polly

Types

A filesystem event.

Polly is careful when handing you events, and makes sure that you always get them in the right order, so if a dictionary with some files was created, you will get the dictionary event first, before any children.

pub type Event {
  Created(String)
  Changed(String)
  Deleted(String)
}

Constructors

  • Created(String)

    A new file or dictionary was created!

  • Changed(String)

    A file got modified!

  • Deleted(String)

    A file or dictionary got deleted :(

Some POSIX file system errors.

Only a handful of errors are handled by Polly, and everything else is collapsed into EUnknown.

pub type FsError {
  Eacces
  Enoent
  Enotdir
  EUnknown(String)
}

Constructors

  • Eacces
  • Enoent
  • Enotdir
  • EUnknown(String)

Polly uses the builder pattern to construct a watcher.

let options: Options = polly.new()
    |> polly.add_dir("src")
    |> polly.interval(3000)
    |> polly.max_depth(10)
    |> polly.filter(polly.default_filter)
pub opaque type Options

A polling file system watcher, built from options.

Watchers are automatically started, and can by stopped by calling stop.

pub opaque type Watcher(state)

Functions

pub fn add_dir(options: Options, path: String) -> Options

Tell Polly which directory to watch. If it does not exist, watch will return an error.

If the directory goes away after watching has started, Polly will continue to check on it to see if it came back.

Paths are not expanded by default, so the paths reported by events and passed to the filter function will be prefixed with whatever you specified here.

pub fn default_filter(path: String) -> Bool

The default filter function, ignoring hidden files starting with a colon "."

pub fn filter(
  options: Options,
  by filter: fn(String) -> Bool,
) -> Options

Filter files using the given predicate.

Polly will ignore files and directories for which the predicate returns False completely, and any event happening for them or for a contained file of them will not get reported.

By default, all hidden files are ignored.

pub fn interval(options: Options, interval: Int) -> Options

Set the interval in-between file-system polls, in milliseconds.

This is the time that Polly rests between calls, so if scanning your directory tree takes 100ms, and you configured 1000ms here, the total time between calls will roughly be 1100ms.

Doing it this way makes sure that Polly doesn’t stumble over herself.

pub fn max_depth(options: Options, max_depth: Int) -> Options

Limit the maximum depth that Polly will walk each directory.

A limit of 0 would mean that Polly only watches the specified list of files or directories. A limit of 1 means that she will also look at the files inside the given directories, but not at any nested directories.

There is no limit by default, but setting a limit might be a good to better control resource usage of the watcher.

pub fn new() -> Options

Start creating a new configuration using the default options.

By default, and interval of 1 second is set, and the default_filter is used.

pub fn stop(watcher: Watcher(a)) -> Nil

Stop this watcher.

If Polly currently scans your directories, she might not hear you right away and may still report events for one run, after which she will stop.

pub fn watch(
  options: Options,
  with callback: fn(Event) -> a,
) -> Result(Watcher(Nil), #(String, FsError))

Tell Polly to start watching all the specified directories for changes.

The callback is called synchronously after collecting all change events since the last run. It is adviseable to move heavier cpu-bound tasks from this callback into their own processes/threads.

On Erlang, the callback will be called in the same process as the watcher runs in.

pub fn watch_with(
  options: Options,
  from initial: a,
  with callback: fn(a, Event) -> a,
) -> Result(Watcher(a), #(String, FsError))

Like watch, but similar to list.fold Polly will also keep some state around for you and pass it back on each invocation.

Search Document