On the image's left side is the color input as Firefox implements it. On the right side the color input of Chrome. (Tested on Linux)

Obviously Chrome's version is as bad as just hacking the hex-code into a text input. No color picker and not even saturation/value adjustment controls.

On my laptop, importing a new album sometimes takes a lot of time (~1minute).
This is too long to not show a progress bar.

• Identify what steps is so time consuming
• Show a progress bar for musicdb add

The "related-songs view" should low-light songs that do not match the selected genre.
The may also be moved to the end of the list.

Hey !

It could be very nice if you can make a Docker image for your project, trying step just for testing your project and it takes a long time to install all dependancies, all python dependancies.. etc :)

Docker image are very fun to easily deploy but still more fun to simply launch and test in 10 seconds ! :)

Regards,

Missing documentation for SetAlbumColor

2019-05-17 18:27:20.748099: F tensorflow/stream_executor/lib/statusor.cc:34] Attempting to fetch value instead of handling error Failed precondition: could not dlopen DSO: libcublas.so.10.1; dlerror: libcublas.so.10.1: cannot open shared object file: No such file or directory

The search function shows all songs it finds.
At least for the quick preview all hated and deactivated songs shall be invisible.

In the full results list they may appear (with low dominance), but maybe they should also be invisible.
At least the deactivated ones.

Exported files that get converted to mp3 have no release year ID3 frame.
Using id3edit (the tool used to set the ID3 Tag) seems not be the problem.
There may be a problem with generating the parameter lists for the id3edit call.

• Did I change the command line interface for setting the release date when I moved to id3edit version 2?
• Did I change the Frame ID of the Frame so that it is no more standard conform (or now it is, and my audio player not?)

Target Device: A Sony Android smartphone using Sonys audio player software

Adding artwork for an album fails, if the artwork path is not absolute.
In the following example, both calls should work, but only the second one succeeds.

cd ~
musicdb artwork -u --artwork tmp.jpg --album /data/music/Artist/2020\ -\ Album
musicdb artwork -u --artwork /home/user/tmp.jpg --album /data/music/Artist/2020\ -\ Album

The error appers in mdbapi/artwork.py:321: if artworkpath and not self.fs.IsFile(artworkpath):
The root for fs is / and therefore relative paths are interpreted as being relative to the systems root directory.
This is correct, because the the command line modules should make all user defined paths absolute before calling MusicDB internal functions.
So mod/artwork.py fails at this point.

Issue

Currently, I use an FlySpray instances running on my private server to track bugs, features and development progress.
The concepts behind the GitHub Issue tracker does not fulfill my requirements so that I really do not want to use it.
I need to find an alternative solution.

Possible Solutions

Some thought on how to solve the problem.

External FlySpray Provider

Just paying for a ready to use FlySpray instance would be my favorite solution.
Sadly there are no provider for ready to use and easy to maintain public FlySpray installations.

External Web Hoster

Paying for some public web space I can run my own FlySpray installation on can be an option.
The problem is the effort for maintaining the installation.
I have to take care to update FlySpray as soon as possible whenever there is an update instead of just blocking the port in the Firewall and update when I have time.
So I have to spend money and time.

External Provider of a Different Bug Tracking System

There may be more than FlySpray. :)

GitHub Issues Tracker

I could just use this Bug Tracker.
Disadvantages are, that it is hard to manage bugs and features.
Especially the fact that the UI appends A LOT of overhead with every change to an Issue (Edit tags, move to different Milestones) is a big downside.
GitHub's Issue Tracker focus lays on fancy GUI, not on the content. That sucks.
Otherwise I could use the Labels to reimplement "Task Types", "Categories", "Difficulty" and "Urgency".

The add module of MusicDB marks invalid song names.
I thought it will automatically clean those names. Of cause it does nothing as long as the user does not explicitly gives the command to do something (in this case pressing c to give the command to clean the names).
Even worse is the fact, that MusicDB crashes when forgetting to press c:

Traceback (most recent call last):
  File "/srv/musicdb/musicdb", line 208, in <module>
    exitcode = modobj.MDBM_Main(args)
  File "/srv/musicdb/mod/add.py", line 824, in MDBM_Main
    self.RunImportProcess(data)
  File "/srv/musicdb/mod/add.py", line 756, in RunImportProcess
    self.AddAlbum(newalbumpath, artist["id"])
  File "/srv/musicdb/mdbapi/database.py", line 531, in AddAlbum
    raise AssertionError("Analysing path \"%s\" failed!", songpaths[0])
AssertionError: ('Analysing path "%s" failed!', 'Die Kreatur/2020 - Panoptikum/Die Kreatur - Panoptikum - 01 DIE KREATUR.mp3')

Problem 1: Make sure the user is aware that the song-name-remarks are indicating problems with the naming, and not an applied solution.
Problem 2: If the user leaves those problems unsolved, a message box should appear allowing the user to solve the issues, or cancel the import process.
Problem 3: The exception listed above should not be unhandled!

