Nyquist Plug-ins Reference

The Audacity Nyquist Interface is implemented in the following files in the Audacity source code:


 * audacity/src/effects/nyquist/Nyquist.cpp


 * audacity/lib-src/libnyquist/nyx.c

Overview
Nyquist plug-ins are simple text files ending with the filename extension ".ny". When Audacity starts, it looks in a plug-ins directory for files and automatically adds the effects it finds there to the Audacity menus. The plug-ins can be written in either LISP or SAL syntax. When the user selects the plug-in from the menu, Audacity causes the Nyquist language runtime system (built into Audacity) to run the plug-in code to process or generate audio. You can also compute strings to display to the user or compute labels to be shown on a new label track.

Before running the plug-in, Audacity parses some of the plug-in text looking for specially formatted comments that describe "widgets" -- controls that the user can use to set parameters for the effect. For example, your plug-in can display a slider to control volume or filter frequency. Values from the widgets are passed through global variables to the plug-in.

For a full description of the plug-in language, read the Nyquist Reference Manual. There are, however, some details of how plug-in code is processed that is unique to plug-ins and therefore not covered in the Nyquist manual. The main thing to know is that the variable s is the selected sound. This is probably the main input to your effect. The second thing to know is how sound is returned from Nyquist to Audacity.

SAL syntax is inherently command oriented. A SAL plug-in should consist of a sequence of SAL commands including define commands to define variables and functions. There must be a function definition for main. After performing the commands in sequence, Audacity calls main (with no parameters). The value returned from main is the result of the plug-in (normally this should be a sound).

One exception to this is the effect. Since defining main just to evaluate and return a simple expression is so awkward, you can simply type return expression Normally, this is not a legal statement as a top-level command in SAL, but in, special processing embeds the return statement into a declaration of function main</tt>, which is then called. This trick also works for regular plug-ins, but defining main</tt> is the preferred style.

For LISP syntax plug-ins, Lisp expressions are read and evaluated one-at-a-time. The value returned by the last expression is the result of the plug-in. The same semantics applies to the effect. There is only one "Nyquist Prompt" for both SAL and LISP syntax. Audacity looks for the first non-space character that is not part of a comment. If it is an open paren "(", then the code is assumed to be LISP; otherwise, SAL is assumed.

The "Nyquist Prompt" effect is not equivalent to plug-ins because Audacity processes the plug-in header information from plug-in files, but the "Nyquist Prompt" effect ignores them (they are just comments).

Nyquist is a superset of the XLISP programming language and as such is limited to single-byte ASCII characters. The results of using characters outside the range of ASCII 32 to 126 in Nyquist plug-ins are unspecified. While the use of characters outside of this range may work satisfactorily for personal use on some operating systems, plug-ins for public distribution must currently be written using only standard characters (ASCII characters 32 to 126). This applies to plug-in code, comments and string data.

Nyquist Plug-in Header
The Nyquist Lisp interpreter as well as the Nyquist SAL compiler in contrast recognize everything after the first semicolon up to the end of the line as a comment and just simply ignores it. So in the Nyquist code, below the plug-in header, comments can also be started with several semicolons in a row.

;nyquist plug-in ;version version ;type type ;name "name" ;action "text" ;categories "text" ;info "text" ;control parameters ;codetype type ;debugflags flags

The "categories", "info", "control", "codetype", and "debugflags" lines are optional and can be omitted.

nyquist plug-in
Tells Audacity "this is a Nyquist plug-in": ;nyquist plug-in

version 1
Only the Slider widget is supported: ;version 1

version 2
The Text input widget was added in Audacity 1.2: ;version 2

version 3
The multiple-Choice widget was added in Audacity 1.3: ;version 3

type generate
Plug-in appears in the Audacity menu: ;type generate

type process
Plug-in appears in the Audacity menu: ;type process

type analyze
Plug-in appears in the Audacity menu: ;type analyze

name
Name of the plug-in as it will appear in the Audacity menu: ;name "name" Note that for plug-ins to be used in Chains, the colon character ":" cannot be used (as it is a special character in the Chain text file).

