Gambit

This manual documents Gambit. It covers release v4.9.4.

1. The Gambit system

The Gambit programming system is a full implementation of the Scheme language which conforms to the R4RS, R5RS, R7RS and IEEE Scheme standards. It consists of two main programs: gsi, the Gambit Scheme interpreter, and gsc, the Gambit Scheme compiler.

The Gambit Scheme compiler translates Scheme code to another target language, currently C or JavaScript. The C target is the most mature and it offers portability and fast execution. The JavaScript target allows writing web apps in Scheme.

Most of the Gambit system, including the interpreter and compiler, is written in Scheme and compiled to portable C code using the compiler. The high portability of the generated C code allows the interpreter, compiler and user programs to be easily compiled and executed on any platform for which a C compiler is available. With appropriate declarations in the source code the compiled Scheme programs run roughly as fast as equivalent C programs.

For the most up to date information on Gambit and related resources please visit the Gambit web page at https://gambitscheme.org. Issues should be reported on the github source code repository https://github.com/gambit/gambit.

1.1 Accessing the system files

Files related to Gambit, such as executables, libraries and header files, are stored in multiple Gambit installation directories. Gambit may be installed on a system according to two different installation models.

In the first model there is a single directory where all the Gambit installation directories are stored. This central installation directory is typically /usr/local/Gambit under UNIX, /Library/Gambit under macOS and C:/Program Files/Gambit under Microsoft Windows. This may have been overridden when the system was built with the command ‘configure --prefix=/my/Gambit’. If the system was built with the command ‘configure --enable-multiple-versions’ then the central installation directory is prefix/version, where version is the system version string (e.g. v4.9.4 for Gambit v4.9.4). Moreover, prefix/current will be a symbolic link which points to the central installation directory. In this model, the Gambit installation directory named X is simply the subdirectory X of the central installation directory.

In the second model some or all of the Gambit installation directories are stored in installation specific directories. The location of these directories is assigned when the system is built using the command ‘configure --bindir=/my/bin --includedir=/my/include --libdir=/my/lib’.

The advantage of the first model is that it is easy to have multiple versions of Gambit coexist and to remove all the files of a given version. However, the second model may be necessary to conform to the package installation conventions of some operating systems.

Executable programs such as the interpreter gsi and compiler gsc can be found in the bin installation directory. Adding this directory to the PATH environment variable allows these programs to be started by simply entering their name. This is done automatically by the macOS and Microsoft Windows installers.

The runtime library is located in the lib installation directory. When the system’s runtime library is built as a shared-library (with the command ‘configure --enable-shared’) all programs built with Gambit, including the interpreter and compiler, need to find this library when they are executed and consequently this directory must be in the path searched by the system for shared-libraries. This path is normally specified through an environment variable which is LD_LIBRARY_PATH on most versions of UNIX, LIBPATH on AIX, SHLIB_PATH on HPUX, DYLD_LIBRARY_PATH on macOS, and PATH on Microsoft Windows. If the shell is sh, the setting of the path can be made for a single execution by prefixing the program name with the environment variable assignment, as in:

$ LD_LIBRARY_PATH=/usr/local/Gambit/lib gsi

A similar problem exists with the Gambit header file gambit.h, located in the include installation directory. This header file is needed for compiling Scheme programs with the Gambit compiler. When the C compiler is being called explicitly it may be necessary to use a -I<dir> command line option to indicate where to find header files and a -L<dir> command line option to indicate where to find libraries.

Access to both of these files can be simplified by creating a link to them in the appropriate system directories (special privileges may however be required):

$ ln -s /usr/local/Gambit/lib/libgambit.a /usr/lib # name may vary
$ ln -s /usr/local/Gambit/include/gambit.h /usr/include

Alternatively these files can be copied or linked in the directory where the C compiler is invoked (this requires no special privileges).

Another approach is to set some environment variables which are used to tell the C compiler where to find header files and libraries. For example, the following settings can be used for the gcc C compiler:

$ export LIBRARY_PATH=/usr/local/Gambit/lib
$ export CPATH=/usr/local/Gambit/include

Note that this may have been done by the installation process. In particular, the macOS and Microsoft Windows prebuilt installers set up the environment so that the gcc compiler finds these files automatically.

2. The Gambit Scheme interpreter

Synopsis:

gsi [-:runtimeoption,…] [-i] [-f] [-h] [-help] [-v]
    [[-] [-e expressions] [-install] [-uninstall] [-update]
     [search-directory-or-module-or-file]]…

The interpreter is executed in batch mode when the command line contains a module or file or a ‘-’, or ‘-e’ option. The interpreter is executed in module management mode when the command line contains the ‘-install’, ‘-uninstall’, or ‘-update’ option. Otherwise the interpreter is executed in interactive mode. The ‘-i’ option is ignored by the interpreter. The initialization file will be examined unless the ‘-f’ option is present (see section Customization). The ‘-h’ and ‘-help’ options print brief usage information on standard output and exit. The ‘-v’ option prints the system version string, system time stamp, operating system type, and configure script options on standard output and exits. Runtime options are explained in Runtime options.

2.1 Interactive mode

In interactive mode a read-eval-print loop (REPL) is started for the user to interact with the interpreter. At each iteration of this loop the interpreter displays a prompt, reads a command and executes it. The commands can be expressions to evaluate (the typical case) or special commands related to debugging, for example ‘,q’ to terminate the process (for a complete list of commands see Debugging or use the ‘,help’ command). Most commands produce some output, such as the value or error message resulting from an evaluation.

The input and output of the interaction is done on the interaction channel. The interaction channel can be specified through the runtime options but if none is specified the system uses a reasonable default that depends on the system’s configuration. Typically the program’s standard input and output are used as the interaction channel. When using the runtime option ‘-:debug=c’, the interaction channel is the user’s console, also known as the controlling terminal in the UNIX world.

When the REPL starts, the ports associated with ‘(current-input-port)’, ‘(current-output-port)’ and ‘(current-error-port)’ all refer to the interaction channel.

Expressions are evaluated in the global interaction environment. The interpreter adds to this environment any definition entered using the define and define-macro special forms. Once the evaluation of an expression is completed, the value or values resulting from the evaluation are output to the interaction channel by the pretty printer. The special void object is not output. This object is returned by most procedures and special forms which are defined as returning an unspecified value (e.g. write, set!, define).

Here is a sample interaction with gsi:

$ gsi
Gambit v4.9.4

> (define (fact n) (if (< n 2) 1 (* n (fact (- n 1)))))
> (map fact '(1 2 3 4 5 6))
(1 2 6 24 120 720)
> (values (fact 10) (fact 40))
3628800
815915283247897734345611269596115894272000000000
> ,q

What happens when errors occur is explained in Debugging.

2.2 Batch mode

In batch mode the command line arguments denote modules and files to execute, REPL interactions to start (‘-’ option), and expressions to be evaluated (‘-e’ option). Those options can be interspersed with the search directories, modules, and files on the command line and can occur multiple times.

In addition to these options the command line may contain 3 types of non-options: search directories, modules, and files.

Search directories

Search directories are locations in the file system that are searched to resolve references to modules. Any command line argument that ends with a path separator or a ‘.’ is treated as a search directory. By default the module search order is initially ~~lib (which contains builtin modules) followed by ~~userlib (which contains user installed modules and is typically the directory .gambit_userlib in the user’s home directory). Search directories on the command line are added to the front of the search order, and thus take precedence over the default module search order.

Modules

Modules are either unversioned or versioned (managed by the git version-control system). There are two flavors of versioned modules: hosted modules have a git repository on a network accessible repository manager site such as github.com and gitlab.com, and local modules have a git repository on the local file system. Module names have a syntax similar to the paths used to identify files. They consist of one or more non-empty parts separated by ‘/’. The last part may end with a suffix of the form @version. Only the first part and version may contain ‘.’, otherwise only the characters a-z, A-Z, 0-9, ‘-’, and ‘_’ are permitted. If there are at least 3 parts and the first part contains at least one ‘.’ and no ‘_’, then it refers to a hosted module (1st part = host, 2nd part = account, 3rd part = repository name). For example github.com/gambit/hello@1.0 is a hosted module reference. Otherwise it refers to a local versioned module or an unversioned module, for example foobar or A/B/C/D.

Files

Files are simple code containers located on the local file system. They are also identified by a path. If a path is a valid module or file, it is interpreted as a module. Note that a path with a last component containing an extension, such as ‘.scm’, and no @, is always interpreted as a file.

The interpreter processes the command line arguments from left to right. Search directories are added to the head of the module search order. Files are executed using the load procedure. Modules are requested using the ##demand-module special form (this form is explained in Modules, but essentially it causes that module to be searched in the module search order and executed once). The ‘-e’ option uses the eval procedure to evaluate expressions in the global interaction environment. After this processing the interpreter exits.

The ports associated with ‘(current-input-port)’, ‘(current-output-port)’ and ‘(current-error-port)’ initially refer respectively to the standard input (‘stdin’), standard output (‘stdout’) and the standard error (‘stderr’) of the interpreter. This is true even in REPLs started with the ‘-’ option. The usual interaction channel is still used to read expressions and commands and to display results. This makes it possible to use REPLs to debug programs which read the standard input and write to the standard output, even when these have been redirected.

