The only way to get new values into a running Synth is using a Control input. Controls are automatically created from arguments of the SynthDef function.

The code you posted looks of the value of d.at("name") once when building the SynthDef. After that, it's hard coded and can't be changed (without rebuilding the SynthDef).

I'd suggest, keep the dictionary on the client side and send the new value as needed using .set --


hjh

Very tricky one. Pproto creates a separate environment that exists only within the pattern while it's executing. The function that you give at the start of Pproto executes within this separate environment. But, your wavetable functions exist in the main environment. Environments don't share values - there's no kind of nested scope like you have with functions. So, when {arg a; ~sin2.(a) } executes, ~sin2 is nil because the function is running inside an environment that knows nothing about ~sin2.

A workaround is to use topEnvironment[\sin2] in place of ~sin2. Or, put inEnvir in front of the function:

Pdef(\toto, ~mkbuf.(inEnvir {arg a; ~sin2.(a); a.postln; }, Pbind(\instrument, \wavetable)));

This connects the function to the outer environment.


Note also that using Buffer.loadCollection in Pproto means that the buffer will not get freed automatically at the end of the pattern. Pproto depends on the event types defined in EventTypesWithCleanup -- if resources are created by other means, Pproto doesn't know what to do with them afterward. The \table event type should suit this purpose.


I recently added a cleanup func to Pfset, which might be more intuitive.




The last one isn't right because _ is not a substitute for a.

{arg a; ~sinx.(a, 3) } is the same as ~sinx.(_, 3) -- but if you call .value on {arg a; ~sinx.(_,3) }, the result will be a Function that is supposed to call ~sinx - but this function will not actually be called.

James

The easiest way to find the line numbers is to use a proper text editor (not the one built into sc.app). A lot of programmers swear by text mate, but there are a bunch of freeware choices. Many editors, especially programming-oriented ones, have the option of showing line numbers in the left margin.

But, looking at the tab file reader class definition, I don't see an easy way to tell it to stop after a given number of lines. I don't have a quick answer to that problem (except that I probably would write my own reader and parser using the outline in my earlier post, especially if this is something I expect to do a lot).



Sure, I'll send to the UK. What was the error message? I can't help you much with the error if I don't know what the error is. (Unfortunately this is all too common in the computer world -- people think it's enough to say "there was an error." So then it takes longer to solve the problem because it's necessary to do another back-and-forth to find out what the problem really was. Sorry to go on about this -- it is a major peeve of mine.)

I can tell you that I've had horrible problems with spam from that page, so I block actually rather a lot of phrases. I also had to put in a timer because a lot of spammers submit the input form instantaneously after visiting the page. The side effect is that if you write the message to me in another editor, then go to my page, paste it in quickly and hit "send," as far as the website is concerned, you look like a spammer. Try waiting at least 15 seconds before sending.

Or, if all else fails, send me a PM here.
James

Nah, don't waste your time with C. (C is still useful for a lot of things, but most of the programming world has moved on to C++ or Java. C is not object-oriented, so it won't teach you much about SC.)

I can't think of a "general programming concepts for SC users" guide... maybe look at the Ruby language. It is somewhat similar to SC in some ways.


Oh, this is interesting. Yeah, you sometimes have to dig into the raw data to see what's really there before writing the final code. Sometimes I do this to look at the raw data:


// prints:
0D 54 69 74 6C 65 20 6F    .Title o
66 20 74 68 65 20 45 78    f the Ex
70 6F 72 74 20 42 72 61    port Bra
69 6E 20 44 61 74 61 3A    in Data:
09 09 09 45 78 70 6F 72    ...Expor
74 20 42 72 61 69 6E 20    t Brain
44 61 74 61 0D 0D 4D 61    Data..Ma
78 20 62 75 66 66 65 72    x buffer
20 73 69 7A 65 3A 09 09     size:..
09 20 30 31 3A 30 30 3A    . 01:00:
30 30 2E 30 30 30 0D 53    00.000.S
74 61 72 74 20 54 69 6D    tart Tim
65 20 69 6E 20 62 75 66    e in buf
66 65 72 3A 09 09 09 20    fer:...
30 30 3A 30 30 3A 30 30    00:00:00
2E 30 30 30 0D 45 6E 64    .000.End
20 54 69 6D 65 20 69 6E     Time in
20 62 75 66 66 65 72 3A     buffer:
09 09 09 20 30 30 3A 30    ... 00:0
32 3A 34 37 2E 35 39 31    2:47.591
0D 0D 49 42 56 41 34 20    ..IBVA4
61 6E 64 20 63 75 72 72    and curr
65 6E 74 20 66 69 6C 65    ent file
20 69 6E 66 6F 72 6D 61     informa
74 69 6F 6E 3A 09 09 09    tion:...

The first thing I see from this is that 0D represents a line break in this file, and getLine doesn't recognize it as a line break. (How do I know this is the line break? By comparing the raw output to the appearance in a text editor.)

The bad characters show up as NUL in notepad++ on my windows machine, meaning ascii code 0. That is a fairly serious data problem that is harder to correct in SC. Since SC's underbelly is C, the C-style rule for strings apply where a 0 byte means the end of the string. Probably the string in memory has all the characters, but they are ignored past 0 because of the C rule.

SC doesn't have a built-in method to strip 0 bytes, so you'll have to write your own getLine function.


Sorry this is such a pain... working with files you didn't create yourself is sometimes hard.

James

A text file is really just a stream of bytes. Some of those bytes are interpreted to mean the end of a line (but it's not required to read the file that way).


