results

A (meta)data-centric, FP approach to error handling in Clojure

Intro

In a language that is focused on data, it makes sense for errors to be data as well. This library allows you to do just that, making it trivial for errors to be passed as messages, for example, in applications that use core.async or in REST APIs.

About

The Problem with Exceptions

The conceptual inconsistency of having exceptions in a functional programming language has been covered elsewhere:

An Error spec? (excellent points about problems with exceptions, e.g., in a core.async context, REST API results, etc.)
try/catch complects: We can do so much better
"Good Enough" error handling in Clojure
Erlang-style supervisor error handling for Clojure
Category Theory and Algebraic abstractions for Clojure and ClojureScript (this supports the use of maybe in a way that is analogous to what Haskell does)

Those are great resources and well worth a read and/or study. I have used dire for several years, and while I have enjoyed it (due to my Erlang background) it never quite felt like a good match for Clojure. I've only ever used cats in a REPL, and it was a lot of fun ... but I've only dabbled in languages that focus on category theory, never deployed or maintained anything with them. As such, while interesting, I find that approach somewhat divergant from idiomatic Clojure.

The Desire for Data-centric Simplicity

What I really wanted was something that felt like Clojure when handling errors (exception handling in Clojure jumps out of Clojure and into Java; I'd like to choose when I do that, not have to do it every time there's an error). To feel like Clojure, a solution would need to be data-focused.

A Simple Library

I worked through several different approaches to this for large organization projects that were ultimately deployed to production in public, user-facing projects. This library takes the lessons learned from all of those, providing:

a means by which error-type data (including aribtrary data, such as user-defined warnings) may be included in results unobtrusively (via vary-meta)
a simple record that implements a results protocol for checking the status of a single result (and for inspecting error-type data accumulated via meta)
a simple function-based API for collecting error-type data from multiple results

The combination of these allows me to create as simple or as complex an errro-handling solution as a project may require, without having to depart from Clojure idioms and without forcing an application into a narrow world-view of errors.

Usage

(require '[clojusc.results.core :as result])

Simple result:

(def r (result/create (+ 1 1)))
r
=> {:data 2}
(result/errors? r)
=> false

As you can see, the data (or expression to be evaulated, which will generate data) is stored by default in a value associated with the :data key. This may be overridden by the developer in an options hash map, when creating a result:

(def r (result/create 42 {:key-name :my-value}))
r
=> {:my-value 42}

Error-producing result:

(def r (result/create (/ 1 0)))
r
=> {:data nil}
(result/errors? r)
=> true
(result/errors r)
=> [{:msg "Divide by zero"}]

Collecting errors:

(def r2 (result/create (hash-map :a)))

(result/collect-errors [r r2])
=> [{:msg "Divide by zero"}
    {:msg "No value supplied for key: :a"}]

You can also collect values. Let's make some non-error results:

(def r3 (result/create (+ 1 2)))
(def r4 (result/create (* 2 (reduce + (range 1 7)))))

Now we can get the values:

(result/collect-values [r r2 r3 r4])
=> (nil nil 3 42)

And you can collect all:

(result/collect [r r2 r3 r4])
=> {:errors ({:msg "Divide by zero"}
             {:msg "No value supplied for key: :a"})
    :values (nil nil 3 42)
    :warnings ()}

... and there's more! See the unit tests for other usage examples (in particular, how to extract stacktraces, etc., from the errors in results).

Releases

Version	Status	Notes
0.2.0-SNAPSHOT	In development
0.1.0	Released	API-compatible with NASA CMR projects

Additional Resources

License ↟

Distributed under the Apache License, Version 2.0.

clojusc/results