Here is a sample use of the interpreter in batch mode, under UNIX:

$ cat h.scm
(display "hello") (newline)
$ cat w.six
display("world"); newline();
$ gsi h.scm - w.six -e "(pretty-print 1)(pretty-print 2)"
hello
> (define (display x) (write (reverse (string->list x))))
> ,c
(#\d #\l #\r #\o #\w)
1
2
$ gsi . h w   # add . to search order to load modules h and w
hello
world

2.3 Module management mode

Package management operations are executed using the command line options ‘-install’, ‘-uninstall’, and ‘-update’ which respectively install, uninstall and update packages. Package installation is explained in detail in Modules, but here are a few examples:

$ gsi -install github.com/gambit/hello
installing github.com/gambit/hello to /Users/feeley/.gambit_userlib/
$ gsi github.com/gambit/hello@1.0
hello world!
$ gsi -uninstall github.com/gambit/hello
uninstalling github.com/gambit/hello from /Users/feeley/.gambit_userlib/

2.4 Customization

There are two ways to customize the interpreter. When the interpreter starts off it tries to execute a ‘(load "~~lib/gambext")’ (for an explanation of how file names are interpreted see Host environment). An error is not signaled when the file does not exist. Interpreter extensions and patches that are meant to apply to all users and all modes should go in that file.

Extensions which are meant to apply to a single user or to a specific working directory are best placed in the initialization file, which is a file containing Scheme code. In all modes, the interpreter first tries to locate the initialization file by searching the following locations: ‘.gambini’ and ‘~/.gambini’ (with no extension, a ‘.sld’ extension, a ‘.scm’ extension, and a ‘.six’ extension in that order). The first file that is found is examined as though the expression (include initialization-file) had been entered at the read-eval-print loop where initialization-file is the file that was found. Note that by using an include the macros defined in the initialization file will be visible from the read-eval-print loop (this would not have been the case if load had been used). The initialization file is not searched for or examined when the ‘-f’ option is specified.

2.5 Process exit status

The status is zero when the interpreter exits normally and is nonzero when the interpreter exits due to an error. Here is the meaning of the exit statuses:

0

The execution of the primordial thread (i.e. the main thread) did not encounter any error. It is however possible that other threads terminated abnormally (by default threads other than the primordial thread terminate silently when they raise an exception that is not handled).

64

The runtime options or the environment variable ‘GAMBOPT’ contained a syntax error or were invalid.

70

This normally indicates that an exception was raised in the primordial thread and the exception was not handled.

71

There was a problem initializing the runtime system, for example insufficient memory to allocate critical tables.

For example, if the shell is sh:

$ gsi -e "(pretty-print (expt 2 100))"
1267650600228229401496703205376
$ echo $?
0
$ gsi -e "(pretty-print (expo 2 100))"
*** ERROR IN (string)@1.16 -- Unbound variable: expo
$ echo $?
70
$ gsi -:debug=0 -e "(pretty-print (expo 2 100))"
$ echo $?
70
$ gsi -:debug=0,unknown # try to use an unknown runtime option
$ echo $?
64
$ gsi -:debug=0 nonexistent.scm # try to load a file that does not exist
$ echo $?
70
$ gsi nonexistent.scm
*** ERROR IN ##load-module-or-file -- No such file or directory
(load "nonexistent.scm")
$ echo $?
70

Note the use of the runtime option ‘-:debug=0’ that prevents error messages from being output.

2.6 Scheme scripts

The load procedure treats specially files that begin with the two characters ‘#!’ and ‘@;’. Such files are called script files and the first line is called the script line. In addition to indicating that the file is a script, the script line provides information about the source code language to be used by the load procedure. After the two characters ‘#!’ and ‘@;’ the system will search for the first substring matching one of the following language specifying tokens:

scheme-r4rs

R4RS language with prefix syntax, case-insensitivity, keyword syntax not supported

scheme-r5rs

R5RS language with prefix syntax, case-insensitivity, keyword syntax not supported

scheme-ieee-1178-1990

IEEE 1178-1990 language with prefix syntax, case-insensitivity, keyword syntax not supported

scheme-srfi-0

R5RS language with prefix syntax and SRFI 0 support (i.e. cond-expand special form), case-insensitivity, keyword syntax not supported

gsi-script

Full Gambit Scheme language with prefix syntax, case-sensitivity, keyword syntax supported

gsc-script

Full Gambit Scheme language with prefix syntax, case-sensitivity, keyword syntax supported

six-script

Full Gambit Scheme language with infix syntax, case-sensitivity, keyword syntax supported

If a language specifying token is not found, load will use the same language as a nonscript file (i.e. it uses the file extension and runtime system options to determine the language).

After processing the script line, load will parse the rest of the file (using the syntax of the language indicated) and then execute it. When the file is being loaded because it is an argument on the interpreter’s command line, the interpreter will:

• Setup the command-line procedure so that it returns a list containing the expanded file name of the script file and the arguments following the script file on the command line. This is done before the script is executed. The expanded file name of the script file can be used to determine the directory that contains the script (i.e. (path-directory (car (command-line)))).
• After the script is loaded the procedure main is called with the command line arguments. The way this is done depends on the language specifying token. For scheme-r4rs, scheme-r5rs, scheme-ieee-1178-1990, and scheme-srfi-0, the main procedure is called with the equivalent of (main (cdr (command-line))) and main is expected to return a process exit status code in the range 0 to 255. This conforms to the “Running Scheme Scripts on Unix SRFI” (SRFI 22). For gsi-script and six-script the main procedure is called with the equivalent of (apply main (cdr (command-line))) and the process exit status code is 0 (main’s result is ignored). The Gambit system has a predefined main procedure which accepts any number of arguments and returns 0, so it is perfectly valid for a script to not define main and to do all its processing with top-level expressions (examples are given in the next section).
• When main returns, the interpreter exits. The command line arguments after a script file are consequently not processed (however they do appear in the list returned by the command-line procedure, after the script file’s expanded file name, so it is up to the script to process them).

2.6.1 Scripts under UNIX and macOS

Under UNIX and macOS, the Gambit installation process creates the executable ‘gsi’ and also the executables ‘six’, ‘gsi-script’, ‘six-script’, ‘scheme-r5rs’, ‘scheme-srfi-0’, etc as links to ‘gsi’. A Scheme script need only start with the name of the desired Scheme language variant prefixed with ‘#!’ and the directory where the Gambit executables are stored. This script should be made executable by setting the execute permission bits (with a ‘chmod +x script’). Here is an example of a script which lists on standard output the files in the current directory:

#!/usr/local/Gambit/bin/gsi-script
(for-each pretty-print (directory-files))

Here is another UNIX script, using the Scheme infix syntax extension, which takes a single integer argument and prints on standard output the numbers from 1 to that integer:

#!/usr/local/Gambit/bin/six-script

function main(n_str)
{
  scmobj n = \string->number(n_str);
  for (scmobj i=1; i<=n; i++)
    \pretty-print(i);
}

For maximal portability it is a good idea to start scripts indirectly through the ‘/usr/bin/env’ program, so that the executable of the interpreter will be searched in the user’s ‘PATH’. This is what SRFI 22 recommends. For example here is a script that mimics the UNIX ‘cat’ utility for text files:

#!/usr/bin/env gsi-script

(define (display-file filename)
  (display (call-with-input-file filename
             (lambda (port)
               (read-line port #f)))))

(for-each display-file (cdr (command-line)))

2.6.2 Scripts under Microsoft Windows

Under Microsoft Windows, the Gambit installation process creates the executable ‘gsi.exe’ and ‘six.exe’ and also the batch files ‘gsi-script.bat’, ‘six-script.bat’, ‘scheme-r5rs.bat’, ‘scheme-srfi-0.bat’, etc which simply invoke ‘gsi.exe’ with the same command line arguments. A Scheme script need only start with the name of the desired Scheme language variant prefixed with ‘@;’. A UNIX script can be converted to a Microsoft Windows script simply by changing the script line and storing the script in a file whose name has a ‘.bat’ or ‘.cmd’ extension:

@;gsi-script %~f0 %*
(display "files:\n")
(pretty-print (directory-files))

Note that Microsoft Windows always searches executables in the user’s ‘PATH’, so there is no need for an indirection such as the UNIX ‘/usr/bin/env’. However the script line must end with ‘%~f0 %*’ to pass the expanded filename of the script and command line arguments to the interpreter.

2.6.3 Compiling scripts

A script file can be compiled using the Gambit Scheme compiler (see section The Gambit Scheme compiler) into a standalone executable. The script line will provide information to the compiler on which language to use. The script line also provides information on which runtime options to use when executing the compiled script. This is useful to set the default runtime options of an executable program.

The compiled script will be executed similarly to an interpreted script (i.e. the list of command line arguments returned by the command-line procedure and the invocation of the main procedure).

For example:

$ cat square.scm
#!/usr/local/Gambit/bin/gsi-script -:debug=0
(define (main arg)
  (pretty-print (expt (string->number arg) 2)))
$ gsi square 30        # gsi will load square.scm
900
$ gsc -exe square      # compile the script to a standalone program
$ ./square 30
900
$ ./square 1 2 3       # too many arguments to main
$ echo $?
70
$ ./square -:debug=1 1 2 3  # ask for error message
*** ERROR IN ##start-main -- Wrong number of arguments passed to procedure
(main "1" "2" "3")

3. The Gambit Scheme compiler

Synopsis:

gsc [-:runtimeoption,…] [-i] [-f] [-h] [-help] [-v]
    [-target target]
    [-prelude expressions] [-postlude expressions]
    [-dynamic] [-exe] [-obj]
    [-nb-gvm-regs n] [-nb-arg-regs n] [-compactness level]
    [-cc compiler] [-cc-options options]
    [-ld-options-prelude options] [-ld-options options]
    [-pkg-config pkg-config-args] [-pkg-config-path pkg-config-path]
    [-warnings] [-verbose] [-report] [-expansion] [-gvm] [-cfg] [-dg]
    [-debug] [-debug-location] [-debug-source]
    [-debug-environments] [-track-scheme]
    [-o output] [-c] [-keep-temp] [-link] [-flat] [-l base]
    [-module-ref module-ref] [-linker-name linker-name]
    [[-] [-e expressions] [-preload] [-nopreload]
     [search-directory-or-module-or-file]]…

The ‘-h’ and ‘-help’ options print brief usage information on standard output and exit. The ‘-v’ option prints the system version string, system time stamp, operating system type, and configure script options on standard output and exits.

