In case there's interest, I'm developing a cython csound wrapper: cycsound

[shakfu]

Hi, In case this is useful somehow, I started work on a cython wrapper for the csound.h api a little while ago called cycsound.

It's still early stage, but so far it wraps a small subset of the api with audio output. It compiles (dynamic linking) on macOS, windows, and linux, with full static linking on macOS, which means it can currently generate a self-contained 2MB python wheel for macOS (which was my initial driver to start this project).

Reading the excellent ctcsound.py source has been enormously helpful and I've been using it as a development guide.

I'll keep chipping away... it helps me to learn csound, but any feedback would be welcome (-:

[fggp]

Beware that in Csound7 (probably released in next September) the host API will be redefined. Maybe you should wait for this new API before going on. François Le sam. 22 juin 2024 à 23:32, Shakeeb Alireza ***@***.***> a écrit :

…

Hi, In case this is useful somehow, I started work on a cython wrapper for the csound.h api a little while ago called cycsound <https://github.com/shakfu/cycsound>. It's still early stage, but so far it wraps a small subset of the api with audio output. It compiles (dynamic linking) on macOS, windows, and linux, with full static linking on macOS, which means it can currently generate a self-contained 2MB python wheel for macOS (which was my initial driver to start this project). Reading the excellent ctcsound.py source has been enormously helpful and I've been using it as a development guide. I'll keep chipping away... it helps me to learn csound, but any feedback would be welcome (-: — Reply to this email directly, view it on GitHub <#25>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABNOAKN5SIX7YSYRUIUZHA3ZIXUPNAVCNFSM6AAAAABJXXYIEOVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM3DQMBTGIYTCMY> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[shakfu]

@fggp Thanks for the heads up...

I'll keep an eye out for changes in the api...

[shakfu]

@fggp It's not clear to me where changes in the api are occurring. Is there some branch somewhere where this is happening?

[fggp]

There's a new branch: feature/newapi. In its include directory the file csound_new.h is the starting place for defining the new API. François Le mar. 25 juin 2024, 04:16, Shakeeb Alireza ***@***.***> a écrit :

…

@fggp <https://github.com/fggp> It's not clear to me where changes in the api are occurring. Is there some branch somewhere where this is happening? — Reply to this email directly, view it on GitHub <#25 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABNOAKKR72ICPK5XR23RBPLZJDHGXAVCNFSM6AAAAABJXXYIEOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOBXHAZDAMZTGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[shakfu]

@fggp Thanks for the info about the new api. Will check it out. 👍

[shakfu]

@fggp The new api is a lot shorter than the old one:

#ifndef CSOUND_H
#define CSOUND_H

//Instantiation
PUBLIC int 	csoundInitialize (int flags)
PUBLIC CSOUND*  csoundCreate (void *hostData) 
PUBLIC void 	csoundDestroy (CSOUND *) 

// Performance
// One compilation function that does everything
// For example, we could potentially use "argc" as a code
// > 0 - argc/argv as main() command-line parameters
// 0 - start csound as a daemon with no code, ignore argv
// -1 - argv[0] is an orchestra string (csoundCompileStr)
// -2 - argv[0] is a complete csound   (csoundCompileCsdText)
// First call to csoundCompile() also starts Csound so we remove csoundStart()
// async defines async operation
// NB - interface with double pointers may not be ideal for non C/C++
PUBLIC int  csoundCompile (CSOUND *, int argc, const char **argv, int async);
PUBLIC int  csoundPerformKsmps (CSOUND *);

// Realtime Audio
// this is now effectively -n since there is no access to buffer, only to spin.
// we need a better name I think
PUBLIC void csoundSetHostImplementedAudioIO(CSOUND *, int state);
// access to spin/spout (no need to give individual sample access)
PUBLIC MYFLT *csoundGetSpin (CSOUND *);
PUBLIC MYFLT *csoundGetSpout (CSOUND *csound);

// no audio IO callbacks - that's only available in the module API now.
// MIDI callbacks can be set with a single new function
typedef struct {
  int(*open)(CSOUND *, void **userData, const char *devName, int mode); // MIDI open callback (mode=CS_MIDI_IN, CS_MIDI_OUT)
  int(*read)(CSOUND *, void *userData, unsigned char *buf, int nBytes);
  int(*write)(CSOUND *, void *userData, unsigned char *buf, int nBytes);
  int(*close)(CSOUND *, void *userData);
} CS_MIDI_CALLBACKS;
PUBLIC void csoundSetMIDICallbacks(CSOUND *, CS_MIDI_CALLBACKS *);
 