Currently, related songs are presented as they appear in the database.
This gets ugly for favorite songs with lots of relations.
Grouping the songs by artist or at least albums will improve the UX.

The documentation for WebSockets on Java Script side needs an update. https://rstemmer.github.io/musicdb/build/html/webui/websockets.html

Some information regarding the configuration are outdated

Artworks cause a significant part of the traffic between MusicDB Server and WebUI Client.
Furthermore a high quality is crucial for good user experience.
Therefore JPEG XL could have a huge impact to the quality of MusicDB and its front end.

• Wait for FireFox support
• Implement JPEG XL support on server side

The FileSystem class does not handle ~ well.
os.path.expanduser will fix this.

Add Drag'n'Drop support to add songs from "Related Songs" list into "Song Queue".
It should be possible to place them at any position in the queue.

Both Views already use the same div-tiles to represent songs. It looks similar and so it should behave.

The css files require some updates to make MusicDB look good on 4k screens.
With 5.1.1 I already fixed the font sizes.
There are some more issues to fix:

• Font Size
• FlexBox layout
• Image sizes
• Use as much relative units as possible

/srv/musicdb/musicdb:20: DeprecationWarning: the imp module is deprecated in favour of importlib; see the module's documentation for alternative uses

Just a warning, but needs to be fixed in near future

via https://twitter.com/whiteblue3/status/1057449660682387457

Possible Implementation
Before sending the audio data to icecast, push them through a second gstreamer pipeline.
This second pipeline can be used to transcode the stream to any final encoding icecast can handle.
Further more it may be possible to make icecast optional and directly use that pipeline for streaming.

New added album does not automatically appear in WebUI until the artists-list gets updated (for example by triggering the genre selection).

After a new Album/Artist/Song got added or removed from the music collection, the WebUI should update its state to be in sync with the server again.

The mechanics already exists and only need to be applied as an event of a music collection update.

When the WebUI is reloaded, it connects successfully.
But the watchdog gets not reset correctly by the heat beat.
Only after the watchdog forces a reconnect, it gets reset with every heart beat.

This is barely notable if not looking at the logs and so it is not critical.

I just skipped some songs that got added to the empty queue by the Random Song Selection Algorithm (Randy).
I skipped them, because I did not like them, even if the song fits perfectly the selected genre (and any other criteria a valid song shall fulfill).
Goal of Randy is to provide a song selection that "feels good". This objective failed at this point.

Possible Improvements

It is important that the improvements do not have any disadvantages regarding discovering less frequently or never played songs.
Also songs with lots of dislikes should still have a chance to get added to the queue because liking/disliking songs is very mood-depending.

Select a favorite after several skipped songs

After a certain amount of "next song" commands or "dislikes", Randy shall add a song that got faved by the user.
The amount of actions until the new constraint counts should be configurable.

Trace the voting and maintain a positive average

Randy could trace each songs voting (like/dislikes) that got played.
A gliding average (configurable) should not be lower than a certain threshold (also configurable).
Is the average vote below the threshold a faved song should be added to the queue at "next" position. Then the average can be reset.
When a song gets liked/dislikes while it is playing, the vote should also be considered.
So a dislike has direct impact to the "quality" of the next songs.

When songs are bought at amazon they are full of "[explicit]" substrings.
The import module should automatically remove them (using the clear-function of muscdb add).
Furthermore the ID3 frames COMM and PRIV can be used to identify the origin of the song.

I just came around the following pages:

• Rename mp3-metadata-interval for clarity
• Config: Rename mp3-metadata-interval to icy-metadata-interval

Implement a way to hide albums.

• Add column in albums table to mark albums as "hidden".
• Improve API to filter hidden albums
• Improve WebUI to hide albums
• How to make hidden albums visible again? CLI-Module?

Reasons why users may want to hide albums:
The albums are interesting for "export" situations. For example albums the user only listens to when sitting in the car.

In the documentation of the StreamManager I warn about calling SongQueue.NextSong directly and not via the StreamManager.
This situation should also be mentioned in the SongQueue Class.

Furthermore there should be a reason given. It may not be obvious that the StreamManager needs to be informed to start streaming a different file, and that the SongQueue does not do this.