action
Text to be displayed while the plug-in is working: ;action "text"

categories
LV2 categories, see Categories below. ;categories "text"

info
Text to be displayed at the top border of the plug-in window: ;info "text"

A two-character sequence "\n" within "text" causes a line break. It is not possible to "quote" a line break in the plug-in code with a backslash '\' at the end of the line. In the plug-in header, the text in the "action" and "info" lines must be written in one single line.

control
Define a plug-in widget. There can be several "control" lines in the plug-in header. Add one for each widget to appear in the dialog box -- see Nyquist Plug-in Widgets below: ;control parameters

codetype lisp
The syntax of the plug-in is Lisp. This is the default if the "codetype" line is missing: ;codetype lisp

codetype sal
The syntax of the plug-in is SAL. The default is "lisp". ;codetype sal

debugflags
See Internal Debug Options below: ;debugflags flags

Nyquist Plug-in Widgets
Every ";control" line gets parsed by Audacity into several tokens, where each token is separated by one or several whitespaces:

Italic words in the table denote data types. Because tokens are separated by whitepace, strings containing whitespace must be written within quotation marks. Do not try to memorize this table, use it as a reference. The detailed syntax for each widget type is described in the following sections.

Slider Widget


Slider widgets are supported in all Audacity Nyquist plug-in versions.

;control variable-name "text-left" variable-type "text-right" initial-value minimum maximum


 * variable-name - a Lisp symbol.


 * text-left - text that will appear to the left of the slider.


 * variable-type - a "number" type, either int or real:


 * int - integer [FIXNUM, an XLISP number without a dot]


 * real - floating point [FLONUM, an XLISP number with a dot]


 * text-right - text that will appear to the right of the slider.


 * initial-value - variable value [and slider position] at the first start of the plug-in.


 * minimum - numerical variable value when the slider is moved to the left border.


 * maximum - numerical variable value when the slider is moved to the right border.

The variable value [the slider position] can be referenced by the variable name in the plug-in code.

In more recent versions of Audacity there appears a text input box beside the slider, where the user can type in a variable value via the keyboard. Note that it's up to the plug-in programmer to catch undesired values typed in by the user.

Text Input Widget


The text input widget is supported in plug-ins version 2 or above.

;control variable-name "text-left" string "text-right" "initial-string"


 * variable-name - a Lisp symbol.


 * text-left - text that will appear to the left of the text input field.


 * text-right - text that will appear to the right of the text input field.


 * initial-string - the string will appear inside the text field.

The text typed in by the user in the text field of the plug-in window can be referred as a string variable from within the plug-in code.

Examples how to use the text input widget can be found in the source code of the Apropos Plug-in.

Multiple-Choice Widget


The multiple choice input widget is supported in plug-ins version 3 or above.

;control variable-name "text-left" choice "string-1,string-2,..." initial-value


 * variable-name - a Lisp symbol.


 * text-left - text that will appear to the left of the multiple-choice list.


 * string-1,... - for every string an entry in a list to choose from will be produced.


 * initial-value - the number of the list entry that will be displayed as the default choice at the first start of the plug-in.

The list entries string-1, string-2, etc. are internally represented by integer numbers. The first, top-most list entry string-1 will be represented by the number 0. The list entry chosen by the user can be determined by the integer value of the variable from within the plug-in code.

Examples how to use the 'choice' widget can be found in the source code of the Apropos Plug-in.

Nyquist Variables
The variables given from Audacity to Nyquist are defined in the file "audacity/lib-src/libnyquist/nyx/nyx.c" in the Audacity source code.

The following variables are given from Audacity to Nyquist:


 * s - the Audacity sound [the selected part of the Audacity audio track]


 * len - the number of samples contained in s


 * *sound-srate* - the sample frequency of the Audacity track


 * *warp* - information that communicates start-time and duration to Nyquist functions. In Audacity, the start-time is always considered to be zero (regardless of the actual start time of the selection) and the duration indicates the duration of the selection.


 * *scratch* - a symbol whose value and property list are preserved from one effect invocation to the next. See "The *scratch* Symbol" below.

