Interfacing to MIDI (sequencer) devices via ALSA on Linux

[Fred vS]

It works with vmpk too ! 

That's odd... it depends on whether vmpk's showing only an "in" or both an "In" and an "Out"... JP's example and your working ones are the latter case. Most likely a bug in vmpk, which isn't encouraging.

MarkMLl

In fact, I get all I need with vkeybd, I did have a little bemol about the number of octave displayed  (default = 3) but there is a parameter to define it (see picture):

Code: Bash  [Select][+][-]

1. > vkeybd --octave 5

So, for me, I am happy with vkeybd.

Hum, new things are not always better:

Code: Bash  [Select][+][-]

1. > vkeybd --help
2. vkeybd -- virtual keyboard
3.   version 0.1.18d
4. Copyright (c) 1997-2000 Takashi Iwai

Code: Bash  [Select][+][-]

1. > vmpk -h
2. VMPK v0.8.7
3. Virtual MIDI Piano Keyboard
4. Copyright (C) 2006-2022 Pedro Lopez-Cabanillas

[Jurassic Pork]

Hello Mark,
Have a look to ChannelSelector and LazRtmidi  which use rtmidi library :

A set of C++ classes that provide a common API for realtime MIDI input/output across Linux (ALSA & JACK), Macintosh OS X (CoreMIDI & JACK) and Windows (Multimedia).

The pascal  Source code of these projects can help you.
Friendly, J.P

[MarkMLl]

Hello Mark,
Have a look to ChannelSelector and LazRtmidi  which use rtmidi library :

A set of C++ classes that provide a common API for realtime MIDI input/output across Linux (ALSA & JACK), Macintosh OS X (CoreMIDI & JACK) and Windows (Multimedia).

The pascal  Source code of these projects can help you.
Friendly, J.P

Apart from the declarations I don't see anything about setting up a filter :-/

As a general point I'd remark that I'm trying to implement this interface to be a straight transcription of the ALSA headers, since that provides the best chance of subsequently being able to ask the developers for help without being bluntly told that we should be using a supported language.

MarkMLl

[Jurassic Pork]

for filters with rtmidi you can see here but it is in python language 

[Fred vS]

Hello Fred,
Strange, for VMPK you have no output with aconnect -l command. Under Ubuntu 20.04 , it seems ok for me :

ubuntu$aconnect -l
client 0: 'System' [type=noyau]
    0 'Timer           '
    1 'Announce        '
client 14: 'Midi Through' [type=noyau]
    0 'Midi Through Port-0'
client 128: 'FLUID Synth (8160)' [type=utilisateur,pid=8160]
    0 'Synth input port (8160:0)'
client 129: 'VMPK Input' [type=utilisateur,pid=4584]
    0 'in              '
client 130: 'VMPK Output' [type=utilisateur,pid=4584]
    0 'out             '
ubuntu$aconnect 130:0 128:0
ubuntu$aconnect -l
client 0: 'System' [type=noyau]
    0 'Timer           '
    1 'Announce        '
client 14: 'Midi Through' [type=noyau]
    0 'Midi Through Port-0'
client 128: 'FLUID Synth (8160)' [type=utilisateur,pid=8160]
    0 'Synth input port (8160:0)'
   Connecté Depuis: 130:0
client 129: 'VMPK Input' [type=utilisateur,pid=4584]
    0 'in              '
client 130: 'VMPK Output' [type=utilisateur,pid=4584]
    0 'out             '
   Connexion À: 128:0

i have installed VMPK with synaptic some other packages have been installed (qsynth, soundfont etc...)

Friendly, J.P

Ooops, I did not see this post! 

Salut J.P.

Hum, indeed strange.
I noted that there is a kind of random with aconnect -l and VMPK.
Some time I get value for  'VMPK Output' , some time it does not appear.

I will use vkeybd for the test, there no problems.

Fre;D

[MarkMLl]

It appears not to work if vmpk doesn't present an output as well as an input port.

Code: [Select]

client 129: 'VMPK Input' [type=utilisateur,pid=4584]
    0 'in              '
client 130: 'VMPK Output' [type=utilisateur,pid=4584]
    0 'out             '

I don't know whether to blame vmpk or something flaky in the ALSA API :-/

MarkMLl

[Fred vS]

It appears not to work if vmpk doesn't present an output as well as an input port.