Missing documentation for GetSongLyrics

Missing documentation for SetSongLyrics

When adding a new genre tag to an album, all its songs should be updated by a suggestion of the same genre.
This should only be applied to songs, that are not tagged at that moment.
The tag should be a "low confident" tag that needs to be confirmed by the user later.

This procedure can be seen as a replacement for MusicAI

There are still some historic columns in music.db and music.db.sql: qskips, qadds, qremoves.

The preview of the results of the search in the WebUI would be much nicer with a blurred transparent background. (Frosted glass effect)

Solution:

backdrop-filter: blur(2px);

Source: developer.mozilla.org

Problem:

Sadly, this feature is not yet implemented - bugzilla.mozilla.org

1. Hard to read code
2. Code needs to be split into several files
3. Does not take care about song meta data
4. Does not automatically handle multi-CD-situations
5. Hard to extend

[M45ch1n3/pts/4:/srv/musicdb] python
Python 3.8.1 (default, Dec 21 2019, 20:57:38) 
[GCC 9.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import tflearn
2020-01-11 14:33:59.393571: I tensorflow/stream_executor/platform/default/dso_loader.cc:44] Successfully opened dynamic library libcudart.so.10.2
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/lib/python3.8/site-packages/tflearn/__init__.py", line 4, in <module>
    from . import config
  File "/usr/lib/python3.8/site-packages/tflearn/config.py", line 5, in <module>
    from .variables import variable
  File "/usr/lib/python3.8/site-packages/tflearn/variables.py", line 7, in <module>
    from tensorflow.contrib.framework.python.ops import add_arg_scope as contrib_add_arg_scope
ModuleNotFoundError: No module named 'tensorflow.contrib'
>>> 

Installed tensorflow:

• python-tensorflow-cuda 2.1.0-1
• tensorflow-cuda 2.1.0-1
• tflearn (0.3.2)

The tflearn module was installed via pip 9.3 (python 3.8).

Issue

The current update process is far too complicated.
Currently there are too many situation the user has to do changes manually.

Solution

I want to provide a singe script (update.sh) to update MusicDB in a way that the user does not have to do more than just executing the script.

Steps to the Solution

• Separate installation (install.sh) and updating (update.sh)
• Organize Databases and configuration files in a way to detect its current version and the steps needed to update them automatically.
• Developing a MusicDB Module to allow updates that cannot be done by an external script (and call it from the update script)
• Test the update and installation process
  • Arch Linux (Raspberry Pi)
  • Arch Linux (ODroid C2)
  • Debian (AMD64 VM)
• Update documentation

(Changes are done on the sprint-summer2018 branch)

Requirements

• This should be the last time that a user has to change things in his setup manually
• Changes should not destroy previous installations

The random song selection algorithm inserted a song from the same artist the song before was (added by me) into the queue

Problem: The randomizer does not track songs added by the user. So that artist was not on the blacklist. This was implemented by intention, but obviously was a bad decision.

Solution: Also add user selection to the blacklists.

• Move blacklist handling from Randy into different class/file
• Update Randy/SongQueue classes

Issue

"Too many open files" after a very long run time.

Log

 LOG[2018-10-07 21:34:49|FATAL] server.py:Run.343: FATAL ERROR (shutting down server!!):
 LOG[2018-10-07 21:34:49|FATAL] server.py:Run.344: [Errno 24] Too many open files: '/data/musicdb/musicdb.fifo'
Traceback (most recent call last):
  File "/srv/musicdb/mdbapi/server.py", line 326, in Run
  File "/srv/musicdb/lib/namedpipe.py", line 93, in ReadLine
  File "/srv/musicdb/lib/namedpipe.py", line 91, in opener
OSError: [Errno 24] Too many open files: '/data/musicdb/musicdb.fifo'

After a fresh install, the WSAPIKEY in config.js and musicdb.ini is not set.
The script should generate a key.

Uncaught ReferenceError: WSAPIKEY is not defined
    <anonymous> http://127.0.0.1/musicdb/webui/config.js:8

Some artist, album or song names are all-caps.
The JS code that generates the HTML for those names should detect this and add a CSS class to the DIVs that increases the letter-spacing a bit.

Change "Foreground" and "Highlight" label for colors to "Primary" and "Secondary" color.
This would be closer to the way how those two colors are used. Actually they are used this way.

This change should not only be applied to the WebUI but also to the backend and the database.

Post Scriptum / Fun-Fact: The name of the three colors of the WebUI (Background, Foreground, Highlight) come from the very first version of MusicDB. From the day, right after executing mkdir musicdb.
It's never too late to fix bad decisions 😄

Adding tags, especially mood-tags are a bit ugly because of the complexity behind the tag management.
A command line "GUI" like for musicdb-repair will be a benefit.

