The result
type captures expressions and functions that might cause error or otherwise fail to produce a value. Using result
can make code more readable and less error-prone by abstracting away the need for nested if
statements and tryCatch
blocks. At its most powerful, result
can be used to build a composable pipeline of functions that will either end with a success
object, or stop gracefully with a failure
object as soon as an error is detected.
Make your own functions fault-tolerant by wrapping your results in success
or failure
for downstream users to safely use:
library(result)
# A function that might fail
risky_fn <- function(good = TRUE) {
if (good) success(42)
else failure("cannot find good answer")
}
# Downstream users are now encouraged to inspect the result for success
risky_result <- risky_fn(good = TRUE)
if (is_success(risky_result)) {
print(paste0("The answer: ", value(risky_result)))
}
#> [1] "The answer: 42"
risky_result <- risky_fn(good = FALSE)
if (is_failure(risky_result)) {
cat(paste0("Failed because ", value(risky_result)))
}
#> Failed because cannot find good answer
If you are using others’ functions, wrap their results safely for immediate inspection, or wrap the whole function for safe use later:
# A third-party function that might crash with error
external_api <- function(good = TRUE) {
if (good) 42
else stop("Cannot connect to API")
}
# We can either safely inspect the results:
api_result <- as_result(external_api(good = TRUE))
is_success(api_result)
#> [1] TRUE
value(api_result)
#> [1] 42
# Or we can wrap the risky function for safe use later!
safely_call_api <- result(external_api)
safely_call_api(good = FALSE) |> is_failure()
#> [1] TRUE
We can chain together functions to create a pipeline that will either complete successfully or else stop gracefully at the first sign of failure.
Operations for our pipeline:
# Start by wrapping the first operation in a result to evaluate later
safely_call_api <- result(external_api)
# Other (risky) operations we want to safely use
add_some <- function(x, y) x + y
times_some <- function(x, y) x * y
times_too_much <- function(x, y) {
z <- x * y
if (z > 100) stop("Result has become too big") else z
}
A successful pipeline binding result
operations together:
process <-
safely_call_api(good = TRUE) |>
then_try(times_some, 2) |>
then_try(add_some, 10)
if (is_success(process)) {
print(paste0("Processed: ", value(process)))
} else {
cat(paste0("Could not process: ", value(process)))
}
#> [1] "Processed: 94"
A failing pipeline that should fail gracefully at the second step (times_too_much
):
process <-
safely_call_api(good = TRUE) |>
then_try(times_too_much, 50) |>
then_try(add_some, 10)
if (is_success(process)) {
print(paste0("Processed: ", value(process)))
} else {
cat(paste0("Could not process: ", value(process)))
}
#> Could not process: Result has become too big
You can install the released version of result from CRAN:
You can install the development version of result from GitHub with:
Result
is a generalization of the Maybe
type that is available to R devs from the maybe
package. Conversely, the Maybe
type is a special case of the Result
type where the error type is Nothing
. Maybe
shines in situations when an operation might return something or nothing. But Result
can convey more information than Maybe
in error situations (e.g., a status code or message explaining the error). In this R implementation, the result()
and as_result()
functions also capture errors automatically, sparing developers from having to wrap expressions in their own tryCatch
blocks.
The Result
type is a special case of the Either
type. Whereas Either
resolves to Left
or Right
, Result
usually resolves to notions of “Ok’ or”Error”. Use of Left
and Right
can be confusing to newcomers to functional concepts, and do not express the relevance of outcomes. In this R implementation of Result
, the result()
and as_result()
functions resolve to success
and failure
respectively, which should be intuitive to many developers. Implementations of Result
in other languages (such as Rust) sometimes use ok
and error
(or err
), but these can conflict with variable names and keywords in other languages.