The ‘-i’ option can be used to force gsc to process the command line like the interpreter. The only difference with the interpreter is that the compilation related procedures listed in this chapter are also available (i.e. compile-file, compile-file-to-target, etc).

3.1 Interactive mode

When no command line argument is present other than options gsc behaves like gsi in interactive mode.

3.2 Customization

Like the interpreter, the compiler will examine the initialization file unless the ‘-f’ option is specified. Runtime options are explained in Runtime options.

3.3 Batch mode

In batch mode gsc accepts on the command line 3 types of non-options which are processed from left to right: search directories, modules, and files. Search directories are added to the list of module search order directories. Every command line argument that is the name of a module that is found in the list of module search order directories will cause that module to be compiled. Similarly, file names (with either no extension, or a C file extension, or some other extension) on the command line will cause that file to be compiled. The compilation is done for the target language specified with the -target target option. target is either js, for JavaScript, or C, which is the default if no target language is specified.

The recognized C file extensions are ‘.c’, ‘.C’, ‘.cc’, ‘.cp’, ‘.cpp’, ‘.CPP’, ‘.cxx’, ‘.c++’, ‘.m’, ‘.M’, and ‘.mm’.

The extension can be omitted from a file name when the Scheme file has a ‘.scm’, ‘.sld’ or ‘.six’ extension. When the extension of the Scheme file is ‘.six’ the content of the file will be parsed using the Scheme infix syntax extension (see Scheme infix syntax extension). Otherwise, gsc will parse the Scheme file using the normal Scheme prefix syntax. Files with a C file extension must have been previously produced by gsc with the C target and the ‘-c’ option, and are used by the C target Gambit linker.

For each Scheme file the compiler creates a file of target code, either ‘file.c’ or ‘file.js’ for the C and js targets respectively. The file’s name is the same as the Scheme file, but the extension is changed to ‘.c’ or ‘.js’ as appropriate. By default the file is created in the same directory as the Scheme file. This default can be overridden with the compiler’s ‘-o’ option.

The files of target code produced by the compiler serve two purposes. They will be processed by a C compiler or JavaScript VM, and they also contain information to be read by Gambit’s linker to generate a link file. The link file is a file of target code that collects various linking information for a group of modules, such as the set of all symbols and global variables used by the modules. The linker is only invoked when the ‘-link’ or ‘-exe’ options appear on the command line.

Compiler options must be specified before the first file name and after the ‘-:’ runtime option (see section Runtime options). If present, the ‘-i’, ‘-f’, and ‘-v’ compiler options must come first. The available options are:

-i

Force interpreter mode.

-f

Do not examine the initialization file.

-h / -help

Print brief usage information on standard output and exit.

-v

Print the system version string, system time stamp, operating system type, and configure script options on standard output and exit.

-target target

Select the target language.

-prelude expressions

Add expressions to the top of the source code being compiled.

-postlude expressions

Add expressions to the bottom of the source code being compiled.

-cc compiler

Select specific C compiler.

-cc-options options

Add options to the command that invokes the C compiler.

-ld-options-prelude options

Add options to the command that invokes the C linker.

-ld-options options

Add options to the command that invokes the C linker.

-pkg-config pkg-config-args

Use the pkg-config program to determine options for the C compiler and C linker.

-pkg-config-path pkg-config-path

Add a path to the PKG_CONFIG_PATH environment variable.

-warnings

Display warnings.

-verbose

Display a trace of the compiler’s activity.

-report

Display a global variable usage report.

-expansion

Display the source code after expansion.

-gvm

Generate a listing of the GVM code.

-cfg

Generate a control flow graph of the GVM code.

-dg

Generate a dependency graph.

-debug

Include all debugging information in the code generated.

-debug-location

Include source code location debugging information in the code generated.

-debug-source

Include the source code debugging information in the code generated.

-debug-environments

Include environment debugging information in the code generated.

-track-scheme

Generate ‘#line’ directives referring back to the Scheme code.

-o output

Set name of output file or directory where output file(s) are written.

-dynamic

Compile Scheme source files to dynamically loadable object files (this is the default).

-exe

Compile Scheme source files to an executable program (machine code or script).

-obj

Compile Scheme source files to object files by invoking the C compiler.

-keep-temp

Keep any intermediate files that are generated.

-c

Compile Scheme source files to target code without generating a link file.

-link

Compile Scheme source files to target code and generate a link file.

-flat

Generate a flat link file instead of the default incremental link file.

-l base

Specify the link file of the base library to use for the link.

-module-ref module-ref

Specify the reference of the generated module.

-linker-name linker-name

Specify the name of the low-level initialization function exported by the module.

-preload

Turn on ‘preload’ linker bit.

-nopreload

Turn off ‘preload’ linker bit. Start REPL interaction.

-e expressions

Evaluate expressions in the interaction environment.

-nb-gvm-regs n

Specify the number of available Gambit virtual machine registers.

-nb-arg-regs n

Specify the number of procedure call parameters passed in Gambit virtual machine registers.

-compactness level

Specify the compactness of the generated code.

The ‘-i’ option forces the compiler to process the remaining command line arguments like the interpreter.

The ‘-target’ option selects the target language of the compilation. It is either js for JavaScript, or C for C (which is the default).

The ‘-prelude’ option adds the specified expressions to the top of the source code being compiled. It can appear multiple times. The main use of this option is to supply declarations on the command line. For example the following invocation of the compiler will compile the file ‘bench.scm’ in unsafe mode:

$ gsc -prelude "(declare (not safe))" bench.scm

The ‘-postlude’ option adds the specified expressions to the bottom of the source code being compiled. It can appear multiple times. The main use of this option is to supply the expression that will start the execution of the program. For example:

$ gsc -postlude "(start-bench)" bench.scm

The ‘-cc’ option is only meaningful when the C target is selected. The ‘-cc’ option selects the specified C compiler for compiling the generated C code. When this option is used, the default C compiler options that were determined to be needed by the configure script are nullified because they are very likely to be invalid for the specified C compiler. Any options needed for this C compiler should be specified explicitly using the ‘-cc-options’, ‘-ld-options-prelude’, and ‘-ld-options’ options. For example:

$ gsc -cc clang -cc-options "-O0 -bundle" bench.scm # clang on macOS
$ gsc -cc tcc -cc-options -shared bench.scm         # tcc on linux

The ‘-cc-options’ option is only meaningful when the C target is selected and a dynamically loadable object file is being generated (neither the ‘-c’ or ‘-link’ options are used). It can appear multiple times. The ‘-cc-options’ option adds the specified options to the command that invokes the C compiler. The main use of this option is to specify the include path, some symbols to define or undefine, the optimization level, or any C compiler option that is different from the default. For example:

$ gsc -cc-options "-U___SINGLE_HOST -O2 -I../include" bench.scm

The ‘-ld-options-prelude’ and ‘-ld-options’ options are only meaningful when the C target is selected and a dynamically loadable object file is being generated (neither the ‘-c’ or ‘-link’ options are used). They can appear multiple times. The ‘-ld-options-prelude’ and ‘-ld-options’ options add the specified options to the command that invokes the C linker (the options in ld-options-prelude are passed to the C linker before the input file and the options in ld-options are passed after). The main use of this option is to specify additional object files or libraries that need to be linked, or any C linker option that is different from the default (such as the library search path and flags to select between static and dynamic linking). For example:

$ gsc -ld-options "-L/usr/X11R6/lib -lX11 -dynamic" app.scm

The ‘-pkg-config’ is only meaningful when the C target is selected. The ‘-pkg-config’ option will cause the pkg-config program to be invoked to determine the options to add to the command that invokes the C compiler and C linker. It can appear multiple times. The pkg-config program is passed the arguments in the string pkg-config-args in addition to either --cflags or --libs. It is typical for pkg-config-args to be the name of a system library, such as "sqlite3", but other pkg-config options can be specified, such as "--static sqlite3". The ‘-pkg-config-path’ option adds a path to the PKG_CONFIG_PATH environment variable for use by the pkg-config program to find ‘.pc’ files. For example:

$ gsc -pkg-config "x11" -pkg-config-path "/usr/share/pkgconfig" app.scm

The ‘-warnings’ option displays on standard output all warnings that the compiler may have.

The ‘-verbose’ option displays on standard output a trace of the compiler’s activity.

The ‘-report’ option displays on standard output a global variable usage report. Each global variable used in the program is listed with 4 flags that indicate whether the global variable is defined, referenced, mutated and called.

The ‘-expansion’ option displays on standard output the source code after expansion and inlining by the front end.

The ‘-gvm’ option generates a listing of the intermediate code for the “Gambit Virtual Machine” (GVM) of each Scheme file on ‘file.gvm’.

