CHANGELOG.md

Change Log

All notable changes to this project will be documented in this file. This change log follows the conventions of keepachangelog.com.

Unreleased

Added

• The ability to provide most features with the Opus Quad, even though that hardware does not really support the Pro DJ Link protocol, by leveraging its ability to work with rekordbox lighting. Thanks to @cprepos for doing most of this work!
• The ability to proxy metadata from mounted archive files corresponding to USB media mounted in the Opus Quad, which cannot itself provide that information.

Fixed

• An error in interpreting database export file format by the Crate Digger library could lead to some rows that were actually present in tables not being found.

7.4.0 - 2024-05-04

May the Fourth be with you.

Fixed

• It turns out there is a bug in Java under Windows that can sometimes return null values in the list of network addresses associated with an interface. Beat Link now has defensive code to protect itself against this and avoid crashing.

Changed

• Replaced the dependency on ConcurrentLinkedHashMap with a much-simplified in-project implementation based on some suggestions kindly shared by Ben Manes, the author of that library, so that Beat Link Trigger can work as a pre-compiled GraalVM native-image, while retaining its backwards compatibility with Java 6 environments (like afterglow-max). Thanks to Noah Zoschke for the pull request that led to this change in direction.
• Upgraded to version 0.2.0 of Crate Digger, which had a couple of backwards-incompatible changes to accommodate a new linter in the latest release of the Kaitai Struct Compiler which identified places where field names were not following the style guide. It also supports creating metadata archives for use with the Opus Quad.
• Upgraded to building Java 11 class files (since the new version of Crate Digger does), so we can take advantage of new APIs.

7.3.0 - 2023-11-24

Added

• Precise position packets sent by CDJ-3000s are processed and used to keep track of the exact playback position of these players with much higher fidelity than is possible for other players, even when they are not currently playing.
• Recognition of dynamic loop information sent by CDJ-3000s, including display of such loops in the WaveformDetailComoponent. This means that even loops which do not exist within the track metadata can be displayed on CDJ-3000s, and for such players inactive loops can be drawn in gray rather than the usual orange.
• A new utility function to calculate a raw pitch integer value given a human-oriented pitch percentage value, because the new precise position packets sent by the CDJ-3000 use the humane approach.
• A convenience method on TrackPositionUpdate which returns the beat within bar represented by that track position.
• We can now recognize loaded menu items that represent a track title plus musical key.

Fixed

• We now send startup and keep-alive packets in a way that is compatible even with CDJ-3000s that are using player numbers 5 and 6 (they no longer display a warning dialog nor drop off the network).
• The calculation of pitch percentages from pitch values found in CDJ status packets was slightly inaccurate because of two transposed digits in a calculation.
• The code that detects pre-nexus CDJs was confused by the CDJ-3000.
• Sometimes beat packets can come after a status packet at the very beginning of a new beat, and this used to cause the beat number to jump up and back down.
• The clipping region was not being set correctly when painting phrase information at the bottom of the waveform detail component, which could cause the fill dots to extend beyond the edge of the component in certain circumstances.
• The code to clear out the device name buffer when setting a new device name for the VirtualCdj was using the wrong value for toIndex (I had thought the argument was a length instead), so long names were not being properly cleared out.

Changed

• The TrackPositionUpdate class has a new field, precise, which reports whether we are receiving precise position updates from the player.
• When determining the item type of a menu item response message, we now mask the menu item type (field 7) with 0xffff, because CDJ-3000s seem to send additional (as yet unknown) information in the higher-order bytes.

7.2.0 - 2023-08-13

Added

• The WaveformPreviewComponent and WaveformDetailComponent now have accessors allowing callers to obtain the underlying waveform information that is driving their display.

Fixed

• The WaveformPreviewComponent and WaveformDetailComponent were not correctly updating the color of their playback indicators if setPlaybackState was called with a change in nothing but the playing boolean.
• Trying to request track analysis file tags using DB Server queries (when using a real player number) was not working because the file extension was being passed without stripping off the leading "." as required by the protocol.