The length of the sound in seconds can be computed one of the following ways:

(/ len *sound-srate*) ; in LISP (get-duration 1) ; in LISP

len / *sound-srate* ; in SAL get-duration(1) </i>; in SAL</i>

The *scratch* Symbol
Since Audacity 1.3.9, there is a new global symbol *scratch*, which is not deleted in-between plug-in runs.

The point of *scratch* is to provide some way for information to survive from one invocation of a plug-in to the next. Because of the possibility that different effects could interact in undesirable ways by setting *scratch*, it is recommended to use only property lists of *scratch*. That way, you get a whole name space rather than a single variable name.

To pass data from plug-in "effectX-partA" to "effectX-partB":

1. Assign a property name based on the effect name, e.g.: 'EFFECTX [or in SAL, which does not support the single-quote notation of LISP, write QUOTE(EFFECTX). ]

2. "effectX-partA" should delete any old property value:

exec remprop(quote(*scratch*), quote(effectx)) ;; in SAL

(remprop '*scratch* 'effectx) ;; in LISP

3. "effectX-partA" should compute a new property value v and save it:

exec putprop(quote(*scratch*), v, quote(effectx)) ;; in SAL

(putprop '*scratch* v 'effectx) ;; in LISP

4. "effectX-partB" should access the property using:

set v = get(quote(*scratch*), quote(effectx)) ;; in SAL

(get '*scratch* 'effectx) ;; in LISP

5. When "effectX-partB" finishes, it should remove the property:

exec remprop(quote(*scratch*), quote(effectx)) ;; in SAL

(remprop '*scratch* 'effectx) ;; in LISP

But there may be cases where you do some analysis and want to use the analysis data multiple times. You might even have multiple analysis plug-ins operating on different inputs to collect data to feed into a plug-in with multiple inputs. In this case, which might be quite common, you should not call REMPROP, but this has the problem of leaving data on the *scratch* property list indefinitely.

In cases where *scratch* data is not deleted immediately after use, and where there is the potential to leave large amounts of memory there, there should be another effect, e.g. "effectX-partCleanup", that simply calls:

exec remprop(quote(*scratch*), quote(effectx)) ;; in SAL

(remprop '*scratch* 'effectx) ;; in LISP

allowing the user to explicitly free up any data stored on the 'EFFECTX property. It would be reasonable to omit the "effectX-partCleanup" effect if the data stored on the property list has a maximum size of, say, 10KB. The problem we want to avoid is properties with unbounded size getting left in the heap until Audacity is restarted.

Stereo Tracks
If a sound from an Audacity stereo track was given to Nyquist, the s variable contains an array of sounds. Because all Nyquist "snd-..." low-level functions only can process mono signals, to use such a function, the s array first must be split into single mono signals and afterwards be re-combined into an array before it is given back to Audacity.

In Sal, one could write: if arrayp(s) then return vector(snd-function(s[0]), snd-function(s[1])) else return snd-function(s)

Or in LISP, one could write:

(if (arrayp s)     (vector (snd-function (aref s 0))  ; left stereo channel (snd-function (aref s 1))) ; right stereo channel      (snd-function s))             ; mono signal


 * (arrayp s) - tests if 's' is an array


 * (vector ... ) - re-combines the two mono signals into a stereo signal. A "vector" is an one-dimensional array


 * (aref s 0) - the left stereo channel [the 0-th slot of the array]


 * (aref s 1 - the right stereo channel [the 1-st slot of the array]

Important: The Nyquist interface within Audacity can handle a maximum of two channels simultaneously [Audacity stereo tracks]. If in Audacity more than one audio track were selected, each of the selected tracks will be given sequentially, one after the other, with a maximum of two channels simultaneously [stereo] to Nyquist for processing. It is not possible with Nyquist in Audacity e.g. to copy audio signals from one Audacity track into another track [Audacity 1.3.9, November 2009].

multichan-expand
In the "nyquist.lsp" file in the Audacity "nyquist" sub-directory there is a function "multichan-expand" defined that simplifies the handling of multi-channel sounds [e.g. stereo tracks]:

(multichan-expand function &rest arguments)

So the "arrayp" constuct from above can also be written: return multichan-expand(quote(snd-function), s) ;; in SAL

(multichan-expand #'snd-function s) ;; in LISP

This looks a bit more cryptic and reads less intelligibly [to some], but it can help to clean up the code in long-winded audio processing functions.