The ‘-cfg’ option generates a visual representation of the control flow graph of the intermediate code for the “Gambit Virtual Machine” (GVM) of each Scheme file on ‘file.cfg’. The file is suitable for processing with the “dot” program. For example, to generate the PDF file ‘file.cfg.pdf’ from ‘file.cfg’ the following command can be used:

$ dot -O -Tpdf file.cfg

The ‘-dg’ option generates a visual representation of the dependency graph of each Scheme file on ‘file.dg’. The file is suitable for processing with the “dot” program. For example, to generate the PDF file ‘file.dg.pdf’ from ‘file.dg’ the following command can be used:

$ dot -O -Tpdf file.dg

The ‘-debug’ option causes all kinds of debugging information to be saved in the code generated. See the documentation of the ‘debug’ declaration for details.

The ‘-debug-location’ option causes source code location debugging information to be saved in the code generated. See the documentation of the ‘debug-location’ declaration for details.

The ‘-debug-source’ option causes source code debugging information to be saved in the code generated. See the documentation of the ‘debug-source’ declaration for details.

The ‘-debug-environments’ option causes environment debugging information to be saved in the code generated. See the documentation of the ‘debug-environments’ declaration for details.

The ‘-track-scheme’ option is only meaningful when the C target is selected. The ‘-track-scheme’ option causes the generation of ‘#line’ directives that refer back to the Scheme source code. This allows the use of a C debugger or profiler to debug Scheme code.

The ‘-o’ option sets the filename of the output file, or the directory in which the output file(s) generated by the compiler are written.

If the ‘-link’ or ‘-exe’ options appear on the command line, the Gambit linker is invoked to generate the link file from the set of files specified on the command line or produced by the Gambit compiler. By default the link file is named after the last file on the compiler’s command line. If the last file stripped of it’s extension is ‘last’ then the link file is ‘last_.c’ for the C target and ‘last_.js’ for the js target. When the ‘-c’ option is specified, the Scheme source files are compiled to target files without invoking the linker, which is useful for separate compilation of modules. When the ‘-exe’ option is specified, the generated target files and link file are combined to produce an executable program whose name defaults to ‘last’ on Unix, and ‘last.exe’ or ‘last.bat’ on Windows depending on whether a machine code executable or script is produced. When the C target is selected and the ‘-obj’ option is specified, the generated C files are compiled using the C compiler to produce object files (‘.o’ or ‘.obj’ extensions). If neither the ‘-link’, ‘-c’, ‘-exe’, or ‘-obj’ options appear on the command line, the Scheme source files are compiled to dynamically loadable object files (‘.on’ extension). The ‘-keep-temp’ option will prevent the deletion of any intermediate files that are generated. Note that in this case the intermediate file will be generated in the same directory as the Scheme source file even if the ‘-o’ option is used.

The ‘-flat’ option is only meaningful when a link file is being generated (i.e. the ‘-link’ or ‘-exe’ options also appear on the command line). The ‘-flat’ option directs the Gambit linker to generate a flat link file. By default, the linker generates an incremental link file (see the next section for a description of the two types of link files).

The ‘-l’ option is only meaningful when an incremental link file is being generated (i.e. the ‘-link’ or ‘-exe’ options appear on the command line and the ‘-flat’ option is absent). The ‘-l’ option specifies the link file (without the ‘.c’ or ‘.js’ extension) of the base library to use for the incremental link. By default the link file of the Gambit runtime library is used (i.e. ‘~~lib/_gambit’).

The ‘-preload’ and ‘-nopreload’ options are only meaningful when a link file is being generated. The ‘-preload’ option turns on the ‘preload’ linker bit for the modules that follow on the command line. The following modules will be loaded unconditionally at program startup and in command line order (this is the default for compatibility with how legacy modules have been handled in the past). The ‘-nopreload’ option turns off the ‘preload’ linker bit. The following modules will be loaded only to satisfy the module dependencies of the ##demand-module form.

The ‘-’ option starts a REPL interaction.

The ‘-e’ option evaluates the specified expressions in the interaction environment.

The ‘-nb-gvm-regs’ option specifies the number of Gambit virtual machine registers that are available for the generated code. The default number depends on configuration options and the target but it is typically 5. All modules and the runtime library must be compiled with the same setting. This option exists mainly for experimentation by the developers. For example:

$ gsc -nb-gvm-regs 10 -c bench.scm

The ‘-nb-arg-regs’ option specifies the number of procedure call parameters passed in Gambit virtual machine registers. The default number depends on configuration options and the target but it is typically 3. All modules and the runtime library must be compiled with the same setting. This option exists mainly for experimentation by the developers. For example:

$ gsc -nb-arg-regs 2 -c bench.scm

The ‘-compactness’ option selects the level of compactness of the generated code. The default level depends on configuration options and the target but it is typically 5. Levels from 0 to 5 cause the generation of increasingly compact code with little or no impact on execution speed. Lower values tend to make the generated code more humanly readable. Above a level of 5 the compiler will trade execution speed for saving code space. The detailed meaning of this option depends on the target, some targets may ignore it and some targets may require all modules and the runtime library to be compiled with the same compactness level. For example:

$ gsc -target js -compactness 0 -c bench.scm

3.4 Link files

Gambit can be used to create programs and libraries of Scheme modules. This section explains the steps required to do so and the role played by the link files.

In general, a program is composed of a set of Scheme modules and modules in the target language. Some of the modules are part of the Gambit runtime library and the other modules are supplied by the user. When the program is started it must setup various global tables (including the symbol table and the global variable table) and then sequentially execute the Scheme modules (more or less as though they were being loaded one after another). The information required for this is contained in one or more link files generated by the Gambit linker from the target files produced by the Gambit compiler.

The order of execution of the Scheme modules corresponds to the order of the modules on the command line which produced the link file. The order is usually important because most modules define variables and procedures which are used by other modules (for this reason the program’s main computation is normally started by the last module).

When a single link file is used to contain the linking information of all the Scheme modules it is called a flat link file. Thus a program built with a flat link file contains in its link file both information on the user modules and on the runtime library. This is fine if the program is to be statically linked but is wasteful in a shared-library context because the linking information of the runtime library can’t be shared and will be duplicated in all programs (this linking information typically takes hundreds of kilobytes).

Flat link files are mainly useful to bundle multiple Scheme modules to make a runtime library (such as the Gambit runtime library) or to make a single file that can be loaded with the load procedure.

An incremental link file contains only the linking information that is not already contained in a second link file (the “base” link file). Assuming that a flat link file was produced when the runtime library was linked, a program can be built by linking the user modules with the runtime library’s link file, producing an incremental link file. This allows the creation of a shared-library which contains the modules of the runtime library and its flat link file. The program is dynamically linked with this shared-library and only contains the user modules and the incremental link file. For small programs this approach greatly reduces the size of the program because the incremental link file is small. A “hello world” program built this way can be as small as 5 Kbytes. Note that it is perfectly fine to use an incremental link file for statically linked programs (there is very little loss compared to a single flat link file).

Incremental link files may be built from other incremental link files. This allows the creation of shared-libraries which extend the functionality of the Gambit runtime library.

3.4.1 Building an executable program

The simplest way to create an executable program is to invoke gsc with the ‘-exe’ option. The compiler will transparently perform all the steps necessary, including compiling Scheme source files to target files, generating the link file, and (when the C target is selected) compiling the C files generated to object files and creating the final executable file using the C linker. The following example shows how to use the C target to build the executable program ‘hello.exe’ which contains the two Scheme modules ‘h.scm’ and ‘w.six’.

$ cat h.scm
(display "hello") (newline)
$ cat w.six
display("world"); newline();
$ gsc -o hello.exe -exe h.scm w.six
h.scm:
/Users/feeley/gambit/doc/h.c:
w.six:
/Users/feeley/gambit/doc/w.c:
/Users/feeley/gambit/doc/w_.c:
$ ./hello.exe
hello
world

The detailed steps which are performed can be viewed by setting the ‘GAMBUILD_VERBOSE’ environment variable to a nonnull value. Alternatively, gsc’s ‘-verbose’ option can be used (it implicitly sets the ‘GAMBUILD_VERBOSE’ environment variable). For example:

$ export GAMBUILD_VERBOSE=yes
$ gsc -o hello.exe -exe h.scm w.six
h.scm:
/Users/feeley/gambit/doc/h.c:
gcc  -O1    -Wno-unused -Wno-write-strings -Wdisabled-optimization
fwrapv -fno-strict-aliasing -fno-trapping-math -fno-math-errno
-fschedule-insns2 -foptimize-sibling-calls -fomit-frame-pointer -fPIC
-fno-common -mpc64   -D___SINGLE_HOST  -I"/usr/local/Gambit/include"
-c -o 'h.o'  'h.c'
w.six:
/Users/feeley/gambit/doc/w.c:
gcc  -O1    -Wno-unused -Wno-write-strings -Wdisabled-optimization
-fwrapv -fno-strict-aliasing -fno-trapping-math -fno-math-errno
-fschedule-insns2 -foptimize-sibling-calls -fomit-frame-pointer -fPIC
-fno-common -mpc64   -D___SINGLE_HOST  -I"/usr/local/Gambit/include"
-c -o 'w.o'  'w.c'
/Users/feeley/gambit/doc/w_.c:
gcc  -O1    -Wno-unused -Wno-write-strings -Wdisabled-optimization
-fwrapv -fno-strict-aliasing -fno-trapping-math -fno-math-errno
-fschedule-insns2 -foptimize-sibling-calls -fomit-frame-pointer -fPIC
-fno-common -mpc64   -D___SINGLE_HOST  -I"/usr/local/Gambit/include"
-c -o 'w_.o'  'w_.c'
gcc     -Wno-unused -Wno-write-strings -Wdisabled-optimization
-fwrapv -fno-strict-aliasing -fno-trapping-math -fno-math-errno
-fschedule-insns2 -foptimize-sibling-calls -fomit-frame-pointer -fPIC
-fno-common -mpc64    -D___SINGLE_HOST  -I"/usr/local/Gambit/include"
-o 'hello.exe'   'w_.o' 'h.o' 'w.o' "/usr/local/Gambit/lib/libgambit.a"