Changed

• The BeatGrid reports requests for out-of-range beats at debug level now, rather than warning, since they were quite noisy and meaningless in the log file.
• The getBarNumber() method in BeatGrid now follows the Rekordbox convention of reporting a value of -1 in any partial bar at the start of a track, and bar 1 always begins with the first downbeat.
• The method requestSongStructureFrom() in the AnalysisTagFinder was an anachronism, and has been corrected to requestAnalysisTagFrom().

7.1.0 - 2023-04-16

Added

• The ArtFinder now tries to find high-resolution album art when it is available (240x240 pixels rather than the original 80x80 pixels). This can be turned off by calling setRequestHighResolutionArt(false).
• Some additional details of the upcoming beat/bar values in the Beat object were clarified.

7.0.1 - 2022-05-30

Changed

• The BeatGrid class no longer throws an exception when it is asked for the offset of a nonexistent beat. Instead, it logs a warning, and returns the offset of the nearest beat.

7.0.0 - 2022-03-07

Added

• Support for channels 5 and 6 in the DJM-V10 channels-on-air messages, #47 for receiving, and #46 for sending.
• Support for obtaining song structure (phrase analysis) information from players, now that rekordbox 6 and later export it.
• Ability to display analyzed phrases on track waveforms.
• Support for obtaining any new type of track analysis which we can create definitions for in Crate Digger and obtain using the general track-analysis section dbserver request.
• The BeatGrid can now tell you the bar number associated with a beat number, accounting for the possibility of a partial bar at the start of the track.
• Information about the timing of upcoming beats and bars present within beat packets is now made available.

Removed

• The ability to create and use metadata caches has not been useful for a couple of years now, since we figured out how to use Crate Digger to reliably obtain track metadata even when there are a full set of real players on the network. So the complicated code which supported it has been removed, rather than trying to update it to keep up with new features like phrase analysis.

Changed

• The versioning numbering scheme has been made more reasonable. This is now version 7.0.0 instead of 0.7.0, to reflect the fact that there have been many releases in active production use.
• We now ensure that if the TimeFinder is running, it is informed about new beats received before any other registered beat listeners, so they can rely on it to provide the actual current beat number.

0.6.3 - 2020-12-28

Added

• The ability to set a player's My Settings configuration by sending a newly-discovered packet.
• The ability to set a background color for the WaveformDetailComponent and WaveformPreviewComponent, for use by the OBS overlay server in Beat Link Trigger. You can also now change the colors used to draw the playback position indicator and tick and beat marks to go with your overall look.

Fixed

• Some Unicode strings were not being properly read by Crate Digger.

0.6.2 - 2020-05-10

Fixed

• The interpretation of colors for memory points and loops (as opposed to hot cues) has been fixed, thanks to a contribution by @ehendrikd
• It turns out that nxs2 cue list entries are not always complete, they are sometimes missing color information or even the comment information that precedes it. We no longer crash when we encounter such partial cues.
• Fixes to the Beat Link and Crate Digger libraries allow them to work properly with new formats for data that rekordbox 6 sends. We may find more problems in the future, because testing with this new version has been limited, but it is already working much better than it did at first.
• An update to Crate Digger avoids crashes when trying to parse track analysis files created with mal-formed vestigial waveform preview tags.
• The interpretation of byte 0x37 in the CDJ status packet has been updated, which forced fixes of the isDiscSlotEmpty() and getDiscTrackCount() methods in CdjStatus.
• Links in the JavaDoc now directly take you to the relevant sections of the new Antora documentation site that replaced the old PDF protocol analysis document.
• Track Load packets now need to have an extra value to tell the recipient on which deck the track should be loaded, because the XDJ-XZ has two decks but only one network address on which to receive these commands. (Sadly for us, it refuses to load tracks from other players even when this byte is sent correctly, it only loads tracks from a rekordbox collection when rekordbox tells it to.)