Return Values
The result of the last computation within the plug-in code will be given back from Nyquist to Audacity. According to the data type of the Nyquist return value one of the following actions will be invoked in Audacity:

Sound: The sound will be re-inserted into the selected part of the Audacity track. If the returned sound is shorter or longer than the original sound, the selection will be reduced or augmented. If a mono sound is returned to a stereo track, so in both channels of the stereo track the same mono sound will be inserted. If a stereo sound is returned to a mono track, an error message will be displayed.

String: A dialog window will appear with the string being displayed as text.

Number: A dialog window will appear with the number being displayed as text.

List: If a specially formatted list is given back to Audacity, a label track will be created below the audio track(s).

For point labels the format is: ((number "string") (number "string") ... ) The list to create a label track must contain one or more lists, each of which must have:
 * number - (an integer or float) is the time in seconds from the beginning of the Audacity selection, where the label will appear.
 * "string" - a string to be displayed in the label's text field.

For region labels (Audacity 1.3) each label list must contain two int-or-float elements, one for the start and one for the end of the label region. ((number number "string") (number number "string") ... )

Math Functions
n-th power of x:

power(x, n) ;; in SAL (power x n) ;; in LISP n-th root of a number:

power(x, 1.0 / n) ;; in SAL (power x (/ 1.0 n)) ;; in LISP

n-th root of a sound s: return s-exp(s-log(s) * (1.0 / n)) ;; in SAL (s-exp (mult (s-log s) (/ 1.0 n))) ;; in LISP

n-th root with two sounds x and n: return s-exp(s-log(x) * s-recip(n))  ;; in SAL (s-exp (mult (s-log x) (s-recip n))) ;; in LISP

Nyquist Workbench
The Nyquist Workbench gives the ability to run arbitrary Nyquist code in Audacity from a graphical IDE.



The Nyquist Workbench is an Audacity module, previously available as standalone source code from http://audacity.homerow.net/index.php?dir=modules%2F.

Nyquist Workbench is included in the Audacity HEAD source code at \lib-src\mod-nyq-bench and has now moved from the Tools menu to the View menu.

Linux
The following steps are correct for Linux as of July 2013 building from the Audacity HEAD development code.

1. Type "./configure " in the Audacity source code directory:

./configure

2. If configure completes successfully, compile Audacity. In the Audacity source code directory type:

make

If you are having problems at either of the above points, try the Compiling Audacity Step by Step Guide.

3. If 'make' completes successfully, still in the Audacity source code directory, type:

sudo make install

which by default installs Audacity to "/usr/local/share/audacity".

4. Open and set the required AUDACITY_DIR variable at the top of the Makefile to the base of your Audacity source code directory, for example:

AUDACITY_DIR ?= /home/edgar/downloads/audacity

5. After saving the changes in the Makefile, compile the Nyquist Workbench. In the Nyquist Workbench directory type:

make

After "make" has finished you should see a message like:

NyqBench.so has been copied to /home/edgar/downloads/audacity/modules

6. On Linux, the Audacity "make install" does not copy the "modules" directory, so change to the base of your Audacity source code directory, then type:

sudo cp -r modules /usr/local/share/audacity

This copies the "modules" directory with its complete contents from the Audacity source code directory to "/usr/local/share/audacity/modules". If you installed Audacity other than to /usr/local/share/audacity, point the copy command to that directory. Alternatively, you could create an "~/.audacity-files/modules" directory:

mkdir -p ~/.audacity-files/modules

then move or copy "NyqBench.so" into that "modules" directory where Audacity will find it irrespective of the location of the Audacity installation directory.

Mac OS X
This information assumes you have already followed Building On Mac and is correct as of July 2013.