No -- file = File(path, mode) doesn't read anything. It just opens a file handle. You can read data from the file handle - or just close the handle without reading anything (which is a way to test for a file's existence on disk, e.g. -- see File.exists in the SC lib).


Try it for yourself -- running some exploratory code will teach you more about reading files.

f = File(path, "r");
f.getLine;   // what do you see in the post window?

f.getLine.dump;  // proves the result is a String, and shows some of the contents

In the source file, there is a line "Left CH1" that occurs at the beginning of the first data section. When that line is read, line != "Left CH1" becomes false and the while loop stops. The file's position should be after the newline character terminating "Left CH1" (i.e., the next character to be read will be the first character on the line following "Left CH1").


Yes. Double-quote syntax for string literals is explained in the Literals help file.

James

Reading a file like this takes a few basic capabilities that most computer languages have.

- reading strings from a file
- searching for characters or substrings in a string
- extracting substrings
- converting strings to numbers

Supercollider can do all of these things, so yes, of course it's possible.

What you need to do is think step-by-step about the parts of the file and how each should be processed.

First is a header, which you just want to ignore up until the line that says "Left CH1." That's pretty easy -- read a line and check to see if it == "Left CH1" -- if so, stop and go onto the next section. The standard way to do this is with a while loop. (Checking for notNil is also important because you want the loop to stop if the file runs out of lines.)


// the above line = ...; while ...; could be written more concisely as:
while { (line = file.getLine).notNil and: { line != "Left CH1" } };

Then follow the data. What steps do we need for each line?

- there are some bad characters; strip them out
- split the string wherever there is white space
- remove any empty substrings in the partitioned string (just in case)
- throw out the first three columns
- convert the rest to floating-point numbers

Oh, and stop when a line doesn't contain any numbers.


Then skip a few lines until you get to "Right CH2" and do the same thing to read the next table.

I'm not sure of the format you need for the final matrix, but array manipulations can certainly get you there.

Solving problems like this is a matter of thinking it through, one step at a time. Looking at the file as a whole, it seems difficult indeed but there is a very clear action to take for each separate line. So look at one line, and break down the process into simple steps.

Also it's a good idea to use protect when processing file data. If there is an error in the middle of file processing, you could leave a file handle open and then have no way to close it, not so good.

General template:


With this, even if an error occurs inside the protect block, the file is always closed.

James

An enterprising SC user has put up a global google map to locate other users.

maps.google.nl/maps...

Kind of fun to see where people are... feel free to add yourself if you like. All it takes is a google account, thus entailing the surrender of your online privacy in perpetuity. (Bit too late for me; I gave that up years ago.)

James

Browsing around, wasting time this fine Christmas Eve morning, I saw this and noticed something odd.



(Original img location: http://www.flickr.com/photos/danstowell/2898577451/in/pool-supercollider/)

What is wrong with this picture?
James

(Metadata is post-3.2 I believe... if you're on 3.2, leave it out and play the synth manually.)

James

Actually the userview/pen thing has some performance problems using Cocoa (OSX) gui objects. Here's another version using an empty statictext, setting the background color. It runs a lot faster in OSX.

James

I'd been meaning to do this for awhile... when I was a kid, I was obsessed with a small program on the Apple II that had been printed in the magazine of the Puget Sound apple users group (I am really dating myself here). It did a two-dimensional bubble sort of colored blocks, ending up with these curved and tapered bands of color.

So I hacked it up in SuperCollider... basically just like I remember it

With 3600 blocks it will take quite some time to complete, but it's fascinating to watch as the random matrix slowly organizes itself from the outside in. A smaller matrix runs faster of course, but the bands are chunkier and not as pretty. If you get bored with it, just close the window and the routine will stop.

James

Typically I like to keep the buffer with other data about it in one object. Identity dictionaries are really good for this because there is no preset structure and you can add stuff to it anytime -- and its subclass Event is even better because it has a really simple syntax.

b = (buf: Buffer.alloc(...));  // (name: value) makes an event

Then in the OSCresponderNode:


Now when you play back, you can use bufnum: b.buf and also have a duration argument to the synth, populated by b.dur (in seconds) or b.dur * b.buf.sampleRate (in frames).


As long as mouse button.kr is running, it's going to respond to every click. One solution might be to designate a block of screen coordinates to start and stop recording, and check using MouseX and MouseY (again, adapt this to your synth):


Another way would be to have some sort of GUI thing to click on, which would trigger an action function in the client. The action function could set an argument in the record synth to replace MouseButton. Many GUI objects respond to mouse down as well as mouse up.

James

If you're going to set other things in the synth, then you need to save the result of r.next in a variable, e.g.

a = r.next.set(...);

James

How about:

r.next.set(\synthArgName, maxrate12);

(Since r.next returns the new Synth, you can use it as the receiver of .set. Then, the comment "can yield anything" is no longer valid -- the synth routine will have to yield the synth object.)

It isn't clear to me what you want to do with maxrate12, actually. If you were trying to set a value on a Synth, the above would work. If not, then I don't think you would need .set in there.

James

The Routine making the Synths needs to .yield something after playing the synth. Right now I think the loop is ending, causing the Routine to exit - at that point you can't do any more with it. But if that Routine loops infinitely and yields after each synth, you can call .next on the synth-routine to get the next synth.

This is how I would do it. For fun, I'll convert your pitch stream to a pattern network 


Then your osc responder can call .next on the synth routine.

James