Added

• Another bit has been explained in the main CdjStatus flags byte: we can now tell when a DJ has forced a player into degraded BPM-only sync mode (by nudging the jog wheel of a synced player).
• We now know that beat grid entries also report the tempo of the track at that beat, so this is properly exposed in the API.
• We now fully support the device number assignment phase of the Pro DJ Link protocol, so when Beat Link is connected to an ethernet port on a mixer that is assigned a dedicated channel number, it will use that device number, and we defend our device number once we have successfully claimed it.

Deprecated

• The classes and methods relating to creating and using metadata caches are no longer needed because we can reliably obtain metadata no matter how many CDJs are on the network.
• The getName and getNumber methods in DeviceAnnouncement were inconsistent with the getDeviceName and getDeviceNumber naming convention used in the entire device update packet hierarchy.

0.6.1 - 2020-02-09

Fixed

• The TimeFinder is no longer tricked by beat packets from pre-nexus players into thinking that it might know where in the track that player is. Track position detection is only possible with nexus and later hardware, because only they report beat numbers.
• When the VirtualCdj is forced to shut down because of apparent network changes, it now also flushes the DeviceFinder’s list of known DJ Link devices, because they are probably no longer reachable. This will allow for immediate recovery attempts by telling the VirtualCdj to restart itself (without this change, restarting it within ten seconds or so would fail because it would complain about being unable to communicate with the ghost devices on a no-longer reachable network, until they disappeared due to lack of recent packets).
• The crate-digger library was updated to fix a misunderstanding of the structure of cue list entries which could crash the parser.
• Our own parsing of nxs2 cue list entries fetched using the dbserver protocol is now robust against missing color bytes, which also seems to happen in the wild.
• Some nxs2 cue list entries are so short that they don't even have space for comments/labels, which we now handle properly too.

Added

• Error messages reported when parsing an ANLZ or EXT file fails now include the path to the file in the source media, to help find it for forensic analysis. This enabled the crate-digger fix mentioned above.
• A flag on the DeviceUpdate class that indicates whether it seems to have been sent by a pre-nexus player (CDJ-900 or CDJ-2000).

Changed

• No longer log a stack trace for the expected situation of a color waveform (or waveform preview) being unavailable. Also handle the case where it is reported available, but has zero size.
• Because the XDJ-1000 seems to calculate the payload size value for its status packets incorrectly (subtracting the header size twice) we now only report that one time, to stop flooding the logs with warnings for each status packet received.

0.6.0 - 2019-11-24

Added

• Support for the XDJ-XZ, which reports multiple devices on a single IP address. This broke assumptions in many places in the code, and required a more sophisticated approach to device matching. Thanks to patient and detailed reports, experiments, packet captures, and videos from Teo Tormo.
• The TrackPositionListener now always sends an initial track position update to newly registered listeners, even if the update is null (meaning nothing is known) to help them get initialized. This is especially useful if their registration is being moved to a different player.

Fixed

• A race condition which caused CrateDigger to retain incorrect track information when a player was removed from the network while it had a mounted USB, then returned to the network with a different USB, has been resolved. Thanks to @ben-xo for discovering and reporting this!
• Players report themselves as stopped when they reach the end of the track playing forwards, but they don't when they reach the start playing backwards. This caused weird display in the Player Status window, and weird generated time code with expressions. The TimeFinder now compensates for this buggy behavior by reporting playback as stopped when it is stuck at 0 milliseconds.

Changed

• tempoChanged events are no longer sent to the MasterListener interface when the tempo is meaningless (i.e. the new master has no track loaded).

0.5.5 - 2019-10-25

Fixed

• CDJs which were powered on after Beat Link was already running would not ever get assigned a valid dbserver port by the ConnectionManager so the MetadataFinder would not be able to perform metadata requests to them without CrateDigger. This seems to have been caused by the players not being quite ready to respond to the port number query right after booting, so we now try again a few times.
• Some extended cue entries encountered by users were missing the color bytes, which prevented Crate Digger from parsing EXT files and accessing color waveforms. These values are now treated as optional.
• Eliminated spurious warnings in the log for cue entries with rekordbox color code 0 which also had explicit green color values stored for nxs2 players.
• Now builds properly under current JDKs, including Amazon Corretto 11 (which is a long-term support release). The minimum JDK for building is now Java 9, but the resulting build is still compatible back to Java 1.6. Building under Java 11 results in much nicer JavaDoc, with search support.