Here is the same example using the js target showing the creation of a shell script invoking nodejs:

$ export GAMBUILD_VERBOSE=yes
$ gsc -target js -o hello.exe -exe h.scm w.six
h.scm:
/Users/feeley/gambit/doc/h.js:
cat h.js > "h.o"
w.six:
/Users/feeley/gambit/doc/w.js:
cat w.js > "w.o"
/Users/feeley/gambit/doc/w_.js:
cat w_.js > "w_.o"
echo "#! /usr/bin/env node" > "hello.exe"
cat w_.o h.o w.o "/usr/local/Gambit/lib/_gambit.js" >> "hello.exe"
chmod +x "hello.exe"

Using a single invocation of gsc with the ‘-exe’ option is sometimes inappropriate when the build process is more complex, for example when the program is composed of several separately compiled modules. In such a case it is useful to decompose the build process into smaller compilation steps. The ‘hello.exe’ executable program could have been built with the C target by separating the generation of C files from the C compilation and linking:

$ gsc -c h.scm
$ gsc -c w.six
$ gsc -o hello.exe -exe h.c w.c

When even finer control is desired the C target’s build process can be decomposed into smaller steps that invoke the C compiler and linker explicitly. This is described in the rest of this section.

The gsc compiler can be invoked to compile each Scheme module into a C file and to create an incremental link file. The C files and the link file must then be compiled with a C compiler and linked (at the object file level) with the Gambit runtime library and possibly other libraries (such as the math library and the dynamic loading library).

Here is for example how a program with three modules (one in C and two in Scheme) can be built. The content of the three source files (‘m1.c’, ‘m2.scm’ and ‘m3.scm’) is:

/* File: "m1.c" */
int power_of_2 (int x) { return 1<<x; }

; File: "m2.scm"
(c-declare "extern int power_of_2 ();")
(define pow2 (c-lambda (int) int "power_of_2"))
(define (twice x) (cons x x))

; File: "m3.scm"
(write (map twice (map pow2 '(1 2 3 4)))) (newline)

The compilation of the two Scheme source files can be done with three invocations of gsc:

$ gsc -c m2.scm        # create m2.c (note: .scm is optional)
$ gsc -c m3.scm        # create m3.c (note: .scm is optional)
$ gsc -link m2.c m3.c  # create the incremental link file m3_.c

Alternatively, the three invocations of gsc can be replaced by a single invocation:

$ gsc -link m2 m3
m2:
m3:

At this point there will be 4 C files: ‘m1.c’, ‘m2.c’, ‘m3.c’, and ‘m3_.c’. To produce an executable program these files must be compiled with a C compiler and linked with the Gambit runtime library. The C compiler options needed will depend on the C compiler and the operating system (in particular it may be necessary to add the options ‘-I/usr/local/Gambit/include -L/usr/local/Gambit/lib’ to access the ‘gambit.h’ header file and the Gambit runtime library).

Here is an example under macOS:

$ uname -srmp
Darwin 20.6.0 x86_64 i386
$ gsc -obj m1.c m2.c m3.c m3_.c
m1.c:
m2.c:
m3.c:
m3_.c:
$ gcc m1.o m2.o m3.o m3_.o -lgambit
$ ./a.out
((2 . 2) (4 . 4) (8 . 8) (16 . 16))

Here is an example under Linux:

$ uname -srmp
Linux 5.10.0-9-amd64 x86_64 unknown
$ gsc -obj m1.c m2.c m3.c m3_.c
m1.c:
m2.c:
m3.c:
m3_.c:
$ gcc m1.o m2.o m3.o m3_.o -lgambit -lm -ldl -lutil
$ ./a.out
((2 . 2) (4 . 4) (8 . 8) (16 . 16))

3.4.2 Building a loadable library

To bundle multiple modules into a single object file that can be dynamically loaded with the load procedure, a flat link file is needed. The compiler’s ‘-o’ option must be used to name the C file generated as follows. If the dynamically loadable object file is to be named ‘myfile.on’ then the ‘-o’ option must set the name of the link file generated to ‘myfile.on.c’ (note that the ‘.c’ extension could also be ‘.cc’, ‘.cpp’ or whatever extension is appropriate for C/C++ source files). The three modules of the previous example can be bundled by generating a link file in this way:

$ gsc -link -flat -o foo.o1.c m2 m3
m2:
m3:
*** WARNING -- "cons" is not defined,
***            referenced in: ("m2.c")
*** WARNING -- "map" is not defined,
***            referenced in: ("m3.c")
*** WARNING -- "newline" is not defined,
***            referenced in: ("m3.c")
*** WARNING -- "write" is not defined,
***            referenced in: ("m3.c")

The warnings indicate that there are no definitions (defines or set!s) of the variables cons, map, newline and write in the set of modules being linked. Before ‘foo.o1’ is loaded, these variables will have to be bound; either implicitly (by the runtime library) or explicitly.

When compiling the C files and link file generated, the flag ‘-D___DYNAMIC’ must be passed to the C compiler and the C compiler and linker must be told to generate a dynamically loadable shared library.

Here is an example under macOS:

$ uname -srmp
Darwin 20.6.0 x86_64 i386
$ gsc -link -flat -o foo.o1.c m2 m3 > /dev/null
m2:
m3:
$ gsc -cc-options "-D___DYNAMIC" -obj m1.c m2.c m3.c foo.o1.c
m1.c:
m2.c:
m3.c:
foo.o1.c:
$ gcc -bundle m1.o m2.o m3.o foo.o1.o -o foo.o1
$ gsi foo.o1
((2 . 2) (4 . 4) (8 . 8) (16 . 16))

Here is an example under Linux:

$ uname -srmp
Linux 5.10.0-9-amd64 x86_64 unknown
$ gsc -link -flat -o foo.o1.c m2 m3 > /dev/null
m2:
m3:
$ gsc -cc-options "-D___DYNAMIC" -obj m1.c m2.c m3.c foo.o1.c
m1.c:
m2.c:
m3.c:
foo.o1.c:
$ gcc -shared m1.o m2.o m3.o foo.o1.o -o foo.o1
$ gsi foo.o1
((2 . 2) (4 . 4) (8 . 8) (16 . 16))

Here is a more complex example, under Solaris, which shows how to build a loadable library ‘mymod.o1’ composed of the files ‘m4.scm’, ‘m5.scm’ and ‘x.c’ that links to system shared libraries (for X-windows):

$ uname -srmp
SunOS ungava 5.6 Generic_105181-05 sun4m sparc SUNW,SPARCstation-20
$ gsc -link -flat -o mymod.o1.c m4 m5
m4:
m5:
*** WARNING -- "*" is not defined,
***            referenced in: ("m4.c")
*** WARNING -- "+" is not defined,
***            referenced in: ("m5.c")
*** WARNING -- "display" is not defined,
***            referenced in: ("m5.c" "m4.c")
*** WARNING -- "newline" is not defined,
***            referenced in: ("m5.c" "m4.c")
*** WARNING -- "write" is not defined,
***            referenced in: ("m5.c")
$ gsc -cc-options "-D___DYNAMIC" -obj m4.c m5.c x.c mymod.o1.c
m4.c:
m5.c:
x.c:
mymod.o1.c:
$ /usr/ccs/bin/ld -G -o mymod.o1 mymod.o1.o m4.o m5.o x.o -lX11 -lsocket
$ gsi mymod.o1
hello from m4
hello from m5
(f1 10) = 22
$ cat m4.scm
(define (f1 x) (* 2 (f2 x)))
(display "hello from m4")
(newline)

(c-declare #<<c-declare-end
#include "x.h"
c-declare-end
)
(define x-initialize (c-lambda (char-string) bool "x_initialize"))
(define x-display-name (c-lambda () char-string "x_display_name"))
(define x-bell (c-lambda (int) void "x_bell"))
$ cat m5.scm
(define (f2 x) (+ x 1))
(display "hello from m5")
(newline)

(display "(f1 10) = ")
(write (f1 10))
(newline)

(x-initialize (x-display-name))
(x-bell 50) ; sound the bell at 50%
$ cat x.c
#include <X11/Xlib.h>

static Display *display;

int x_initialize (char *display_name)
{
  display = XOpenDisplay (display_name);
  return display != NULL;
}

char *x_display_name (void)
{
  return XDisplayName (NULL);
}

void x_bell (int volume)
{
  XBell (display, volume);
  XFlush (display);
}
$ cat x.h
int x_initialize (char *display_name);
char *x_display_name (void);
void x_bell (int);

3.4.3 Building a shared-library

A shared-library can be built using an incremental link file or a flat link file. An incremental link file is normally used when the Gambit runtime library (or some other library) is to be extended with new procedures. A flat link file is mainly useful when building a “primal” runtime library, which is a library (such as the Gambit runtime library) that does not extend another library. When compiling the C files and link file generated, the flags ‘-D___LIBRARY’ and ‘-D___SHARED’ must be passed to the C compiler. The flag ‘-D___PRIMAL’ must also be passed to the C compiler when a primal library is being built.

