I was answering this question where commenters suggested !!ensym, and I thought this might be a good place to use curly curly {{ but I couldn't get it to work (maybe not applicable?).
How might I do this filter operation, without using filter_, eval/parse, or quote-unquote? Would ~ help?
My solution (1g) uses filter_ and conditions built up with paste. 1a works (but could it use {{ }} somehow?)
And what if we wanted to filter by multiple variables? This is where you see 2g working below (while 2a does not work anymore).
library(tidyverse)
set.seed(1234)
A <- matrix(rnorm(30),nrow = 10, ncol = 3) %>% as_tibble() %>% set_names(paste("var", seq(1:3), sep = ""))
varnames_1 <- c("var2")
(expected_result_1 <- filter(A, var2 > 0))
#> # A tibble: 3 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -2.35 0.0645 0.460
#> 2 0.429 0.959 -0.694
#> 3 -0.890 2.42 -0.936
(answer_1a <- filter(A,!!ensym(varnames_1) > 0)) # works (thanks joran and aosmith)
#> # A tibble: 3 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -2.35 0.0645 0.460
#> 2 0.429 0.959 -0.694
#> 3 -0.890 2.42 -0.936
(answer_1b <- filter_(A, varnames_1 > 0)) # filter_ not doing what I thought it might
#> Warning: filter_() is deprecated.
#> Please use filter() instead
#>
#> The 'programming' vignette or the tidyeval book can help you
#> to program with filter() : https://tidyeval.tidyverse.org
#> This warning is displayed once per session.
#> # A tibble: 10 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -1.21 -0.477 0.134
#> 2 0.277 -0.998 -0.491
#> 3 1.08 -0.776 -0.441
#> 4 -2.35 0.0645 0.460
#> 5 0.429 0.959 -0.694
#> 6 0.506 -0.110 -1.45
#> 7 -0.575 -0.511 0.575
#> 8 -0.547 -0.911 -1.02
#> 9 -0.564 -0.837 -0.0151
#> 10 -0.890 2.42 -0.936
(answer_1c <- filter(A, {{varnames_1}} > 0)) # curly curly not doing what I thought it might
#> # A tibble: 10 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -1.21 -0.477 0.134
#> 2 0.277 -0.998 -0.491
#> 3 1.08 -0.776 -0.441
#> 4 -2.35 0.0645 0.460
#> 5 0.429 0.959 -0.694
#> 6 0.506 -0.110 -1.45
#> 7 -0.575 -0.511 0.575
#> 8 -0.547 -0.911 -1.02
#> 9 -0.564 -0.837 -0.0151
#> 10 -0.890 2.42 -0.936
(answer_1d <- filter(A, {{varnames_1 > 0}})) # curly curly not doing what I thought it might
#> `arg` must be a symbol
conditions_1 <- paste(varnames_1, "> 0")
(answer_1e <- filter(A, conditions_1)) # does not work
#> Error: Argument 2 filter condition does not evaluate to a logical vector
(answer_1f <- filter(A, {{conditions_1}})) # curly curly not doing what I thought it might
#> Error: Argument 2 filter condition does not evaluate to a logical vector
(answer_1g <- filter_(A, conditions_1)) # works
#> # A tibble: 3 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -2.35 0.0645 0.460
#> 2 0.429 0.959 -0.694
#> 3 -0.890 2.42 -0.936
# what if we wanted to filter multiple variables?
varnames_2 <- c("var2", "var3")
(expected_result_2 <- filter(A, var2 > 0 & var3 > 0))
#> # A tibble: 1 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -2.35 0.0645 0.460
(answer_2a <- filter(A,!!ensym(varnames_2) > 0)) # does not work
#> Only strings can be converted to symbols
conditions_2 <- paste(paste(varnames_2, "> 0"), collapse = " & ")
(answer_2f <- filter(A, {{conditions_2}})) # curly curly not doing what I thought it might
#> Error: Argument 2 filter condition does not evaluate to a logical vector
(answer_2g <- filter_(A, conditions_2)) # works
#> # A tibble: 1 x 3
#> var1 var2 var3
#> <dbl> <dbl> <dbl>
#> 1 -2.35 0.0645 0.460
Created on 2019-08-28 by the reprex package (v0.3.0)
As Lionel points out, curly-curly works inside functions. To use it with filter, you thus have to wrap the call inside a function.
f <- function(.df, v) {
filter(.df, {{ v }} > 0)
}
# Curly-curly provides automatic NSE support
f( A, var2 )
# # A tibble: 3 x 3
# var1 var2 var3
# <dbl> <dbl> <dbl>
# 1 -2.35 0.0645 0.460
# 2 0.429 0.959 -0.694
# 3 -0.890 2.42 -0.936
# Strings have to be first converted to symbols
f( A, !!sym("var3") )
# # A tibble: 3 x 3
# var1 var2 var3
# <dbl> <dbl> <dbl>
# 1 -1.21 -0.477 0.134
# 2 -2.35 0.0645 0.460
# 3 -0.575 -0.511 0.575
Curly-curly is meant to reference a single argument. You can extend it to work with multiple variables through sequential application with the help of purrr::reduce. (Don't forget to convert your strings into actual variable names first!):
syms(varnames_2) %>% reduce(f, .init=A)
# # A tibble: 1 x 3
# var1 var2 var3
# <dbl> <dbl> <dbl>
# 1 -2.35 0.0645 0.460