0.5.2 - 2019-09-02

Added

• CueList entries can now include nxs2-style DJ comments when they are present, along with colors and hot cues beyond C.
• Helper functions to search CueList entries for the closest cue before or after a specific time in a track.
• The WaveformPreviewComponent can return its CueList for the convenience of code that wants to add informative tool tips.
• The WaveformDetailComponent draws labels for hot cues and for any cue or loop that has been assigned a comment.
• MountListener instances registered with the MetadataFinder will be examined to see if they also implement MediaDetailsListener. If they do, they will be informed when details are available for newly-mounted media.

0.5.1 - 2019-03-05

Fixed

• The SignatureFinder would crash trying to calculate signatures for tracks without artists.

Added

• The MenuLoader now supports loading tracks from the Label, Bit Rate, Original Artist, and Remixer menus.
• Track Metadata now includes bit rate, when applicable.

0.5.0 - 2019-02-23

Changed

⚠️ These are breaking API changes. Code that used some cache-related and metadata methods will need to be rewritten.

• The handling of metadata cache files has been moved out of the MetadataFinder into a new class focused on this task. It also implements a new interface that can be used by client objects to offer cached metadata to Beat Link (for example track cue lists that group metadata from multiple different media sources).
• Track comments and dates added are just plain strings, not searchable items, so the API has been changed to reflect this.
• WaveformPreviewComponent and WaveformDetailComponent have been generalized to support tracking multiple simultaneous playback positions, so they can be used in the context of a Beat Link Trigger show file, where a track can be loaded and even playing on multiple players at once.
• It is now possible to construct a fully-functional WaveformPreviewComponent or WaveformDetailComponent even if you don’t have access to actual TrackMetadata objects, as long as you pass them the few individual pieces of information they need. Again, this supports the needs of the new Beat Link Trigger show interface.

Added

• We now can use Crate Digger to reliably obtain metadata even when there are four players on the network, and we are using player number 5.
• We can now retrieve and display the full-color waveforms used by nxs2 players and rekordbox, and prefer them when available. This can be turned off through a WaveformFinder property.
• When going online, the VirtualCdj now reports information about the chosen network interface to help with troubleshooting problematic network environments.
• There is a new interface, TrackPositionBeatListener, that can be registered with the TimeFinder in order to learn about new beats as soon as they happen, along with the actual beat number within the track represented by the beat (which is missing from the raw beat packet available through the BeatFinder).
• The TimeFinder takes advantage of track metadata, (when it is available) to detect when the DJ has jumped to a cue, or when a track has loaded with auto-cue to a memory point, so it can more accurately infer where the player has stopped.
• MenuLoader supports loading several more menu types from the dbserver.
• We can offer track years as part of the metadata when rekordbox is indexing that.
• When disconnecting from a player’s dbserver, we politely send a teardown message and let it close the connection from its side.
• A few new methods that can help troubleshoot problematic network topologies are available. Beat Link Trigger uses them to provide better help at startup.

Fixed

• A mistake in the TimeFinder which could trap the VirtualCdj packet delivery thread in an infinite loop, which would then cause client applications to lose touch with what was happening on the players, with no recourse except going offline and back online (and even this would gradually waste more and more CPU time on each occurrence). Thanks to @Kevinnns for the thread dumps which helped finally find this!
• The code that was supposed to notice computers running rekordbox and report their media collections as being available on the network had always been broken, but was easy enough to fix. Now Beat Link Trigger can tell players to load tracks from rekordbox and rekordbox mobile.
• The various metadata finders were supposed to report the loss of associated metadata when a player disappeared from the network, but they were not doing so.
• At some point (possibly when adding the ability to show multiple player locations in a waveform) the memory point and hot cue markers stopped being displayed in the waveform preview. They have been restored.
• Cue Lists were not being sorted into the correct order when they were loaded from Crate Digger. Now they are sorted regardless of how we obtain them.
• The colors used for playback position markers were inconsistent between the waveform preview and waveform detail (one used red when the other used white for playing/stopped). Now they both use the constants defined in the WaveformDetailComponent, which also adds a bit of transparency to the preview markers.
• It turns out that rekordbox mobile does not report meaningful values for its media name or creation date, so we no longer attempt to parse them.
• It seems rekordbox sometimes sends mixer status packets with a different subtype structure which was causing us to log warnings about a length mismatch. We now properly recognize this packet subtype.

