<scherz>
Think about Zelda: when hens are free to propagate
everywhere, they attack you and it becomes a mess. Your application
state is like your hens: it's safe when it is securely contained in a
coop with automated doors to prevent run-aways.</scherz>
This small Clojure library tries to get as close as possible to the
bare essential complexity of state management in code. It defines a
new type of Clojure map that you may (.close m)
. <scherz>
See it in
action above</scherz>
.
This library defines a new type of map that you may use like any other
map. This map may contain stateful Java objects like a server, a Kafka
producer, a file output stream. When you want to clean your state, you
just .close
the map and all nested stateful objects will be closed
recursively.
It is a tiny alternative to more capable projects:
- Application state management in a map: juxt/clip
- Simple library to initialise an app stack: sathyavijayan/upit
- Application state management: stuartsierra/component, weavejester/integrant, tolitius/mount, et al.
- Extension of
with-open
: jarohen/with-open - Representing state in a map: robertluo/fun-map
In your project, require:
(require '[piotr-yuxuan.closeable-map :as closeable-map :refer [close-with with-tag close!]])
Define an application that can be started, and closed.
(defn start
"Return a map describing a running application, and which values may
be closed."
[config]
(closeable-map/closeable-map
{;; Kafka producers/consumers are `java.io.Closeable`.
:producer (kafka-producer config)
:consumer (kafka-consumer config)}))
You can start/stop the app in the repl with:
(comment
(def config (load-config))
(def system (start config))
;; Stop/close all processes/resources with:
(.close system)
)
It can be used in conjunction with with-open
in test file to create
well-contained, independent tests:
(with-open [{:keys [consumer] :as app} (start config)]
(testing "unit test with isolated, repeatable context"
(is (= :yay/🚀 (some-business/function consumer)))))
You could also use thi library while live-coding to stop and restart your application whenever a file is changed.
(defn start
"Return a map describing a running application, and which values may
be closed."
[config]
(closeable-map/closeable-map
{;; Kafka producers/consumers are `java.io.Closeable`.
:producer (kafka-producer config)
:consumer (kafka-consumer config)
;; File streams are `java.io.Closeable` too:
:logfile (io/output-stream (io/file "/tmp/log.txt"))
;; Closeable maps can be nested. Nested maps will be closed before the outer map.
:backend/api {:response-executor (close-with (memfn ^ExecutorService .shutdown)
(flow/utilization-executor (:executor config)))
:connection-pool (close-with (memfn ^IPool .shutdown)
(http/connection-pool {:pool-opts config}))
;; These functions receive their map as argument.
::closeable-map/before-close (fn [m] (backend/give-up-leadership config m))
::closeable-map/after-close (fn [m] (backend/close-connection config m))}
;; Any exception when closing this nested map will be swallowed
;; and not bubbled up.
:db ^::closeable-map/swallow {;; Connection are `java.io.Closeable`, too:
:db-conn (jdbc/get-connection (:db config))}
;; Some libs return a zero-argument function which when called
;; stops the server, like:
:server (with-tag ::closeable-map/fn (http/start-server (api config) (:server config)))
;; Gotcha: Clojure meta data can only be attached on 'concrete'
;; objects; they are lost on literal forms (see above).
:forensic ^::closeable-map/fn #(metrics/report-death!)
::closeable-map/ex-handler
(fn [ex]
;; Will be called for all exceptions thrown when closing this
;; map and nested items.
(println (ex-message ex)))}))
When (.close system)
is executed, it will:
-
Recursively close all instances of
java.io.Closeable
andjava.lang.AutoCloseable
; -
Recursively call all stop zero-argument functions tagged with
^::closeable-map/fn
; -
Skip all nested
Closeable
under a^::closeable-map/ignore
; -
Silently swallow any exception with
^::closeable-map/swallow
; -
Exceptions to optional
::closeable-map/ex-handler
in key or metadata; -
If keys (or metadata)
::closeable-map/before-close
or::closeable-map/after-close
are present, they will be assumed as a function which takes one argument (the map itself) and used run additional closing logic:(closeable-map {;; This function will be executed before the auto close. ::closeable-map/before-close (fn [this-map] (flush!)) ;; Kafka producers/consumers are java.io.Closeable :producer (kafka-producer config) :consumer (kafka-consumer config) ;; This function will be executed after the auto close. ::closeable-map/after-close (fn [this-map] (garbage/collect!))})
Some classes do not implement java.lang.AutoCloseable
but present
some similar method. For example instances of
java.util.concurrent.ExecutorService
can't be closed but they can be
.shutdown
:
{:response-executor (close-with (memfn ^ExecutorService .shutdown)
(flow/utilization-executor (:executor config)))
:connection-pool (close-with (memfn ^IPool .shutdown)
(http/connection-pool {:pool-opts config}))}
A Java object does not implement interface clojure.lang.IObj
so it
is unable to carry Clojure metadata. As such, you can't give a it a
tag like ::closeable-map/fn
. You may also extend this library by
giving new dispatch values to multimethod close!
