watchr.coffee | |
|---|---|
# | Watchr is used to be nofitied when a change happens to, or within a directory.
You will not be notified what file was changed, or how it was changed.
It will track new files and their changes too, and remove listeners for deleted files appropriatly.
The source code here is written as an experiment of literate programming
Which means you would be able to understand it, without knowing code |
# | |
| Require the file system module for node.js This provides us with what we need to interact with the file system (aka files and directories) | fs = require('fs') |
| Let's set the debugging mode We will use this later on when outputting messages that make our code easier to debug Note: when we publish the module, we want to set this as off, as we don't want the application using us to spurt our all our debug messages! | debug = false |
| Now to make watching files more convient and managed, we'll create a class which we can use to attach to each file It'll provide us with the API and abstraction we need to accomplish difficult things like recursion We'll also store a global store of all the watchers and their paths so we don't have multiple watchers going at the same time for the same file - as that would be quite ineffecient | watchers = {}
Watcher = class |
| The path this class instance is attached to | path: null |
| Is it a directory or not? | isDirectory: null |
| Our fs.stat object, it contains things like change times, size, and is it a directory | stat: null |
| The events that we will trigger when we detect a change We make this as an array as otherwise we would have to have one listener for every event as that would be quite slow So instead we have one listener, with many events | events: [] |
| The node.js file watcher instance, we have to open and close this, it is what notifies us of the events | fswatcher: null |
| We also want to setup a delay between change events, as if we changed a lot of files, we just want to be notified once, instead of like 1000 times | timeout: null
delay: 500 |
| The watchers for the children of this watcher will go here This is for when we are watching a directory, we will scan the directory and children go here | children: [] |
| We have to store the current state of the watcher and it is asynchronous (things can fire in any order) as such, we don't want to be doing particular things if this watcher is deactivated (closed) | state: 'pending' |
| The method we will use to watch the files Preferably we use watchFile, however we may need to use watch in case watchFile doesn't exist (e.g. windows) | method: null
|
| Now it's time to construct our watcher We give it a path, and give it some events to use Then we get to work with watching it | constructor: (path,events=[]) ->
@children = []
@events = []
@path = path
@addEvents events
fs.stat @path, (err,stat) =>
return if @state is 'closed'
throw err if err
@stat = stat
@isDirectory = stat.isDirectory()
@watch()
|
| Let's now add our events We should support being passed a list of events, as well as one event, or no events (for error prevention) We should also not add an event if it is already added | addEvents: (events) ->
unless events instanceof Array
if events
events = [events]
else
events = []
for event in events
found = false
for storedEvent in @events
if storedEvent is event
found = true
break
if not found
@events.push(event)
@
|
| Before we start watching, we'll have to setup the functions our watcher will need | |
| It will need function to trigger all of our watcher's events | trigger: ->
console.log "trigger: #{@path}" if debug
for event in @events
event()
@
|
| We also need something so when a file is changed, we wait a while until things have calmed down and once they have calmed down, then trigger our events | changed: ->
console.log "changed: #{@path}" if debug
if @timeout
clearTimeout(@timeout)
@timeout = false
@timeout = setTimeout(
=> @trigger()
@delay
)
@
|
| We will need something to close our listener for removed or renamed files As renamed files are a bit difficult we will want to close and delete all the watchers for all our children too Essentially it is like a self-destruct without the body parts | close: ->
return @ if @state is 'closed'
console.log "close: #{@path}" if debug |
| Close and destroy children | for watcher,index in @children
watcher.close()
delete @children[index]
@children = [] |
| Close ourself | if @state isnt 'closed' |
| Close listener | if @method is 'watchFile'
fs.unwatchFile @path
else if @method is 'watch' and @fswatcher
@fswatcher.close()
@fswatcher = null
|
| Updated state | @state = 'closed' |
| Delete our watchers reference | delete watchers[@path] if watchers[@path]? |
| Chain | @
|
| We need something to figure out what to do when a file is changed It will check if we are still active, and if so, then Handle the events appropriatly | handler: (action) -> |
| Ignore if we are closed | return if @state is 'closed'
|
| Using fs.watchFile | if arguments.length is 2 and typeof arguments[0] isnt 'string'
console.log "watchFile: #{@path}" if debug
curr = arguments[0]
prev = arguments[1]
console.log arguments if debug
return @ if curr.mtime.getTime() is prev.mtime.getTime() or curr.size is prev.size
@changed()
@watch()
return @
|
| Using fs.watch | console.log "#{action}: #{@path}" if debug |
| Renames and new files If we are a file then stop our close our watcher, as an event will also have fired for the parent directory If we are the parent directory, then trigger our change event, then re-initialise all our listernes, as we want to close listerners for deleted files and add new listeners for added files | if action is 'rename'
if @isDirectory is false
@close()
else
@changed()
try
@watch()
|
| Changed files If we were a change, then let's check that something did actually change If it did, then trigger our change event | else if action is 'change'
fs.stat @path, (err,stat) => |
| Ignore if we are closed | return if @state is 'closed'
throw err if err
return if stat.mtime.getTime() is @stat.mtime.getTime() and stat.size is @stat.size
@stat = stat
@changed()
|
| Chain | @
|
| Setup the watching for our path If we are already watching this path then let's start again (call close) Then if we are a directory, let's recurse Finally, let's initialise our node.js watcher that'll let us know when things happen and update our state to active | watch: ->
@close()
console.log "watch: #{@path}" if debug
if @isDirectory
fs.readdir @path, (err,files) =>
throw err if err
for file in files
continue if /^[\.~]/.test file
filePath = @path+'/'+file
watcher = watch filePath, => @changed()
@children.push watcher
try
fs.watchFile @path, => @handler.apply(@,arguments)
@method = 'watchFile'
catch err
@fswatcher = fs.watch @path, => @handler.apply(@,arguments)
@method = 'watch'
@state = 'active'
@ |
| Provide our interface to the applications that use watchr This will create our new Watcher class for the path we want (or use an existing one, and add the events) Watcher also uses this too | watch = (path,events) -> |
| Check if we are already watching that path | if watchers[path]? |
| We do, so let's use that one instead | watchers[path].addEvents(events)
else |
| We don't, so let's create a new one | watchers[path] = new Watcher(path,events) |
| Now let's provide node.js with our public API In other words, what the application that calls us has access to | module.exports = {watch}
|