0.4.1 - 2018-10-28

Added

• Metadata caches now store the details of the media from which they were created to more easily and reliably auto-attach them and survive small changes.

Fixed

• When a Nexus player is reporting that it is preloading hot cues, we no longer incorrectly consider it to be playing.
• When scanning metadata caches to consider them for auto-attachment, we were not closing the ones which failed to match. This was probably eventually being taken care of by the garbage collector and finalizers, but there is no guarantee if or when that would actually happen.
• When handling a Fader Start command, a missing break statement caused the BeatFinder to log warning which incorrectly reported that the command had not been recognized. This no longer happens.

0.4.0 - 2018-10-07

Added

• Metadata can now be retrieved for non-rekordbox tracks, including unanalyzed files in the media slots, and on audio CDs and data discs.
• The mounting and removal of discs is reported to registered media mount listeners.
• Details about all mounted media slots can be discovered, including the name assigned in rekordbox, creation date, number of rekordbox tracks and playlists (and whether there is a rekordbox database at all, which is needed to correctly request the root menu for that slot), size, and free space.
• The number of tracks on a disc can be discovered.
• The VirtualCdj can send status and beat packets, simulate playing, and become the tempo master.
• The VirtualCdj has new methods allowing you to tell other devices to turn sync mode on or off, or to become the tempo master.
• The VirtualCdj has a new method allowing you to tell other players to start or stop playing.
• The VirtualCdj has a new method allowing you to tell other players to load a particular track from any rekordbox database on the network (in a player's media slot, or on a laptop running rekordbox).
• The VirtualCdj has a new method allowing you to tell players whether they are on or off the air in the absence of a DJM mixer.
• The Message.KnownType enum has many new entries describing menu requests we now know how to perform.
• The Message.MenuItemType enum has new entries for the menus that can appear in the root menu response.
• A new class, MenuLoader, provides methods for navigating the menu hierarchies served by players.
• A new Enum which captures all known packet types, to improve the readability and compactness of code that works with them.

Fixed

• Waveform details are properly loaded for tracks that are found on multiple players on the network. This affected time-remaining calculations too.
• Now assembles entire dbserver messages into a buffer to write them to the network as a single operation, avoiding the chance of them being split into multiple packets, because Windows rekordbox can't handle when they are.
• Removed a potential source of crashes in the waveform rendering code.
• Improved the clarity of an exception thrown when trying to ask for metadata from a player for which we did not find a db server port.
• Improved protection against problems that can occur when delivering events (status updates, beats, etc.) to registered listeners; previously we were only catching exceptions, but some kinds of problems in the listener classes would lead to other kinds of throwables, which could kill our event delivery loops.
• The count parameter is now passed in menu render requests in a way that is more consistent with the way CDJs do it, although this has not seemed to cause any problems.

Changed

• We no longer simply reject packets with unexpected lengths; if they are longer than the minimum value we expect, we try to process them after logging a warning. If they are too short, we just log the warning.
• More use of ByteBuffers to efficiently assemble and compare packets.
• The TRACK_LIST_REQ message type has been renamed TRACK_MENU_REQ to fit in with the large number of other menus that have been added.

0.3.7 - 2018-03-26

Fixed

• Although the TrackMetadata object received the Album field, and would display it in the toString() method, there was no accessor by which the value could be obtained and used by other code!