. Once evaluated,
this will work accross all your code. The multimethod is dispatched on
the concrete class of its argument:
(import '(java.util.concurrent ExecutorService))
(defmethod close! ExecutorService
[x]
(.shutdown ^ExecutorService x))
(import '(io.aleph.dirigiste IPool))
(defmethod close! IPool
[x]
(.shutdown ^IPool x))
(import '(clojure.lang Atom))
(defmethod close! Atom
[x]
(reset! x nil))
(import '(clojure.core.async.impl.protocols Channel))
(defmethod close! Channel
[x]
(async/close! x))
(import '(xtdb.node XtdbNode))
(defmethod c/close! XtdbNode
[^Closeable x]
(doto x
xt/sync
(.close)))
A Java object may be wrapped as a closeable*
.
You may also avoid partially open state when an exception is thrown
when creating a CloseableMap
. This is where closeable-map*
comes
handy. It outcome in one of the following:
-
Either everything went right, and all inner forms wrapped by
closeable
correctly return a value; you get an open instance ofCloseableMap
. -
Either some inner form wrapped by
closeable
didn't return a closeable object but threw an exception instead. Then allcloseable
forms are closed, and finally the exception is bubbled up.
(closeable-map*
{:server (closeable* (http/start-server (api config)))
:kafka {:consumer (closeable* (kafka-consumer config))
:producer (closeable* (kafka-producer config))
:schema.registry.url "https://localhost"}})
Closeable objects not directly within the map may still be closed if
wrapped in closeable*
:
(closeable-map*
(closeable* (kafka-consumer config))
(closeable* (kafka-producer config))
{:server (closeable* (http/start-server (api config)))
:schema.registry.url "https://localhost"})
In some circumstances you may need to handle exception on the creation of a closeable map. If an exception happens during the creation of the map, values already evaluated will be closed. No closeable objects will be left open with no references to them.
For instance, this form would throw an exception:
(closeable-map/closeable-map {:server (http/start-server (api config))
:kafka {:consumer (kafka-consumer config)
:producer (throw (ex-info "Exception" {}))}})
;; => (ex-info "Exception" {})
with-closeable*
prevents that kind of broken, partially open states for its bindings:
(with-closeable* [server (http/start-server (api config))
consumer (kafka-consumer config)
producer (throw (ex-info "Exception" {}))]
;; Your code goes here.
)
;; Close consumer,
;; close server,
;; finally throw `(ex-info "Exception" {})`.
You now have the guarantee that your code will only be executed if
all these closeable are open. In the latter example an exception is
thrown when producer
is evaluated, so consumer
is closed, then
server
is closed, and finally the exception is bubbled up. Your
code is not evaluated. In the next example the body is evaluated,
but throws an exception: all bindings are closed.
(with-closeable* [server (http/start-server (api config))
consumer (kafka-consumer config)
producer (kafka-producer config)]
;; Your code goes here.
(throw (ex-info "Exception" {})))
;; Close producer,
;; close consumer,
;; close server,
;; finally throw `(ex-info "Exception" {})`.
When no exception is thrown, leave bindings open and return like a
normal let
form. If you prefer to close bindings, use with-open
as
usual.
(with-closeable* [server (http/start-server (api config))
consumer (kafka-consumer config)
producer (kafka-producer config)]
;; Your code goes here.
)
;; All closeable in bindings stay open.
;; => result
Some Clojure datastructures implement IFn
:
({:a 1} :a) ;; => 1
(remove #{:a} [:a :b :c]) ;; => '(:b :c)
([:a :b :c] 1) ;; => :b
Clojure maps (IPersistentMap
) implement IFn
, for invoke()
of one
argument (a key) with an optional second argument (a default value),
i.e. maps are functions of their keys. nil
keys and values are fine.
This library defines a new data strucure, CloseableMap. It is exposed
as an instance of java.io.Closeable
which is a subinterface of
java.lang.AutoCloseable
. When trying to close its values, it looks
for instances of the latter. As such, it tries to be most general.
(require '[clojure.data])
(clojure.data/diff
(ancestors (class {}))
(ancestors CloseableMap))
;; =>
[;; Ancestors of Clojure map only but not CloseableMap.
#{clojure.lang.AFn ; Concrete type, but see below for IFn.
clojure.lang.APersistentMap
clojure.lang.IEditableCollection
clojure.lang.IKVReduce
clojure.lang.IMapIterable
java.io.Serializable}
;; Ancestors of CloseableMap only.
#{clojure.lang.IType
java.io.Closeable
java.lang.AutoCloseable
java.util.Iterator
potemkin.collections.PotemkinMap
potemkin.types.PotemkinType}
;; Ancestors common to both types.
#{clojure.lang.Associative
clojure.lang.Counted
clojure.lang.IFn
clojure.lang.IHashEq
clojure.lang.ILookup
clojure.lang.IMeta
clojure.lang.IObj
clojure.lang.IPersistentCollection
clojure.lang.IPersistentMap
clojure.lang.MapEquivalence
clojure.lang.Seqable
java.lang.Iterable
java.lang.Object
java.lang.Runnable
java.util.Map
java.util.concurrent.Callable}]