A shared-library ‘mylib.so’ containing the two first modules of the previous example can be built this way:

$ uname -srmp
Linux 5.10.0-9-amd64 x86_64 unknown
$ gsc -link -o mylib.c m2
$ gsc -obj -cc-options "-D___SHARED" m1.c m2.c mylib.c
m1.c:
m2.c:
mylib.c:
$ gcc -shared  m1.o m2.o mylib.o -o mylib.so

Note that this shared-library is built using an incremental link file (it extends the Gambit runtime library with the procedures pow2 and twice). This shared-library can in turn be used to build an executable program from the third module of the previous example:

$ gsc -link -l mylib m3
$ gsc -obj m3.c m3_.c
m3.c:
m3_.c:
$ gcc m3.o m3_.o mylib.so -lgambit
$ LD_LIBRARY_PATH=.:/usr/local/lib ./a.out
((2 . 2) (4 . 4) (8 . 8) (16 . 16))

3.4.4 Other compilation options

The performance of the code can be increased by passing the ‘-D___SINGLE_HOST’ flag to the C compiler. This will merge all the procedures of a module into a single C procedure, which reduces the cost of intra-module procedure calls. In addition the ‘-O2’ option can be passed to the C compiler. For large modules, it will not be practical to specify both ‘-O2’ and ‘-D___SINGLE_HOST’ for typical C compilers because the compile time will be high and the C compiler might even fail to compile the program for lack of memory. It has been observed that lower levels of optimization (e.g. ‘-O1’) often give faster compilation and also generate faster code. It is a good idea to experiment.

Normally C compilers will not automatically search ‘/usr/local/Gambit/include’ for header files so the flag ‘-I/usr/local/Gambit/include’ should be passed to the C compiler. Similarly, C compilers/linkers will not automatically search ‘/usr/local/Gambit/lib’ for libraries so the flag ‘-L/usr/local/Gambit/lib’ should be passed to the C compiler/linker. Alternatives are given in Accessing the system files.

A variety of flags are needed by some C compilers when compiling a shared-library or a dynamically loadable library. Some of these flags are: ‘-shared’, ‘-call_shared’, ‘-rdynamic’, ‘-fpic’, ‘-fPIC’, ‘-Kpic’, ‘-KPIC’, ‘-pic’, ‘+z’, ‘-G’. Check your compiler’s documentation to see which flag you need.

3.5 Procedures specific to compiler

The Gambit Scheme compiler features the following procedures that are not available in the Gambit Scheme interpreter.

The file parameter must be a string. If expression is not specified, file must name an existing file containing Scheme source code. The extension can be omitted from file when the Scheme file has a ‘.scm’, ‘.sld’ or ‘.six’ extension. By default, this procedure compiles the source file into a file containing C code. A different target language can be selected in the options. The generated file is named after file with the extension replaced with ‘.c’ or ‘.js’, as appropriate for the target selected. The name of the generated file can also be specified directly with the output parameter. If output is a string naming a directory then the generated file is created in that directory. Otherwise the name of the generated file is output.

Compilation options are specified through the options parameter which must be an association list. Any combination of the following options can be used: ‘target’, ‘verbose’, ‘report’, ‘expansion’, ‘gvm’, ‘debug’, ‘module-ref’, and ‘linker-name’.

When expression is specified, the file parameter is not open or read. Instead, expression is used as though it was the content of the file. This makes it possible to compile source code without having to create a file to contain the code. Note that file is used in error messages and to determine the output file name if output is not specified.

When the compilation is successful, compile-file-to-target returns the name of the file generated. When there is a compilation error, #f is returned.

$ cat h.scm
(display "hello") (newline)
$ gsc
Gambit v4.9.4

> (compile-file-to-target "h")
"/Users/feeley/gambit/doc/h.c"

The file, options, output, and expression parameters have the same meaning as for the compile-file-to-target procedure, except that file may be a Scheme source file or a file possibly generated by the Gambit Scheme compiler (for example with the compile-file-to-target procedure). The cc-options parameter is a string containing the options to pass to the C compiler and the ld-options-prelude and ld-options parameters are strings containing the options to pass to the C linker (the options in ld-options-prelude are passed to the C linker before the input file and the options in ld-options are passed after).

The compile-file procedure compiles the source file file into an object file, which is either a file dynamically loadable using the load procedure, or a C linkable object file destined to be linked with the C linker (for example to create a standalone executable program). The presence of the obj option in options will cause the creation of a C linkable object file and therefore the options ld-options-prelude and ld-options are ignored, otherwise a dynamically loadable file is created. In both cases, if file is a Scheme source file, the compiler first compiles file to a C file which is created in the same directory as file regardless of the output parameter. Then the C file is compiled with the C compiler.

When the compilation is successful, compile-file returns the name of the object file generated. When there is a compilation error, #f is returned.

The name of the object file can be specified with the output parameter. If output is a string naming a directory then the object file is created in that directory. Otherwise the name of the object file is output.

In the case of a dynamically loadable object file, by default the object file is named after file with the extension replaced with ‘.on’, where n is a positive integer that acts as a version number. The next available version number is generated automatically by compile-file.

When dynamically loaded object files are loaded using the load procedure, the ‘.on’ extension can be specified (to select a particular version) or omitted (to load the file with a ‘.on’ extension with the highest n consecutively from 1). When the ‘.on’ extension is not specified and older versions are no longer needed, all versions must be deleted and the compilation must be repeated (this is necessary because the file name, including the extension, is used to name some of the exported symbols of the object file).

Note that dynamically loadable object files can only be generated on host operating systems that support dynamic loading.

$ cat h.scm
(display "hello") (newline)
$ gsc
Gambit v4.9.4

> (compile-file "h")
"/Users/feeley/gambit/doc/h.o1"
> (load "h")
hello
"/Users/feeley/gambit/doc/h.o1"
> (compile-file-to-target "h" output: "h.o99.c")
"/Users/feeley/gambit/doc/h.o99.c"
> (compile-file "h.o99.c")
"/Users/feeley/gambit/doc/h.o99"
> (load "h.o99")
hello
"/Users/feeley/gambit/doc/h.o99"
> (compile-file-to-target "h")
"/Users/feeley/gambit/doc/h.c"
> (compile-file "h.c" options: '(obj))
"/Users/feeley/gambit/doc/h.o"

The first parameter must be a non empty list of strings naming Scheme modules to link (the file extension may be omitted). An incremental link file is generated for the modules specified in module-list. By default the link file generated is named ‘last_.ext’, where last is the name of the last module, without the file extension, and ext is the appropriate extension for the target. The name of the generated link file can be specified with the output parameter. If output is a string naming a directory then the link file is created in that directory. Otherwise the name of the link file is output.

The base link file is specified by the base parameter, which must be a string. By default the base link file is the Gambit runtime library link file ‘~~lib/_gambit’ (with extension appropriate for the target). However, when base is supplied it is the name of the base link file (the file extension may be omitted).

The warnings? parameter controls whether warnings are generated for undefined references.

The following example shows how to build the executable program ‘hello’ which contains the two Scheme modules ‘h.scm’ and ‘w.six’.

$ uname -srmp
Darwin 8.1.0 Power Macintosh powerpc
$ cat h.scm
(display "hello") (newline)
$ cat w.six
display("world"); newline();
$ gsc
Gambit v4.9.4

> (compile-file-to-target "h")
"/Users/feeley/gambit/doc/h.c"
> (compile-file-to-target "w")
"/Users/feeley/gambit/doc/w.c"
> (link-incremental '("h" "w") output: "hello.c")
"/Users/feeley/gambit/doc/hello_.c"
> ,q
$ gsc -obj h.c w.c hello.c
h.c:
w.c:
hello.c:
$ gcc h.o w.o hello.o -lgambit -o hello
$ ./hello
hello
world

The first parameter must be a non empty list of strings naming Scheme modules to link (the file extension may be omitted). The first string must be the name of a Scheme module or the name of a link file and the remaining strings must name Scheme modules. A flat link file is generated for the modules specified in module-list. By default the link file generated is named ‘last_.ext’, where last is the name of the last module, without the file extension, and ext is the appropriate extension for the target. The name of the generated link file can be specified with the output parameter. If output is a string naming a directory then the link file is created in that directory. Otherwise the name of the link file is output. If a dynamically loadable object file is produced from the link file ‘output’, then the name of the dynamically loadable object file must be ‘output’ stripped of its file extension.

The warnings? parameter controls whether warnings are generated for undefined references.

The following example shows how to build the dynamically loadable object file ‘lib.o1’ which contains the two Scheme modules ‘m6.scm’ and ‘m7.scm’.

$ uname -srmp
Darwin 8.1.0 Power Macintosh powerpc
$ cat m6.scm
(define (f x) (g (* x x)))
$ cat m7.scm
(define (g y) (+ n y))
$ gsc
Gambit v4.9.4

> (compile-file-to-target "m6")
"/Users/feeley/gambit/doc/m6.c"
> (compile-file-to-target "m7")
"/Users/feeley/gambit/doc/m7.c"
> (link-flat '("m6" "m7") output: "lib.o1.c")
*** WARNING -- "*" is not defined,
***            referenced in: ("m6.c")
*** WARNING -- "+" is not defined,
***            referenced in: ("m7.c")
*** WARNING -- "n" is not defined,
***            referenced in: ("m7.c")
"/Users/feeley/gambit/doc/lib.o1.c"
> ,q
$ gcc -bundle -D___DYNAMIC m6.c m7.c lib.o1.c -o lib.o1
$ gsc
Gambit v4.9.4

> (load "lib")
*** WARNING -- Variable "n" used in module "m7" is undefined
"/Users/feeley/gambit/doc/lib.o1"
> (define n 10)
> (f 5)
35
> ,q

The warnings indicate that there are no definitions (defines or set!s) of the variables *, + and n in the modules contained in the library. Before the library is used, these variables will have to be bound; either implicitly (by the runtime library) or explicitly.