Code: [Select]

client 129: 'VMPK Input' [type=utilisateur,pid=4584]
    0 'in              '
client 130: 'VMPK Output' [type=utilisateur,pid=4584]
    0 'out             '

I don't know whether to blame vmpk or something flaky in the ALSA API :-/

MarkMLl

I propose to trust the ALSA API and forget vmpk for the moment.

[Fred vS]

If you edit GetFilteredEvents() you can select event types which /should/ get filtered out: the fact that I've not worked out how to do that is one of the issues.

I'm watching, but if I go quiet for a day or so it's because some "legal crap" has arrived which has to be dealt with promptly.

MarkMLl

Hello Mark.

I did try with SND_SEQ_EVENT_NOTE and all seems to work in the terminal amidi -p virtual -d. 

Code: Pascal  [Select][+][-]

1. procedure GetFilteredEvents(out ets: seq_event_type_set);
2. begin
3. // Default for debugging.
4. //  ets := [SND_SEQ_EVENT_CONTROLLER] ;
5.  
6.   ets := [SND_SEQ_EVENT_NOTE]  ; // here changed
7. end ;

Code: Bash  [Select][+][-]

1. ~> amidi -p virtual -d
2.  
3. 903C7F
4. 803C00
5. 903C7F
6. 803C00
7. 90477F
8. 804700
9. 90477F
10. 804700
11. 90477F

[MarkMLl]

Yes, but what you're seeing there is that the note events are still going through to amidi, rather than being filtered and potentially dumped as hex by the filter program... try setting a breakpoint in the filter procedure.

MarkMLl

[Fred vS]

Yes, but what you're seeing there is that the note events are still going through to amidi, rather than being filtered and potentially dumped as hex by the filter program... try setting a breakpoint in the filter procedure.

MarkMLl

Indeed.
But I am sure you will find the trick how to do it. 

Fre;D

[MarkMLl]

Yes, but what you're seeing there is that the note events are still going through to amidi, rather than being filtered and potentially dumped as hex by the filter program... try setting a breakpoint in the filter procedure.

MarkMLl

Indeed.
But I am sure you will find the trick how to do it. 

Fre;D

Well I certainly haven't yet :-(

MarkMLl

[Jurassic Pork]

with python and the mido library it is easy to do a filter. Here is a python script to create a virtual port and wait for notes messages to display them :

Code: Python  [Select][+][-]

1. import mido
2. def accept_notes(port):
3.     """Only let note_on and note_off messages through."""
4.     for message in port:
5.         if message.type in ('note_on', 'note_off'):    
6.             yield message
7. """Create a New Midi Virtual Port and wait for filtered messages"""        
8. with mido.open_input('New Port', virtual=True) as inport:
9.     for message in  accept_notes(inport):
10.         print(message)

When i change the channel on the virtual midi keyboard the message doen't display.

[MarkMLl]

I have less than zero interest in doing it in Python, and from what I can see that doesn't use the ALSA filtering facility which is what I'm asking for help with.

MarkMLl

[Jurassic Pork]

Code: Pascal  [Select][+][-]

1. procedure GetFilteredEvents(out ets: seq_event_type_set);
2. begin
3.   ets := [SND_SEQ_EVENT_CONTROLLER]
4. end { GetFilteredEvents } ;
5.  
6. function ApplyFilter(var seq: Psnd_seq_t; var event: snd_seq_event_t): boolean;
7. begin
8.   result := true;
9.   WriteLn(event.type_:3);
10. end { ApplyFilter } ;
11. // ...
12.          if ApplyFilter(seq, ev^) then begin
13. //            snd_seq_ev_set_source(ev, my_port);
14.             AsoundSeq.snd_seq_ev_set_subs(ev);
15.             AsoundSeq.snd_seq_ev_set_direct(ev);
16.             AsoundSeq.snd_seq_event_output(seq, ev);
17.             AsoundSeq.snd_seq_drain_output(seq)
18.           end
19. // ...
20.  

in the ApplyFilter function you can add a parameter of type seq_event_type_set , you can check if the event is in the set and if not return false ?

[MarkMLl]

in the ApplyFilter function you can add a parameter of type seq_event_type_set , you can check if the event is in the set and if not return false ?

ApplyFilter() is not being called. My suspicion is that it's incompatible with subscriptions (which are handled internal to ALSA itself), but the documentation doesn't say what it /is/ compatible with.

MarkMLl