// Messages
PUBLIC void 	   csoundCreateMessageBuffer (CSOUND *csound, int toStdOut)
PUBLIC const char *csoundGetFirstMessage (CSOUND *csound)
PUBLIC void 	   csoundPopFirstMessage (CSOUND *csound)
PUBLIC int 	       csoundGetMessageCnt (CSOUND *csound)
void PUBLIC 	   csoundDestroyMessageBuffer (CSOUND *csound)

// Channels
PUBLIC int csoundListChannels(CSOUND *, controlChannelInfo_t **lst);
PUBLIC void csoundDeleteChannelList(CSOUND *, controlChannelInfo_t *lst);
// basic types
PUBLIC MYFLT csoundGetControlChannel(CSOUND *csound, const char *name, int *err);
PUBLIC void csoundSetControlChannel(CSOUND *csound, const char *name, MYFLT val);
PUBLIC void csoundGetAudioChannel(CSOUND *csound, const char *name, MYFLT *samples);
PUBLIC void csoundSetAudioChannel(CSOUND *csound, const char *name, MYFLT *samples);
PUBLIC void csoundGetStringChannel(CSOUND *csound, const char *name, char *string);
PUBLIC  void csoundSetStringChannel(CSOUND *csound, const char *name, char *string);
// generic data access (e.g. for other types, structs, etc)
PUBLIC int csoundGetChannelPtr(CSOUND *, MYFLT **p, const char *name, int type);
PUBLIC int csoundGetChannelDatasize(CSOUND *csound, const char *name);

// events:
// async selects asynchronous operation
// new function - replaces csoundInputMessage
PUBLIC void  csoundEventString (CSOUND *, const char *message, int async);
// new function - replaces csoundScoreEvent
// type 0 - instrument instance     CS_INSTR_EVENT
// type 1 - functiob table instance CS_TABLE_EVENT
PUBLIC void  csoundEvent (CSOUND *, int type, MYFLT *params, int async);

//Tables
PUBLIC int 	csoundTableLength (CSOUND *, int table);
PUBLIC void csoundTableCopyIn (CSOUND *csound, int table, MYFLT *src);
PUBLIC void csoundTableCopyOut (CSOUND *csound, int table, MYFLT *dest);

//Opcodes - adapted for new internals
PUBLIC int 	csoundAppendOpcode (CSOUND *, const char *opname, int dsblksiz, int flags,
                                    const char *outypes, const char *intypes, int(*init)(CSOUND *, void *),
                                    int(*perf)(CSOUND *, void *), int(*deinit)(CSOUND *, void *));

//Threading and concurrency

#endif  /* CSOUND_H */

[fggp]

It's just a draft starting point. We're working on it. In the attachment you can read some proposals. Any suggestions are welcome. François Le ven. 28 juin 2024 à 16:28, Shakeeb Alireza ***@***.***> a écrit :

