Skip to content

otto-de/nom

Repository files navigation

nom

Utilities for working with the happy path in the face of anomalies.

(The name »nom« is a bad pun on the word »anomaly«. »Anomaly« is actually composed from »an-« and »homalos«, so the antonym of »anomal« should be »homal«. However, a funny name is obviously more important than etymological accuracy.)

Rationale

One popular way of signalling errors in Clojure is to return anomalies, i. e. data structures that carry information about the error and can be identified as such.

This library supports the handling of potential anomalies by providing tools to transparently propagate anomalies without obscuring the logic flow.

The convention we propose hereby is: functions don't need to expect anomalies in their formal parameters. Instead, whenever we have an anomaly at hand, we immediately propagate it to any dependent values. The actual handling of whatever anomalies may have occurred is deferred to the outside, e. g. at the level of an HTTP handler.

Usage

Dependency: [de.otto/nom "0.3.0"]

Require: [de.otto.nom.core :as nom]

Creation

You can create a nom anomaly using fail:

(nom/fail ::my-category {:some data})

or

(nom/fail ::my-category :some data)

which both produce the same:

[::nom/anomaly ::my-category {:some data}]

Our own anomaly format is a bit different from the cognitect.anomalies format, but those are recognized too and automatically adapted. In order to extend this behaviour to your own anomaly formats, extend the abominable? and adapt multimethods using a proper predicate, e. g.:

(defn my-anomaly? [x]
  ...whatever...)

(defmethod nom/abominable? my-anomaly? [_]
  true)

(defmethod nom/adapt my-anomaly? [x]
  (nom/fail (yaba x) :daba (doo x)))

Propagation

The basic principle is maybe best illustrated by showing nom. If you have a form like this:

(foo arg0 arg1)

where both arg0 or arg1 might be anomalies, then you can prepend nom:

(nom/nom foo arg0 arg1)

which does the same as the form above, but if arg0 or arg1 is an anomaly, then it will immediately return that anomaly and not call foo at all.

You can do this in a threading manner very similar to some-> and some->>, with nom-> and nom->>:

(nom/nom-> foo
           (bar baz)
           quux
           (wizzle wozzle))

which would be equivalent to:

(nom/nom wizzle
         (nom/nom quux
                  (nom/nom bar foo baz))
         wozzle)

i. e. thread the values just as in ->, but if any of the intermediate values is an anomaly, short-circuit and return that anomaly.

Another pattern is checking a number of values, then executing a body, but if any of the values is an anomaly, short-circuit and return the anomaly instead. That's what with-nom is for:

(nom/with-nom [something (something else) (and another thing)]
  (frob something)
  (wozzle another thing))

Of course, this begs for binding those values; we have let-nom> for that:

(nom/let-nom> [a something
               b (something else)
               c (and another thing)]
  (frob a b c))

This short-ciruits as soon as an anomaly is encountered in the bindings; in that case, no further binding nor the body is even evaluated. You might want to have anomalies propagate through the bindings instead and then do something about it in the body. That's what let-nom (without the >) does.

Maintainers

See MAINTAINERS.

Lifecyle status

See OSSMETADATA.

License

Copyright © 2021 OTTO GmbH & Co. KG

This program and the accompanying materials are made available under the terms of the Apache License 2.0, see the LICENSE file or https://www.apache.org/licenses/LICENSE-2.0.html

About

Tools for ignoring anomalies

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Contributors 3

  •  
  •  
  •