Carica is a flexible configuration library.
It offers:
[leathekd/carica "1.3.0"]
;; *or*, when not using JSON config files:
[leathekd/carica "1.3.0" :exclusions [[cheshire]]]
Carica looks for the config files on the classpath.
Leiningen will add a directory called "resources" to the classpath even though the directory is not created by default, so create a "resources" directory at the root of your project. Now, create and open "resources/config.edn" in your favorite editor. (The following will be evaluated by the eval middleware.)
{:foobar-timeout 300 #_"In seconds"
:favorite-hour-of-day 8 #_"0-23"
:blacklist nil
:export-dir "/mnt/export"
:timeout-ms (* 20 60 1000) #_"20 minutes"
:db {:classname "org.postgresql.Driver"
:subprotocol "postgresql"
:subname "//localhost/test"
:username "cosmo"
:password "toomanysecrets"}}
(If you're wondering about the #_"" comments, we've found that they're less prone to errors in the configuration files. That is, from accidental newlines or from pulling up the closing brace into a line with a ;; comment.)
Now, with all of that in place, open a new REPL session:
(use '[carica.core])
(config :export-dir)
;;=> "/mnt/export"
(config :db :username)
;;=> "cosmo"
(config :blacklist)
;;=> nil
(config :non-existent-key)
;;=> nil (with a warning message logged)
That's it!
Maybe you already have a config file with a different name, or a
config.edn that you use for a different purpose. No problem. To
override what files Carica loads you can create your own config
function using the configurer
function.
Given a list of resources in the format expected by get-configs,
configurer
will return a function to query configuration.
The list of resources can contain Strings, Files or URLs.
If you override the default config
function in this manner you must
also override the override-config
function if you intend to use it.
(ns my-proj.config
(:require [carica.core :refer [configurer
overrider
resources]]))
(def config (configurer (resources "proj_config.edn")))
(def override-config (overrider config))
Calling my-proj.config/config
will work the same as calling
carica.core/config
except that it will use your config file.
You can change the resulting config map in Carica by adding
middleware. For instance, if you trust the data you are reading and
want to have it evaluated, there is an eval-config
middleware
provided. If you want the config to be read only once from disk, you
can use the cache-config
middleware.
(ns my-proj.config
(:require [carica.core :refer [configurer
resources]]))
(def config (configurer (resources "proj_config.edn")
[eval-config
cache-config]))
In typical middleware fashion, eval-config
and cache-config
are
functions that take a function as their only argument and return a
function that takes a list of resources as its only input. For
instance, an example cache-config
function:
(defn cache-config
"Config middleware that will cache the config map so that it is
loaded only once."
[f]
(memoize (fn [resources]
(f resources))))
Middleware can also expose some internal state and options to the
outside world by wrapping the function that is returned. As an
example, the actual cache-config
function:
(defn cache-config
"Config middleware that will cache the config map so that it is
loaded only once."
[f]
(let [mem (atom {})]
(wrap-middleware-fn
(fn [resources]
(if-let [e (find @mem resources)]
(val e)
(let [ret (f resources)]
(swap! mem assoc resources ret)
ret)))
{:carica/mem mem})))
This ultimately returns:
{:carica/options {:carica/mem <the-cache-atom>}
:carica/fn <the-config-fn>}
This map is built up from all of the defined middleware so the keys
used in the cache-config
function should be unique. To access
the options, call config
as normal with the beginning path of
:carica/middleware
.
For example:
(swap! (cached-cfg :carica/middleware :carica/mem) empty)
Note: Specifically in the the cache-config
case there is a
clear-config-cache!
function that can be called. See the doc
for that function if you use a custom defined config
function.
Out of the box Carica supports Clojure, EDN, and JSON config files. More can be added as described below.
There are some differences in behavior between EDN and CLJ. Clojure
config files will be loaded by the Clojure reader and will read-eval
forms if *read-eval*
is true. EDN config files, however, do not
support any notion of read-eval
. The eval middleware will still
evaluate it.
There are also some differences in how some forms are read. For
instance, {:foo '[a b c]}
will prevent an EDN config file from
loading but will load as {:foo (quote [a b c])}
in a CLJ config
file.
(defmethod carica/load-config :carica/yaml [resource]
(try
;; ...
(catch Throwable t
(log/warn t "error reading config" resource)
(throw
(Exception. (str "error reading config " resource) t)))))
To reuse the existing parsers for a different file extension:
(derive :carica/cfg :carica/edn)
Sometimes, during tests, it's handy to be able to override config values:
(with-redefs [config (override-config :db :password "swordfish")]
(config :db :password)
;;=> "swordfish"
(config :db :username))
;;=> "cosmo"
Or:
(with-redefs [config (override-config :db {:username "wagstaff"
:password "swordfish"})]
(config :db :password)
;;=> "swordfish"
(config :db :username))
;;=> "wagstaff"
Only the provided values will be overwritten.
Carica looks for resources on the classpath and merges them in order, meaning that configuration files earlier in the classpath will take precedence over those that come later.
When using Leiningen the classpath is built such that test comes first, then src, followed by resources-path. Keep the order in mind when dealing with multiple configuration files with Leiningen and when building the classpath by hand.
Copyright © 2012-2016 Sonian, Inc.
Distributed under the Eclipse Public License, the same as Clojure.
Can you improve this documentation? These fine people already did:
David Leatherman, Kurt Harriger, Joe Gallo, Matthew Daley, abp & Lee HinmanEdit on GitHub
cljdoc is a website building & hosting documentation for Clojure/Script libraries
× close