0.3.6 - 2017-11-30

Fixed

• When looping a track that has audio data that extends well past the final beat in the beat grid, players sometimes report playing a beat that does not exist in the beat grid. This previously caused an exception in the log, and the reported playback position would keep growing without bound as long as the loop continued. This situation is now handled better by interpolating missing beats at the end of the beat grid, so there is no exception and the looping of the player is properly reflected by the TimeFinder.

0.3.5 - 2017-10-08

Fixed

• Creating a metadata cache would fail if a playlist contained more than one copy of the same track, as it would try to create the same ZIP file entry more than once. Now redundant copies of the same track are skipped.

0.3.4 - 2017-09-05

Added

• Exposed the function which opens and returns a metadata cache file, so that other projects can explore their contents, for example to offer a view of what they contain.

0.3.2 - 2017-08-08

Added

• More logging about the detection of media being mounted and removed from players, to help track down problems creating metadata caches.

Fixed

• There was a potential for the thread that reports media being mounted or removed to terminate if any exceptions occurred while handling player updates. It now logs these and continues.

0.3.1 - 2017-07-22

Added

• The TimeFinder now keeps a list of listeners that want help tightly following track position changes, for example to generate SMPTE time code that is tied to the current playback position.

0.3.0 - 2017-06-25

Changed

• The entire metadata retrieval approach has been fundamentally rewritten, based on a much deeper understanding of the message and field format, thanks to tips and sample code from @awwright.
• This led to a deep refactoring of the library, and a split into several packages, now that there are many more classes to support all of these new features. API compatibility was not maintained with previous releases.

Added

• Many more kinds of information can be retrieved from players, and the classes and API documentation have been enriched to reflect our deepening understanding of the protocol.
• Metadata cache files can be created by downloading either entire collections or individual playlists from players on the network, and then attached for use in environments (such as busy shows with four physical players) where metadata queries are not reliable.
• Metadata cache files can be monitored for automatic attachment to player ports when media that matches the cache is mounted by the player.
• New listener interfaces to provide information about changes to the available metadata caches, as well as changes to metadata about tracks loaded in players on the network.

Fixed

• The approach to concurrency was examined carefully, and almost all synchronized methods were eliminated in favor of newer concurrent classes for improved performance and safety.
• When the last device disappears from the Pro DJ Link network, we clear the DeviceFinder's notion of when we first saw one, so that the VirtualCDJ's device-number conflict-avoidance code can work properly when we later encounter a new network.

0.2.1 - 2017-03-14

Fixed

• Several issues which prevented Beat Link from reliably detecting all other players on the network when self-assigning the number to use for the VirtualCDJ. First, keep track of when we actually received our first device announcement packet, rather than just when we started trying to find the Pro DJ Link network, because that provides a much more meaningful indication of how long to wait for players on the network. Second, remove the mutual dependency between the DeviceFinder and the VirtualCDJ, because the latter is in a synchronized method during startup, which was preventing the DeviceFinder from being able to actually process any device announcement packets which came in while the VirtualCDJ was trying to wait for them!

Added

• A configuration option to tell the VirtualCDJ to try to use a standard player number when self-assigning.
• When stopping the VirtualCDJ, it sets its device number to 0 so that the default at startup is always self-assignment, to avoid conflicts. You can still set a specific number before calling start if you want to force a device number.
• Support new, slightly larger CDJ status packets from new Pioneer firmware.
• Allow you to request all metadata for a given media slot.

0.2.0 - 2016-12-11

Added

• The ability to retrieve track metadata (artist, title, etc.) from the CDJs, thanks to discoveries by @EvanPurkhiser. See his prolink-go project for a Go language version of a library like beat-link, with no need for a Java virtual machine.
• The ability to get track artwork images.

0.1.9 - 2016-11-09

Changed

• The values returned by the getTimestamp() method on device update objects have been changed to be based on System.nanoTime() rather than System.currentTimeMillis(). This makes them easier to work with as relative times, and also makes them compatible with the time values used by Ableton Link, for better interoperability.

