A.3 Function short-cut

apply one function to your data to create a new variable: mutate(mod=map(data,function))
instead of using i in 1:length(object): for (i in seq_along(object))
apply multiple function: map_dbl
apply multiple function to multiple variables:map2
autoplot(data) plot times series data
mod_tidy = linear(reg) %>% set_engine('lm') %>% fit(price ~ ., data=data) fit lm model. It could also fit other models (stan, spark, glmnet, keras)

• Sometimes, data-masking will not be able to recognize whether you’re calling from environment or data variables. To bypass this, we use .data$variable or .env$variable. For example data %>% mutate(x=.env$variable/.data$variable
  
• Problems with data-masking:
  
  • Unexpected masking by data-var: Use .data and .env to disambiguate
    
  • Data-var cant get through:
    
  • Tunnel data-var with {{}} + Subset .data with [[]]
• Passing Data-variables through arguments

library("dplyr")
mean_by <- function(data,by,var){
    data %>%
        group_by({{{by}}}) %>%
        summarise("{{var}}":=mean({{var}})) # new name for each var will be created by tunnel data-var inside strings
}

mean_by <- function(data,by,var){
    data %>%
        group_by({{{by}}}) %>%
        summarise("{var}":=mean({{var}})) # use single {} to glue the string, but hard to reuse code in functions
}

• Trouble with selection:
  

library("purrr")
name <- c("mass","height")
starwars %>% select(name) # Data-var. Here you are referring to variable named "name"

starwars %>% select(all_of((name))) # use all_of() to disambiguate when 

averages <- function(data,vars){ # take character vectors with all_of()
    data %>%
        select(all_of(vars)) %>%
        map_dbl(mean,na.rm=TRUE)
} 

x = c("Sepal.Length","Petal.Length")
iris %>% averages(x)


# Another way
averages <- function(data,vars){ # Tunnel selectiosn with {{}}
    data %>%
        select({{vars}}) %>%
        map_dbl(mean,na.rm=TRUE)
} 

x = c("Sepal.Length","Petal.Length")
iris %>% averages(x)