4. Runtime options

Both gsi and gsc as well as executable programs compiled and linked using gsc take a ‘-:’ option which supplies parameters to the runtime system. This option must appear first on the command line. The colon is followed by a comma separated list of options with no intervening spaces. The available options are:

min-heap=SIZE or the shorthand mSIZE

Set minimum heap size.

max-heap=SIZE or the shorthand hSIZE

Set maximum heap size.

live-ratio=RATIO or the shorthand lRATIO

Set the ratio of heap that is live after a garbage collection.

gambit or the (deprecated) shorthand S

Select Gambit Scheme mode. This is the default mode.

r5rs or the (deprecated) shorthand s

Select R5RS Scheme mode.

r7rs

Select R7RS Scheme mode.

debug[=[OPT...]] or the shorthand d[OPT...]

Set debugging options.

~~NAME=DIRECTORY

Override the NAME installation directory.

add-arg=ARGUMENT or the shorthand +ARGUMENT

Add ARGUMENT to the command line before other arguments.

io-settings=[IO...] or the shorthand i[IO...]

Set general I/O settings.

file-settings=[IO...] or the shorthand f[IO...]

Set general file I/O settings.

stdio-settings=[IO...] or the shorthand -[IO...]

Set general stdio settings.

0[IO...]

Set stdin settings.

1[IO...]

Set stdout settings.

2[IO...]

Set stderr settings.

terminal-settings=[IO...] or the shorthand t[IO...]

Set terminal I/O settings.

search=[DIR]

Set or reset module search order.

whitelist=[SOURCE]

Set or reset the whitelist of trusted sources for automatic installation of hosted modules.

ask-install=WHEN

Set automatic installation confirmation mode.

The min-heap=SIZE and max-heap=SIZE options set limits on the size of the heap. The SIZE is an integer that may be followed by G (gigabytes), M (megabytes), or K or nothing (kilobytes). The heap will not shrink lower than the minimum heap size which defaults to 0. The heap will not grow larger than the maximum heap size if it is set (by default the heap may grow until the virtual memory is exhausted).

The live-ratio=RATIO option sets the percentage of the heap that will be occupied with live objects after the heap is resized at the end of a garbage collection. RATIO is an integer between 1 and 100 inclusively indicating the desired percentage. The garbage collector resizes the heap to reach this percentage occupation (roughly), within the limits of the min-heap and max-heap options. By default, the percentage is 50.

The gambit, r5rs and r7rs options configure the runtime system to conform to Gambit Scheme, R5RS Scheme and R7RS Scheme respectively. The reader is case-insensitive in r5rs mode, and is case-sensitive in r7rs and gambit modes. The reader supports keywords only in gambit mode, which is the default mode.

The debug=OPT,... option sets various debugging options. The equal sign is followed by a sequence of letters indicating suboptions.

p

Uncaught exceptions will be treated as “errors” in the primordial thread only.

a

Uncaught exceptions will be treated as “errors” in all threads.

r

When an “error” occurs a new REPL will be started.

s

When an “error” occurs a new REPL will be started. Moreover the program starts in single-stepping mode.

q

When an “error” occurs the program will terminate with a nonzero exit status.

R

When a user interrupt occurs a new REPL will be started. User interrupts are typically obtained by typing <^C>. Note that with some system configurations <^C> abruptly terminates the process. For example, under Microsoft Windows, <^C> works fine with the standard console but with the MSYS terminal window it terminates the process.

D

When a user interrupt occurs it will be deferred until the parameter object current-user-interrupt-handler is set or bound.

Q

When a user interrupt occurs the program will terminate with a nonzero exit status.

LEVEL

The verbosity level is set to LEVEL, a digit from 0 to 9. At level 0 the runtime system will not display error messages and warnings. At level 1 and above error messages and warnings are displayed. At level 2 and above a backtrace is displayed. At level 3 and above variable bindings are displayed in the backtrace. At level 5 and above garbage collection reports are displayed during program execution.

c

The REPL interaction channel will be the console.

-

The REPL interaction channel will be standard input and standard output.

+

The REPL interaction channel will be standard input and standard output and standard error.

@[HOST][:PORT]

When a REPL is started by a thread a connection will be established with the address HOST:PORT and that will be the REPL’s interaction channel. The default HOST is 127.0.0.1 and the default PORT is 44556.

$[INTF][:PORT]

The runtime system will open a socket to listen on port number PORT for incoming connections on the network interface with address INTF. The default INTF is 127.0.0.1 and the default PORT is 44555.

The default debugging options are equivalent to debug=pqQ1- (i.e. an uncaught exception in the primordial thread terminates the program after displaying an error message). When the option debug is used without suboptions it is equivalent to debug=prR1- (i.e. a new REPL is started only when an uncaught exception occurs in the primordial thread). When gsi and gsc are running the main REPL, the debugging options are changed to cause errors in the primordial thread and user interrupts to start a nested REPL.

The ~~NAME=DIRECTORY option overrides the setting of the NAME installation directory. If NAME is empty, it will override the central installation directory.

The add-arg=ARGUMENT option adds the text that follows to the command line before other arguments.

The option io-settings=[IO...] sets the default I/O settings of all types of ports. The option file-settings=[IO...] sets the default I/O settings for ports associated to files. The option stdio-settings=[IO...] sets the default I/O settings for ports associated to stdio (but finer control is possible with 0[IO...], 1[IO...], and 2[IO...] that set the I/O settings of stdin, stdout, and stderr respectively). The option terminal-settings=[IO...] overrides the default I/O settings for ports associated to terminals. The default character encoding, end-of-line encoding and buffering can be set. Moreover, for terminals the line-editing feature can be enabled or disabled. Each IO is a one or two letter code as follows:

A

ASCII character encoding.

1

ISO-8859-1 character encoding.

2

UCS-2 character encoding.

4

UCS-4 character encoding.

6

UTF-16 character encoding.

8

UTF-8 character encoding.

U

UTF character encoding with fallback to UTF-8 on input if no BOM is present.

UA

UTF character encoding with fallback to ASCII on input if no BOM is present.

U1

UTF character encoding with fallback to ISO-8859-1 on input if no BOM is present.

U6

UTF character encoding with fallback to UTF-16 on input if no BOM is present.

U8

UTF character encoding with fallback to UTF-8 on input if no BOM is present.

L

If the LC_ALL or LC_CTYPE or LANG environment variables end with ‘.UTF-8’ or ‘.ISO-8859-1’ or ‘.LATIN-1’ (or a variation) set the character encoding accordingly.

c

End-of-line is encoded as CR (carriage-return).

l

End-of-line is encoded as LF (linefeed)

cl

End-of-line is encoded as CR-LF.

u

Unbuffered I/O.

n

Line buffered I/O (‘n’ for “at newline”).

f

Fully buffered I/O.

r

Illegal character encoding is treated as an error (exception raised).

R

Silently replace illegal character encodings with Unicode character #xfffd (replacement character).

e

Enable line-editing (applies to terminals only).

E

Disable line-editing (applies to terminals only).

The search=[DIR] option adds DIR to the head of the list of module search order directories, unless DIR is empty, in which case it is set to the empty list. The initial setting of the list of module search order directories is ~~lib followed by ~~userlib.

When a hosted module can’t be found in the directories on the list of module search order directories it will be automatically installed if it is from a source on the whitelist of trusted sources, which initially contains only github.com/gambit. The whitelist=[SOURCE] option adds SOURCE to the whitelist, unless SOURCE is empty in which case the whitelist will be set to the empty list (no source is trusted).

The ask-install=WHEN option sets the automatic installation mode confirmation mode to WHEN, which is one of always, repl, and never. When a hosted module can’t be found in the directories on the list of module search order directories and it is from a source not on the whitelist the runtime system will ask for installation confirmation when WHEN is always, or when a REPL has already been started for the current thread and WHEN is repl. In the never mode the runtime system will not install the module automatically. The default mode is repl.

When a program’s execution starts, the runtime system obtains the runtime options by processing in turn various sources of runtime options: the defaults, the environment variable ‘GAMBOPT’, the script line of the source code, and, unless the program is an interpreted script, the first command line argument of the program. Any runtime option can be overriden by a subsequent source of runtime options. It is sometimes useful to prevent overriding the runtime options of the script line. This can be achieved by starting the script line runtime options with ‘-:,’. In this case the environment variable ‘GAMBOPT’ is ignored, and the first command line argument of the program is not used for runtime options (it is treated like a normal command line argument even if it starts with ‘-:’).

For example:

$ export GAMBOPT=debug=0,~~=~/my-gambit2
$ gsi -e '(pretty-print (path-expand "~~")) (/ 1 0)'
"/Users/feeley/my-gambit2/"
$ echo $?
70
$ gsi -:debug=1 -e '(pretty-print (path-expand "~~")) (/ 1 0)'
"/Users/feeley/my-gambit2/"
*** ERROR IN (string)@1.35 -- Divide by zero
(/ 1 0)

5. Debugging

5.1 Debugging model

The evaluation of an expression may stop before it is completed for the following reasons:

1. An evaluation error has occured, such as attempting to divide by zero.
2. The user has interrupted the evaluation (usually by typing <^C>).
3. A breakpoint has been reached or (step) was evaluated.
4. Single-stepping mode is enabled.

When an evaluation stops, a message is displayed indicating the reason and location where the evaluation was stopped. The location information includes, if known, the name of the procedure where the evaluation was stopped and the source code location in the format ‘stream@line.column’, where stream is either a string naming a file or a symbol within parentheses, such as ‘(console)’.