…

@fggp <https://github.com/fggp> The new api is a lot shorter than the old one: #ifndef CSOUND_H#define CSOUND_H //InstantiationPUBLIC int csoundInitialize (int flags)PUBLIC CSOUND* csoundCreate (void *hostData) PUBLIC void csoundDestroy (CSOUND *) // Performance// One compilation function that does everything// For example, we could potentially use "argc" as a code// > 0 - argc/argv as main() command-line parameters// 0 - start csound as a daemon with no code, ignore argv// -1 - argv[0] is an orchestra string (csoundCompileStr)// -2 - argv[0] is a complete csound (csoundCompileCsdText)// First call to csoundCompile() also starts Csound so we remove csoundStart()// async defines async operation// NB - interface with double pointers may not be ideal for non C/C++PUBLIC int csoundCompile (CSOUND *, int argc, const char **argv, int async);PUBLIC int csoundPerformKsmps (CSOUND *); // Realtime Audio// this is now effectively -n since there is no access to buffer, only to spin.// we need a better name I thinkPUBLIC void csoundSetHostImplementedAudioIO(CSOUND *, int state);// access to spin/spout (no need to give individual sample access)PUBLIC MYFLT *csoundGetSpin (CSOUND *);PUBLIC MYFLT *csoundGetSpout (CSOUND *csound); // no audio IO callbacks - that's only available in the module API now.// MIDI callbacks can be set with a single new functiontypedef struct { int(*open)(CSOUND *, void **userData, const char *devName, int mode); // MIDI open callback (mode=CS_MIDI_IN, CS_MIDI_OUT) int(*read)(CSOUND *, void *userData, unsigned char *buf, int nBytes); int(*write)(CSOUND *, void *userData, unsigned char *buf, int nBytes); int(*close)(CSOUND *, void *userData); } CS_MIDI_CALLBACKS;PUBLIC void csoundSetMIDICallbacks(CSOUND *, CS_MIDI_CALLBACKS *); // MessagesPUBLIC void csoundCreateMessageBuffer (CSOUND *csound, int toStdOut)PUBLIC const char *csoundGetFirstMessage (CSOUND *csound)PUBLIC void csoundPopFirstMessage (CSOUND *csound)PUBLIC int csoundGetMessageCnt (CSOUND *csound)void PUBLIC csoundDestroyMessageBuffer (CSOUND *csound) // ChannelsPUBLIC int csoundListChannels(CSOUND *, controlChannelInfo_t **lst);PUBLIC void csoundDeleteChannelList(CSOUND *, controlChannelInfo_t *lst);// basic typesPUBLIC MYFLT csoundGetControlChannel(CSOUND *csound, const char *name, int *err);PUBLIC void csoundSetControlChannel(CSOUND *csound, const char *name, MYFLT val);PUBLIC void csoundGetAudioChannel(CSOUND *csound, const char *name, MYFLT *samples);PUBLIC void csoundSetAudioChannel(CSOUND *csound, const char *name, MYFLT *samples);PUBLIC void csoundGetStringChannel(CSOUND *csound, const char *name, char *string);PUBLIC void csoundSetStringChannel(CSOUND *csound, const char *name, char *string);// generic data access (e.g. for other types, structs, etc)PUBLIC int csoundGetChannelPtr(CSOUND *, MYFLT **p, const char *name, int type);PUBLIC int csoundGetChannelDatasize(CSOUND *csound, const char *name); // events:// async selects asynchronous operation// new function - replaces csoundInputMessagePUBLIC void csoundEventString (CSOUND *, const char *message, int async);// new function - replaces csoundScoreEvent// type 0 - instrument instance CS_INSTR_EVENT// type 1 - functiob table instance CS_TABLE_EVENTPUBLIC void csoundEvent (CSOUND *, int type, MYFLT *params, int async); //TablesPUBLIC int csoundTableLength (CSOUND *, int table);PUBLIC void csoundTableCopyIn (CSOUND *csound, int table, MYFLT *src);PUBLIC void csoundTableCopyOut (CSOUND *csound, int table, MYFLT *dest); //Opcodes - adapted for new internalsPUBLIC int csoundAppendOpcode (CSOUND *, const char *opname, int dsblksiz, int flags, const char *outypes, const char *intypes, int(*init)(CSOUND *, void *), int(*perf)(CSOUND *, void *), int(*deinit)(CSOUND *, void *)); //Threading and concurrency #endif /* CSOUND_H */ — Reply to this email directly, view it on GitHub <#25 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABNOAKIX6EZW64PEVH2IZS3ZJVXHBAVCNFSM6AAAAABJXXYIEOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOJXGA3TCNBRGI> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[shakfu]

@fggp wrote

In the attachment you can read some proposals.

I'm sorry but I did not see any attachment.

[fggp]

Maybe it's been automatically removed by the list reception? Anyway the conclusion summaries the main ideas: "In summary one should develop a new host API. This API should be simpler with less granularity when possible and mainly based on C functions using an opaque pointer as first argument to mimic an Object Oriented paradigm. Direct memory access through pointers and length information should be used for buffers, tables, arrays, fsigs and other similar types, with care about memory management. State information struct should be accessed through getter and setter functions. No variadic functions and try to avoid callbacks when possible. Thread and concurrency mechanisms should be treated apart. Complex displays and graphics should be left to third party programs, the host API transmitting raw data." François Le dim. 30 juin 2024 à 05:54, Shakeeb Alireza ***@***.***> a écrit :

…

@fggp <https://github.com/fggp> wrote In the attachment you can read some proposals. I'm sorry but I did not see any attachment. — Reply to this email directly, view it on GitHub <#25 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ABNOAKOBGAEJAXNKVCDKFYLZJ56OLAVCNFSM6AAAAABJXXYIEOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCOJYGQZDGNZUGQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[shakfu]

Thanks for the summary. I completely agree with its contents.

This is good news since it looks like the csound7 api is really mostly a subset of the csound6 api.

I might as well continue my earlier work then.