0.1.8 - 2016-07-04

Added

• Initial support for NXS2 players. Will now recognize status update packets, even though they are larger than from other players, and can recognize all the state flags we already know about. We still don't know how to interpret any new features represented in the new section of the packet.
• Now uses slf4j to abstract over whatever logging framework is being used by projects that embed beat-link.
• Can now obtain track metadata (title and artist, so far), from the CDJs! Not yet reliable or safe, however, so not documented.
• As part of that, the rekordbox track id and source player / slot information have been found in the CDJ status packets. This part is safe to use, and beat-link-trigger is happily doing so.

0.1.7 - 2016-05-31

Added

• Unless you assign an explicit deviceNumber to VirtualCdj, it will watch the DJ Link network looking for an unused player number between 5 and 15 to assign itself when it starts up.

0.1.6 - 2016-05-27

Fixed

• The byte being used to determine device number did not work for beat packets, so Beat objects always returned zero from getDeviceNumber().

0.1.5 - 2016-05-25

Added

• A new method, DeviceFinder.getLatestStatus() that returns the most recent status updates received from all active devices.

Changed

• Device updates, beat announcements, and master announcements are time-sensitive, so they are now delivered directly on the thread that is receiving them from the network, rather than being added to the Event Dispatch Queue. This will reduce latency, but means listener methods need to be very fast, and delegate any lengthy, non-time-sensitive operations to another thread themselves.
• The threads which receive device updates and beat announcements now run at maximum priority, also to reduce latency. Device announcements are less time-sensitive, so those are still received at normal priority, and sent on the event dispatch thread, to make it easier to update user interface elements.

0.1.4 - 2016-05-18

Fixed

• Accept shorter, 208-byte CDJ status packets sent by non-nexus players.
• The value of F seems to always be zero for non-nexus players, so we need to use a different way to check playing status for them.
• The value of B_b is not meaningful for non-rekordbox tracks and non-nexus players, so reflect that in CdjStatus.isBeatWithinBarMeaningful().
• The values reported for P₂ are different for non-nexus players.

Changed

• VirtualCdj.start() now returns a boolean value to indicate whether the attempt was successful, with false meaning that no DJ Link devices could be found.
• DeviceWatcher now updates its device lists before posting lost/found announcements, so recipients can see the results as a whole.

Added

• The firmware version number has been found in the CDJ status packets, so it is now available through CdjStatus.getFirmwareVersion().
• When the beat number within the track is not meaningful (because a non-rekordbox track is being played, or a non-nexus player sent the update), CdjStatus.getBeatNumber() now returns -1.
• A new method DeviceFinder.getLatestAnnouncementFrom() and a new overload of VirtualCdj.getLatestStatusFor() which take a device (player) number and look for matching announcement or status reports.

0.1.3 - 2016-05-12

Fixed

• Device Update objects (including Beats) were missing the first character of the device name.
• The VirtualCdj socket was being bound to the wildcard address rather than the address of the interface on which DJ-Link traffic had been detected.
• Resolved a crash when stopping the VirtualCdj because it was calling one of its own methods after marking itself inactive.
• During startup, the VirtualCdj waits longer for DJ-Link traffic before giving up.

Added

• Device updates have a boolean property which identifies whether you can expect getBeatWithinBar() to give you meaningful results.
• The virtual device we create to communicate with other DJ-Link devices is now filtered out from the list of devices we see on the network, since you don’t ever want to interact with it.

0.1.2 - 2016-05-09

Fixed

• The CdjStatus object was looking at the wrong byte for B_b so getBeatWithinBar() was always returning 0.

Changed

• Trying to call methods that require the DeviceFinder or VirtualCdj to be running in order to give you valid results will now throw an IllegalStateException if those objects are not running, rather than giving you meaningless results.

0.1.1 - 2016-05-08

Added

• Initial release of the full implementation, including BeatFinder, VirtualCdj and the various listener interfaces and status objects they provide.

0.1.0 - 2016-05-05

Added

• Initial early release of DeviceFinder.