A Clojure library designed to watch files and directories.
Like most clojure file watchers, hawk wraps the JDK 7 java.nio.file.WatchService API. This works great when Java has a native implementation for your platform of choice. However, this is not the case on OS X, so hawk also wraps the Barbary WatchService to provide performant monitoring even if you're using a mac. An appropriate implementation is chosen automatically, so all you have to do is install hawk and relax.
Hawk also features:
- A facility to handle "stateful" watches
- The ability to filter watches, and some handy built-in filters
- A name that is also a kind of bird
To install, add the following dependency to your project.clj file:
To get started with using hawk, require hawk.core in your project:
(ns hawk.sample
(:require [hawk.core :as hawk]))
To start a simple watch:
(hawk/watch! [{:paths ["src/main/hawk"]
:handler (fn [ctx e]
(println "event: " e)
(println "context: " ctx)
ctx)}])
The :handler function is passed both the group's context value and the event being handled, and is expected to return the new context value. The event hash has the following structure:
{:kind :create ;; the event kind, one of: #{:create :modify :delete}
:file #java.io.File{} } ;; the (canonicalized) affected file
To start a simple filtered watch, use :filter:
(hawk/watch! [{:paths ["src/main/hawk"]
:filter hawk/file?
:handler (fn [_ _]
(println "look ma, just files!"))}])
The :filter function is also passed both the group's context and event being filtered. Hawk has some built-in filters, like hawk/file?, but any predicate function will work.
To start a stateful watch, use :context to initialize the state, and then just return the updated state from the :handler. In this example, the context is initialized to 1 when the watch is started, and then incremented whenever an event is handled:
(hawk/watch! [{:paths ["src/main/hawk"]
:context (constantly 1)
:handler (fn [ctx _] (inc ctx))}])
The :context function is passed only the group's current context, and is expected to return the new context value.
Hawk supports multiple watches with a single call to watch!. watch! accepts an arbitrary number of arrays, which can hold an arbitrary number of watch specs. Every spec in an an array will share the same state, and events are processed sequentially. Every array gets its own state, and events are processed (potentially) in parallel:
(hawk/watch!
;; here we pass 2 groups to watch!
[{:paths ["src/main/hawk"]
;; for the first group, context is set to 1...
:context (constantly 1)
:handler (fn [ctx _]
(println "I'm always first!")
(inc ctx))}
{:paths ["src/test/hawk"]
;; ...and then incremented to 2 by the second watch in the group
:context (fn [ctx] (inc ctx))
:handler (fn [_ _]
(println "I'm always second place."))}]
;; here is the second watch group
[{:paths ["src/main/hawk"]
;; its context (3) is totally separate from the above group
:context (constantly 3)
:handler (fn [_ _]
(println "I'm also always first!"))}])
There are cases (such as within virtual environments) where the operating system will not receive events when a file has changed. Hawk provides a polling watcher as a fallback mechanism for handling these cases:
(hawk/watch! {:watcher :polling}
[{:paths ["src/main/hawk"]
:handler (fn [ctx e]
(println "event: " e)
(println "context: " ctx)
ctx)}])
You may also provide a :sensitivity argument of :high (the default), :medium, or :low depending on how often you want polling to occur.
To stop a watch, use hawk/stop!:
(let [watcher (hawk/watch! [...])]
(hawk/stop! watcher))
Happy watching!
Copyright (c) 2015 Will Farrell
Distributed under the Eclipse Public License either version 1.0 or (at your option) any later version.