns_naked which we attempt to build using a vector so what we do is we specify this instead of as a vector, as a quos, so as a quosure containing multiple variables.
useful_columns_naked <- quos(date, receiver.city, number.of.items)If you have multiple arguments within a quosure, you need to use three exclamation marks to unquote.
transport_data %>%
select(!!!useful_columns_naked)## # A tibble: 112 x 3
## date receiver.city number.of.items
## <date> <chr> <dbl>
## 1 1884-08-10 Limbach 381
## 2 1885-07-01 Limbach 372
## 3 1885-11-14 Limbach 255
## 4 1877-03-14 Limbach 351
## 5 1880-02-04 Limbach 492
## 6 1880-08-19 Limbach 347
## 7 1880-12-05 Limbach 462
## 8 1881-05-11 Limbach 500
## 9 1881-08-03 Limbach 414
## 10 1882-02-08 Limbach 426
## # ... with 102 more rowsLearning the r tidyverse course by Charlie Joey Hadley