watchpack

Wrapper library for directory and file watching.

Concept

watchpack high level API doesn't map directly to watchers. Instead a three level architecture ensures that for each directory only a single watcher exists.

• The high level API requests DirectoryWatchers from a WatcherManager, which ensures that only a single DirectoryWatcher per directory is created.
• A user-faced Watcher can be obtained from a DirectoryWatcher and provides a filtered view on the DirectoryWatcher.
• Reference-counting is used on the DirectoryWatcher and Watcher to decide when to close them.
• The real watchers are created by the DirectoryWatcher.
• Files are never watched directly. This should keep the watcher count low.
• Watching can be started in the past. This way watching can start after file reading.
• Symlinks are not followed, instead the symlink is watched.

API

constWatchpack=require("watchpack");constwp=newWatchpack({// options:aggregateTimeout: 1000,// fire "aggregated" event when after a change for 1000ms no additional change occurred// aggregated defaults to undefined, which doesn't fire an "aggregated" eventpoll: true,// poll: true - use polling with the default interval// poll: 10000 - use polling with an interval of 10s// poll defaults to undefined, which prefer native watching methods// Note: enable polling when watching on a network path// When WATCHPACK_POLLING environment variable is set it will override this optionfollowSymlinks: true,// true: follows symlinks and watches symlinks and real files// (This makes sense when symlinks has not been resolved yet, comes with a performance hit)// false (default): watches only specified item they may be real files or symlinks// (This makes sense when symlinks has already been resolved)ignored: "**/.git",// ignored: "string" - a glob pattern for files or folders that should not be watched// ignored: ["string", "string"] - multiple glob patterns that should be ignored// ignored: /regexp/ - a regular expression for files or folders that should not be watched// ignored: (entry) => boolean - an arbitrary function which must return truthy to ignore an entry// For all cases expect the arbitrary function the path will have path separator normalized to '/'.// All subdirectories are ignored too});// Watchpack.prototype.watch({// files: Iterable<string>,// directories: Iterable<string>,// missing: Iterable<string>,// startTime?: number// })wp.watch({files: listOfFiles,directories: listOfDirectories,missing: listOfNotExistingItems,startTime: Date.now()-10000,});// starts watching these files and directories// calling this again will override the files and directories// files: can be files or directories, for files: content and existence changes are tracked// for directories: only existence and timestamp changes are tracked// directories: only directories, directory content (and content of children, ...) and// existence changes are tracked.// assumed to exist, when directory is not found without further information a remove event is emitted// missing: can be files or directories,// only existence changes are tracked// expected to not exist, no remove event is emitted when not found initially// files and directories are assumed to exist, when they are not found without further information a remove event is emitted// missing is assumed to not exist and no remove event is emittedwp.on("change",(filePath,mtime,explanation)=>{// filePath: the changed file// mtime: last modified time for the changed file// explanation: textual information how this change was detected});wp.on("remove",(filePath,explanation)=>{// filePath: the removed file or directory// explanation: textual information how this change was detected});wp.on("aggregated",(changes,removals)=>{// changes: a Set of all changed files// removals: a Set of all removed files// watchpack gives up ownership on these Sets.});// Watchpack.prototype.pause()wp.pause();// stops emitting events, but keeps watchers open// next "watch" call can reuse the watchers// The watcher will keep aggregating events// which can be received with getAggregated()// Watchpack.prototype.close()wp.close();// stops emitting events and closes all watchers// Watchpack.prototype.getAggregated(): { changes: Set<string>, removals: Set<string> }const{ changes, removals }=wp.getAggregated();// returns the current aggregated info and removes that from the watcher// The next aggregated event won't include that info and will only emitted// when futher changes happen// Can also be used when paused.// Watchpack.prototype.collectTimeInfoEntries(fileInfoEntries: Map<string, Entry>, directoryInfoEntries: Map<string, Entry>)wp.collectTimeInfoEntries(fileInfoEntries,directoryInfoEntries);// collects time info objects for all known files and directories// this include info from files not directly watched// key: absolute path, value: object with { safeTime, timestamp }// safeTime: a point in time at which it is safe to say all changes happened before that// timestamp: only for files, the mtime timestamp of the file// Watchpack.prototype.getTimeInfoEntries()constfileTimes=wp.getTimeInfoEntries();// returns a Map with all known time info objects for files and directories// similar to collectTimeInfoEntries but returns a single map with all entries// (deprecated)// Watchpack.prototype.getTimes()constfileTimesOld=wp.getTimes();// returns an object with all known change times for files// this include timestamps from files not directly watched// key: absolute path, value: timestamp as number

Environment variables

• WATCHPACK_POLLING: when set, overrides the poll option (see above).
• WATCHPACK_RETRIES: number of times to retry fs.lstat when it returns EBUSY (default: 3). Useful on Windows where anti-virus scanners, indexers or editors briefly lock files — without retries watchpack would see a spurious remove and stop tracking the file. Set to 0 or "false" to disable retrying.