Documentation

Inotify represents an inotify descriptor, to which watches can be added and events can be read from. Internally, it also includes a buffer of events that have been delivered to the application from the kernel but haven't been processed.

Watch represents a watch descriptor, which is used to identify events and to cancel the watch. Every watch descriptor is associated with a particular inotify descriptor and can only be used with that descriptor; incorrect behavior will otherwise result.

Represents the mask, which in inotify terminology is a union of bit flags representing various event types and watch options.

The type parameter is a phantom type that tracks whether a particular flag is used to set up a watch (WatchFlag) or when receiving an event. (EventFlag) Polymorphic parameters mean that the flag may appear in either context.

Compute the intersection (bitwise and) of two masks

Are the bits of the first mask a subset of the bits of the second?

Do the two masks have any bits in common?

An empty type used to denote Mask values that can be sent to the kernel when setting up an inotify watch.

An empty type used to denote Mask values that can be received from the kernel in an inotify event message.

A newtype wrapper for the cookie field of the Event.

Creates an inotify socket descriptor that watches can be added to and events can be read from.

Closes an inotify descriptor, freeing the resources associated with it. This will also raise an IOException in any threads that are blocked on getEvent.

Attempting to use a descriptor after it is closed will safely raise an exception. It is perfectly safe to call close multiple times, which is idempotent and will not result in an exception.

Descriptors will be closed after they are garbage collected, via a finalizer, although it is often preferable to call close yourself.

Has the inotify descriptor been closed?

Creates an inotify socket descriptor with custom configuration options. Calls inotify_init1(IN_NONBLOCK | IN_CLOEXEC).

Additional configuration options for creating an Inotify descriptor.

Default configuration options

Adds a watch on the inotify descriptor, returns a watch descriptor. The mask controls which events are delivered to your application, as well as some additional options. This function is thread safe.

A variant of addWatch that operates on a RawFilePath, which is a file path represented as strict ByteString. One weakness of the current implementation is that if addWatch_ throws an IOException, then any unicode paths will be mangled in the error message.

Stops watching a path for changes. This watch descriptor must be associated with the particular inotify port, otherwise undefined behavior can happen.

This function is thread safe. This binding ignores inotify_rm_watch's errno when it is EINVAL, so it is ok to delete a previously removed or non-existent watch descriptor.

However long lived applications that set and remove many watches should still endeavor to avoid calling rmWatch on removed watch descriptors, due to possible wrap-around bugs.

Returns an inotify event, blocking until one is available.

If the inotify descriptor is closed, this function will return an event from the buffer, if available. Otherwise, it will throw an IOException.

It is safe to call this function from multiple threads at the same time.

Returns an inotify event only if one is immediately available.

If the inotify descriptor is closed, this function will return an event from the buffer, if available. Otherwise, it will throw an IOException.

One possible downside of the current implementation is that returning Nothing necessarily results in a system call.

It is safe to call this function from multiple threads at the same time.

Returns an inotify event only if one is available in 'Inotify'\'s buffer. This won't ever make a system call, and should not ever throw an exception.

It is safe to call this function from multiple threads at the same time.

Returns an inotify event, blocking until one is available.

After this returns an event, the next read from the inotify descriptor will return the same event. This read will not result in a system call.

If the inotify descriptor is closed, this function will return an event from the buffer, if available. Otherwise, it will throw an IOException.

It is safe to call this function from multiple threads at the same time.

Returns an inotify event only if one is immediately available.

If this returns an event, then the next read from the inotify descriptor will return the same event, and this read will not result in a system call.

If the inotify descriptor is closed, this function will return an event from the buffer, if available. Otherwise, it will throw an IOException.

One possible downside of the current implementation is that returning Nothing necessarily results in a system call.

It is safe to call this function from multiple threads at the same time.

Returns an inotify event only if one is available in 'Inotify'\'s buffer. This won't ever make a system call, and should not ever throw an exception.

If this returns an event, then the next read from the inotify descriptor will return the same event, and this read will not result in a system call.

It is safe to call this function from multiple threads at the same time.

File was accessed. Includes the files of a watched directory.

Metadata changed, e.g., permissions, timestamps, extended attributes, link count (since Linux 2.6.25), UID, GID, etc. Includes the files of a watched directory.

File was closed. This is not a separate flag, but a convenience definition such that in_CLOSE == in_CLOSE_WRITE <> in_CLOSE_NOWRITE

File opened for writing was closed. Includes the files of a watched directory.

File not opened for writing was closed. Includes the files of a watched directory.

File/directory created in watched directory.

File/directory deleted from watched directory.

Watched file/directory was itself deleted.

File was modified. Includes the files of a watched directory.

Watched file/directory was itself moved.

File was moved. This is not a separate flag, but a convenience definition such that in_MOVE == in_MOVED_FROM <> in_MOVED_TO.

File moved out of watched directory. Includes the files of a watched directory.

File moved into watched directory. Includes the files of a watched directory.

File was opened. Includes the files of a watched directory.

A union of all flags above; this is not a separate flag but a convenience definition.

(since Linux 2.6.15) Don't dereference pathname if it is a symbolic link.

(since Linux 2.6.36) By default, when watching events on the children of a directory, events are generated for children even after they have been unlinked from the directory. This can result in large numbers of uninteresting events for some applications (e.g., if watching /tmp, in which many applications create temporary files whose names are immediately unlinked). Specifying IN_EXCL_UNLINK changes the default behavior, so that events are not generated for children after they have been unlinked from the watched directory.

Add (OR) events to watch mask for this pathname if it already exists (instead of replacing mask).

Monitor pathname for one event, then remove from watch list.

(since Linux 2.6.15) Only watch pathname if it is a directory.

Watch was removed explicitly (rmWatch) or automatically (file was deleted, or file system was unmounted).

Subject of this event is a directory.

Event queue overflowed (wd is -1 for this event). The size of the queue is available at procsysfsinotify/max_queued_events.

File system containing watched object was unmounted.