<ol><li> It is not currently possible to build mod-nyq-bench at the command-line using the Makefile. Therefore it is suggested that you build both Audacity and mod-nyq-bench in the Xcode IDE. First, build Audacity using the Xcode IDE. <li>Next, in the "Groups & Files" column in Xcode, expand "Targets" and select "mod-nyq-bench". Right-click over the selection and choose Build "mod-nyq-bench". Alternatively from the drop-down box top left, under the "Active Target" list, click on "mod-nyq-bench" (don't change the "Active Configuration"), then click the "Build" icon. <li> Assuming you built Release Static configuration, separate PPC and i386 builds of mod-nyq-bench.so will be built in their own folder inside. However the build process uses lipo to create a single universal binary mod-nyq-bench.so in. This universal binary is the module Audacity will load. </ol>

Loading/Enabling mod-nyq-bench
Each time you launch Audacity you will be asked if you want to load mod-nyq-bench. If you choose "yes" and the build of mod-nyq-bench matches with the version string of Audacity, you should then have an additional "Nyquist Workbench" entry at the bottom of the "View" Menu. Until r12363 (16 June 2013), when you launched Audacity after a first build of mod-nyq-bench, you had to enable the module in a "Modules" section of Preferences, then if so enabled Audacity would load mod-nyq-bench on next launch if it was matched with the Audacity version string.

Rebuilding Audacity and Nyquist Workbench
When you update your Audacity source code tree and rebuild Audacity on a subsequent day, you must also rebuild mod-nyq-bench.
 * It is recommended to run "make clean" or "make distclean" before configuring and building Audacity.
 * In Xcode on Mac, select "Audacity" under "Targets" and choose Clean "Audacity"  before rebuilding Audacity.
 * After building Audacity, change to the Nyquist Workbench directory, type "make clean" to remove NyqBench.so and NyqBench.o, then type "make" to rebuild NyqBench.so.
 * In Xcode on Mac, select "mod-nyq-bench" under "Targets" and choose Clean "mod-nyq-bench"  before rebuilding the module.

Nyquist Apropos Plug-in
Because not all functions, documented in the Nyquist manual, are implemented in Audacity and also not even all Nyquist functions are documented in the Nyquist manual [Lisp is a "programmable programming language", only functions which are considered as stable are documented], it is often helpful to have a tool to find out whether a particular Nyquist function is implemented or not.

Basically only Nyquist functions beginning with "snd-..." are implemented in C within the Nyquist interpreter [in the Nyquist manual they are called 'low-level' functions], while all other Nyquist functions are implemented in the Lisp files within the Audacity "nyquist" sub-directory and can be changed, improved and extended with no need to re-compile Audacity [as long as no C-implemented Nyquist "low-level" function is needed which is not included with Nyquist in Audacity].

All Nyquist/XLISP symbols [e.g. all variable and function names] are stored in the XLISP *obarray*. To find out from within Audacity, whether a particular function is implemented or not, you can e.g. first create and select a "dummy" audio track with any generator from the Audacity menu, then open  [doesn't work without a selected audio track] and copy the following line into the "Nyquist Prompt" text input field:

(print (fboundp 'snd-abs))

Important: now please click "Debug" instead of the "OK" button.

First a dialog appears telling you: "Nyquist did not return audio" [or similar]. In this dialog, click the "OK" button. Then another dialog "Nyquist output" appears with the result of the "print" function:


 * T - meaning "true", function is implemented


 * NIL - meaning "false", function is not implemented

I had typed this so often in the past that I have written an Audacity Nyquist "Apropos" Plug-in:

Download: Apropos-Plugin

Right-Click on the plug-in link, choose "save target as", and save the file "apropos.ny" file in the Audacity "Plug-Ins" sub-directory. After a re-start of Audacity you can find the "Apropos" Plug-in in the menu, the only menu that works without a "dummy" audio track.

The "Apropos" Plug-in offers a pattern search through the Nyquist/XLISP *obarray*. You can also choose between "Functions only", "Variables only" and "All symbols":



Note: with no search pattern [empty field], the plug-in matches all symbol names.

First appears a dialog that reminds me that I have forgotten to press the "Debug" button:



In the first dialog above, just klick the "OK" button. Afterwards, but only if you have clicked the "Debug" button in the plug-in window, appears the window with the results, sorted in alphabetical order:



Bugs
Details of current bugs that affect Nyquist in Audacity are listed on Bugzilla.