tidy evaluation cannot work in function in R

Question

Context

I am learning tidy evaluation on this website, And I see an example:

x <- sym("height")

expr(transmute(starwars, bmi = mass / (!! x)^2))
#> transmute(starwars, bmi = mass/(height)^2)

transmute(starwars, bmi = mass / (!! x)^2)
#> # A tibble: 87 x 1
#>     bmi
#>   <dbl>
#> 1  26.0
#> 2  26.9
#> 3  34.7
#> 4  33.3
#> # ... with 83 more rows

Then, I imitated the above example in my own code using sym and !!, but it reported an error.

library(survival)
library(rms)
data(cancer)

x = sym('meal.cal')

expr(cph(Surv(time, status) ~ rcs(!! x), data = lung))
# cph(Surv(time, status) ~ rcs(meal.cal), data = lung)

cph(Surv(time, status) ~ rcs(!!x), data = lung)
# Error in !x : invalid argument type

Question

Why do I use sym and !! in my code will report an error and how to fix it.

Is it because of the rcs().

By the way you're reading a very old version of this dplyr vignette. We no longer recommend using `!!` in simple applications. Instead, we now provide the easier `{{` and `.data` patterns. See https://dplyr.tidyverse.org/articles/programming.html. However these patterns won't work with base functions, only with tidyverse ones. — Lionel Henry, Sep 26 '22 at 14:24

TimTeaFan · Accepted Answer · 2022-09-26T13:16:04.240

We can use rlang::inject() to splice arguments with !! into functions that usually don't support tidy evaluation. This saves us from using eval(expr(...)) and also answers your question why we don't need rlang::inject() with dplyr::transmute(). The latter already supports tidy evaluation while cph() does not.

library(survival)
library(rms)
data(cancer)

x = sym('meal.cal')

rlang::inject(cph(Surv(time, status) ~ rcs(!!x), data = lung))
#> Frequencies of Missing Values Due to Each Variable
#> Surv(time, status)           meal.cal 
#>                  0                 47 
#> 
#> Cox Proportional Hazards Model
#>  
#>  cph(formula = Surv(time, status) ~ rcs(meal.cal), data = lung)
#>  
#>  
#>                          Model Tests    Discrimination    
#>                                                Indexes    
#>  Obs        181    LR chi2      0.72    R2       0.004    
#>  Events     134    d.f.            4    R2(4,181)0.000    
#>  Center -0.3714    Pr(> chi2) 0.9485    R2(4,134)0.000    
#>                    Score chi2   0.76    Dxy      0.048    
#>                    Pr(> chi2) 0.9443                      
#>  
#>              Coef    S.E.   Wald Z Pr(>|Z|)
#>  meal.cal    -0.0006 0.0013 -0.48  0.6299  
#>  meal.cal'    0.0007 0.0051  0.14  0.8860  
#>  meal.cal''   0.0010 0.0261  0.04  0.9682  
#>  meal.cal''' -0.0132 0.0676 -0.19  0.8456  
#>

Without tidy evaluation we can stay in base R and use eval(bquote(...)) and splice x with .(x).

library(survival)
library(rms)
data(cancer)

x = sym('meal.cal')

eval(bquote(cph(Surv(time, status) ~ rcs(.(x)), data = lung)))
#> Frequencies of Missing Values Due to Each Variable
#> Surv(time, status)           meal.cal 
#>                  0                 47 
#> 
#> Cox Proportional Hazards Model
#>  
#>  cph(formula = Surv(time, status) ~ rcs(meal.cal), data = lung)
#>  
#>  
#>                          Model Tests    Discrimination    
#>                                                Indexes    
#>  Obs        181    LR chi2      0.72    R2       0.004    
#>  Events     134    d.f.            4    R2(4,181)0.000    
#>  Center -0.3714    Pr(> chi2) 0.9485    R2(4,134)0.000    
#>                    Score chi2   0.76    Dxy      0.048    
#>                    Pr(> chi2) 0.9443                      
#>  
#>              Coef    S.E.   Wald Z Pr(>|Z|)
#>  meal.cal    -0.0006 0.0013 -0.48  0.6299  
#>  meal.cal'    0.0007 0.0051  0.14  0.8860  
#>  meal.cal''   0.0010 0.0261  0.04  0.9682  
#>  meal.cal''' -0.0132 0.0676 -0.19  0.8456  
#>

^{Created on 2022-09-26 by the reprex package (v2.0.1)}

Thanks @Tim. I am using `rlang::inject()` and `rms::Predict()` jointly, but an error occur. I would appreciate it if you could help me look at it. The question is post at [here](https://stackoverflow.com/questions/73878304/inject-function-from-rlang-package-cannot-work-with-predict-function-from-rm). — zhiwei li, Sep 28 '22 at 08:26

score 2 · Answer 2 · answered Sep 26 '22 at 12:49

2

You can use eval()

eval(expr(cph(Surv(time, status) ~ rcs(!! x), data = lung)))

answered Sep 26 '22 at 12:49

langtang

22,248
1
12
27

Thanks @langtang. It works in my code. But I don't why `transmute(starwars, bmi = mass / (!! x)^2)` works fine without `eval()`. However, I need to wrap my code in `eval()`. – zhiwei li Sep 26 '22 at 12:55
1

see @TimTeaFan for nice explanation, and alternative options. – langtang Sep 26 '22 at 13:18

tidy evaluation cannot work in function in R

2 Answers2