Build a call to get expression for do.call execution

Giving this example:

lstParams <- list()
lstParams$formula <- as.formula("Sepal.Length ~ Sepal.Width + Petal.Length")                         
lstParams$data <- substitute(iris)
lstParams$distribution <- "bernoulli"

mdl <- do.call("gbm", lstParams)
mdl
gbm(formula=Sepal.Length ~ Sepal.Width + Petal.Length, distribution="bernoulli", 
    data=iris)
# A gradient boosted model with bernoulli loss function.
# 100 iterations were performed.
# There were 2 predictors of which 0 had non-zero influence.

I can get the call to gbm function by :

mdl$call
gbm(formula=Sepal.Length ~ Sepal.Width + Petal.Length, distribution="bernoulli", 
    data=iris)

Since some error can happen in: mdl<-do.call("gbm",lstParams) (e.g some NAs found in the target variable or wrong formula):

lstParams <- list()
lstParams$formula <- as.formula("Sepal.Length ~ Sepal.Width + Petal.Lengthx")                        
lstParams$data <- substitute(iris)
lstParams$distribution <- "bernoulli"

mdl <- do.call("gbm", lstParams)
# Error in eval(predvars, data, env) : object 'Petal.Lengthx' not found

I would like to get the call to gbm function in advance in order to debug it.

Is there any way to get that call before the execution? Thanks

Solution

as.call is your friend. quote function name as first list element first.

lstParams <- list()
lstParams$name <- quote(gbm::gbm)
lstParams$formula <- as.formula("Sepal.Length ~ Sepal.Width + Petal.Length")                       
lstParams$data <- substitute(iris)
lstParams$distribution <- "bernoulli"
g <- as.call(lstParams)

Call

g
# gbm::gbm(formula = Sepal.Length ~ Sepal.Width + Petal.Length, 
#          data = iris, distribution = "bernoulli")

Evaluation

eval(g)
# gbm::gbm(formula = Sepal.Length ~ Sepal.Width + Petal.Length, 
#          distribution = "bernoulli", data = iris)
# A gradient boosted model with bernoulli loss function.
# 100 iterations were performed.
# There were 2 predictors of which 0 had non-zero influence.

You can still use do.call.

do.call("gbm", as.list(g)[-1])
# A gradient boosted model with bernoulli loss function.
# 100 iterations were performed.
# There were 2 predictors of which 0 had non-zero influence.