Issue

Overkill-Template

The current website is based on a random template I found on the internet.
I striped a lot of overhead from the site, but there is still far more code that necessary for a static website. This makes it uncomfortable to edit the page.

• I simply have to write a bare html/css page. No big deal
• Better screenshots (see next section)

Worst Screenshots Ever

For copyright reasons I have to blur all album covers on the screenshots.
I need to find lots of creative commons album covers I can use.
All tries to find some suitable album artworks ended up detecting copyright violations by the artists of those covers which again makes it impossible to use them.
This is a serious problem.

• I need at least 20 free artworks
• 5 fake artist names
• 20 fake album names
• 30 fake song names

Dynamic reloading of sub genre tags when updating the main genre must be implemented.
This should be similar to how song tags will be handled inside the Album View

The implementation of handling the Autobahn event loop broke with the update to Python 3.7 / Autobahn 18.8.1.

Pure asyncio event loop

 LOG[2018-08-25 09:03:42|DEBUG] websocket.py:__init__.40: Using WebSocket module AutobahnPython/18.8.1
 LOG[2018-08-25 09:03:42|INFO ] server.py: MusicDB WebSocket Server (TLSv1.2 secured) listening to port 9001
 LOG[2018-08-25 09:03:42|INFO ] server.py: Setup complete. Executing server.
 LOG[2018-08-25 09:03:42|FATAL] server.py:Run.343: FATAL ERROR (shutting down server!!):
 LOG[2018-08-25 09:03:42|FATAL] server.py:Run.344: cannot reuse already awaited coroutine
Traceback (most recent call last):
  File "/srv/musicdb/mdbapi/server.py", line 324, in Run
    tlswsserver.HandleEvents()
  File "/srv/musicdb/lib/ws/server.py", line 175, in HandleEvents
    self.eventloop.run_until_complete(self.coro)
  File "/usr/lib/python3.7/asyncio/base_events.py", line 568, in run_until_complete
    return future.result()
RuntimeError: cannot reuse already awaited coroutine

With uvloop as event loop

 LOG[2018-08-21 18:13:24|DEBUG] websocket.py:__init__.40: Using WebSocket module AutobahnPython/18.8.1
 LOG[2018-08-21 18:13:24|INFO ] server.py: MusicDB WebSocket Server (TLSv1.2 secured) listening to port 9001
 LOG[2018-08-21 18:13:24|INFO ] server.py: Setup complete. Executing server.
 LOG[2018-08-21 18:13:24|FATAL] server.py:Run.344: FATAL ERROR (shutting down server!!):
 LOG[2018-08-21 18:13:24|FATAL] server.py:Run.345: cannot reuse already awaited coroutine
Traceback (most recent call last):
  File "/srv/musicdb/mdbapi/server.py", line 325, in Run
    tlswsserver.HandleEvents()
  File "/srv/musicdb/lib/ws/server.py", line 181, in HandleEvents
    self.eventloop.run_until_complete(self.coro)
  File "uvloop/loop.pyx", line 1448, in uvloop.loop.Loop.run_until_complete
RuntimeError: cannot reuse already awaited coroutine

Use the new logo for the WebUI's favicon.

Is it possible to differentiate between light and dark themes?

This bug is most likely only in the develop-version.
I just leave the report here to not forget about it.

p.s.
I need to create a label for "next generation issues"

musicdb upgrade should upgrade the database, but musicdb fails because when opening the database, it detects that it is outdated.

When a song name has a slash, its file name uses a Division Slash (U+2215), because Slash (U+002F) is reserved as path separator.
For example 08 Mensch∕Maschine.mp3.
Using a monospace font, the is no notable visual difference between a Slash and a Division Slash. But rendering the name in the WebUI, the kerning information are notable different.
Therefore it is necessary to replace the Division Slash before rendering the font name:
Mensch∕Maschine -> Mensch/Maschine.

This can be done in the WebUI on the fly.
I thought I did this in the past. Maybe there is something broken in the Java Script, or it never left the state of "I should do it".

Another solution is to do this on importing the song. Then the "correct" name is already stored in the database. Doing it in the backend may lead to a problem, because the songname-entry mismatches the song name part of the path-entry. The database was originally only used to cache filesystem information.
Following this strategy requires reviewing all parts of MusicDB that interact with the database and the filesystem. This is additional beneficial to improve stability and reliability of the code.

At the moment I prefer to follow both strategies. On-the-fly replacement in the WebUI, and replacing during import.