A nested REPL is then initiated in the context of the point of execution where the evaluation was stopped. The nested REPL’s continuation and evaluation environment are the same as the point where the evaluation was stopped. For example when evaluating the expression ‘(let ((y (- 1 1))) (* (/ x y) 2))’, a “divide by zero” error is reported and the nested REPL’s continuation is the one that takes the result and multiplies it by two. The REPL’s lexical environment includes the lexical variable ‘y’. This allows the inspection of the evaluation context (i.e. the lexical and dynamic environments and continuation), which is particularly useful to determine the exact location and cause of an error.

The prompt of nested REPLs includes the nesting level; ‘1>’ is the prompt at the first nesting level, ‘2>’ at the second nesting level, and so on. An end of file (usually <^D>) will cause the current REPL to be terminated and the enclosing REPL (one nesting level less) to be resumed.

At any time the user can examine the frames in the REPL’s continuation, which is useful to determine which chain of procedure calls lead to an error. A backtrace that lists the chain of active continuation frames in the REPL’s continuation can be obtained with the ‘,b’ command. The frames are numbered from 0, that is frame 0 is the most recent frame of the continuation where execution stopped, frame 1 is the parent frame of frame 0, and so on. It is also possible to move the REPL to a specific parent continuation (i.e. a specific frame of the continuation where execution stopped) with the ‘,N’, ‘,N+’, ‘,N-’, ‘,+’, ‘,-’, ‘,++’, and ‘,--’ commands. When the frame number of the frame being examined is not zero, it is shown in the prompt after the nesting level, for example ‘1\5>’ is the prompt when the REPL nesting level is 1 and the frame number is 5.

Expressions entered at a nested REPL are evaluated in the environment (both lexical and dynamic) of the continuation frame currently being examined if that frame was created by interpreted Scheme code. If the frame was created by compiled Scheme code then expressions get evaluated in the global interaction environment. This feature may be used in interpreted code to fetch the value of a variable in the current frame or to change its value with set!. Note that some special forms (define in particular) can only be evaluated in the global interaction environment.

5.2 Debugging commands

In addition to expressions, the REPL accepts the following special “comma” commands:

,? and ,help

Give a summary of the REPL commands.

,(h subject)

This command will show the section of the Gambit manual with the definition of the procedure or special form subject, which must be a symbol. For example ‘,(h time)’ will show the section documenting the time special form. Please see the help procedure for additional information.

,h

This command will show the section of the Gambit manual with the definition of the procedure which raised the exception for which this REPL was started.

,q

Terminate the process with exit status 0. This is equivalent to calling (exit 0).

,qt

Terminate the current thread (note that terminating the primordial thread terminates the process).

,t

Return to the outermost REPL, also known as the “top-level REPL”.

,d

Leave the current REPL and resume the enclosing REPL. This command does nothing in the top-level REPL.

,(c expr)

Leave the current REPL and continue the computation that initiated the REPL with a specific value. This command can only be used to continue a computation that signaled an error. The expression expr is evaluated in the current context and the resulting value is returned as the value of the expression which signaled the error. For example, if the evaluation of the expression ‘(* (/ x y) 2)’ signaled an error because ‘y’ is zero, then in the nested REPL a ‘,(c (+ 4 y))’ will resume the computation of ‘(* (/ x y) 2)’ as though the value of ‘(/ x y)’ was 4. This command must be used carefully because the context where the error occured may rely on the result being of a particular type. For instance a ‘,(c #f)’ in the previous example will cause ‘*’ to signal a type error (this problem is the most troublesome when debugging Scheme code that was compiled with type checking turned off so be careful).

,c

Leave the current REPL and continue the computation that initiated the REPL. This command can only be used to continue a computation that was stopped due to a user interrupt, breakpoint or a single-step.

,s

Leave the current REPL and continue the computation that initiated the REPL in single-stepping mode. The computation will perform an evaluation step (as defined by step-level-set!) and then stop, causing a nested REPL to be entered. Just before the evaluation step is performed, a line is displayed (in the same format as trace) which indicates the expression that is being evaluated. If the evaluation step produces a result, the result is also displayed on another line. A nested REPL is then entered after displaying a message which describes the next step of the computation. This command can only be used to continue a computation that was stopped due to a user interrupt, breakpoint or a single-step.

,l

This command is similar to ‘,s’ except that it “leaps” over procedure calls, that is procedure calls are treated like a single step. Single-stepping mode will resume when the procedure call returns, or if and when the execution of the called procedure encounters a breakpoint.

,N

Move to frame number N of the continuation. After changing the current frame, a one-line summary of the frame is displayed as if the ‘,y’ command was entered.

,N+

Move forward by N frames in the chain of continuation frames (i.e. towards older continuation frames). After changing the current frame, a one-line summary of the frame is displayed as if the ‘,y’ command was entered.

,N-

Move backward by N frames in the chain of continuation frames (i.e. towards more recent continuation frames). After changing the current frame, a one-line summary of the frame is displayed as if the ‘,y’ command was entered.

,+

Equivalent to ‘,1+’.

,-

Equivalent to ‘,1-’.

,++

Equivalent to ‘,N+’ where N is the number of continuation frames displayed at the head of a backtrace.

,--

Equivalent to ‘,N-’ where N is the number of continuation frames displayed at the head of a backtrace.

,y

Display a one-line summary of the current frame. The information is displayed in four fields. The first field is the frame number. The second field is the procedure that created the frame or ‘(interaction)’ if the frame was created by an expression entered at the REPL. The remaining fields describe the subproblem associated with the frame, that is the expression whose value is being computed. The third field is the location of the subproblem’s source code and the fourth field is a reproduction of the source code, possibly truncated to fit on the line. The last two fields may be missing if that information is not available. In particular, the third field is missing when the frame was created by a user call to the ‘eval’ procedure or by a compiled procedure not compiled with the declaration ‘debug-location’, and the last field is missing when the frame was created by a compiled procedure not compiled with the declaration ‘debug-source’.

,b

Display a backtrace summarizing each frame in the chain of continuation frames starting with the current frame. For each frame, the same information as for the ‘,y’ command is displayed (except that location information is displayed in the format ‘stream@line:column’). If there are more than 15 frames in the chain of continuation frames, some of the middle frames will be omitted.

,be

Like the ‘,b’ command but also display the environment.

,bed

Like the ‘,be’ command but also display the dynamic environment.

,(b expr)

Display the backtrace of expr’s value, X, which is obtained by evaluating expr in the current frame. X must be a continuation or a thread. When X is a continuation, the frames in that continuation are displayed. When X is a thread, the backtrace of the current continuation of that thread is displayed.

,(be expr)

Like the ‘,(b expr)’ command but also display the environment.

,(bed expr)

Like the ‘,(be expr)’ command but also display the dynamic environment.

,i

Pretty print the procedure that created the current frame or ‘(interaction)’ if the frame was created by an expression entered at the REPL. Compiled procedures will only be pretty printed when they are compiled with the declaration ‘debug-source’.

,e

Display the environment which is accessible from the current frame. The lexical environment is displayed, followed by the dynamic environment if the parameter object repl-display-dynamic-environment? is not false. Global lexical variables are not displayed. Moreover the frame must have been created by interpreted code or code compiled with the declaration ‘debug-environments’. Due to space safety considerations and compiler optimizations, some of the lexical variable bindings may be missing. Lexical variable bindings are displayed using the format ‘variable = expression’ (when variable is mutable) or ‘variable == expression’ (when variable is immutable, which may happen in compiled code due to compiler optimization) and dynamically-bound parameter bindings are displayed using the format ‘(parameter) = expression’. Note that expression can be a self-evaluating expression (number, string, boolean, character, ...), a quoted expression, a lambda expression or a global variable (the last two cases, which are only used when the value of the variable or parameter is a procedure, simplifies the debugging of higher-order procedures). A parameter can be a quoted expression or a global variable. Lexical bindings are displayed in inverse binding order (most deeply nested first) and shadowed variables are included in the list.

,ed

Like the ‘,e’ command but the dynamic environment is always displayed.

,(e expr)

Display the environment of expr’s value, X, which is obtained by evaluating expr in the current frame. X must be a continuation, a thread, a procedure, or a nonnegative integer. When X is a continuation, the environment at that point in the code is displayed. When X is a thread, the environment of the current continuation of that thread is displayed. When X is a procedure, the lexical environment where X was created is combined with the current continuation and this combined environment is displayed. When X is an integer, the environment at frame number X of the continuation is displayed.

,(ed expr)

Like the ‘,(e expr)’ command but the dynamic environment is always displayed.

,st

Display the state of the threads in the current thread’s thread group. A thread can be: uninitialized, initialized, active, and terminated (normally or abnormally). Active threads can be running, sleeping and waiting on a synchronization object (mutex, condition variable or port) possibly with a timeout.

,(st expr)

Display the state of a specific thread or thread group. The value of expr must be a thread or thread group.

,(v expr)

Start a new REPL visiting expr’s value, X, which is obtained by evaluating expr in the current frame. X must be a continuation, a thread, a procedure, or a nonnegative integer. When X is a continuation, the new REPL’s continuation is X and evaluations are done in the environment at that point in the code. When X is a thread, the thread is interrupted and the new REPL’s continuation is the point where the thread was interrupted. When X is a procedure, the lexical environment where X was created is combined with the current continuation and evaluations are done in this combined environment. When X is an integer, the REPL is started in frame number X of the continuation.

5.3 Debugging example

Here is a sample interaction with gsi:

$ gsi
Gambit v4.9.4

> (define (invsqr x) (/ 1 (expt x 2)))
> (define (mymap fn lst)
    (define (mm in)
      (if (null? in)
          '()
          (cons (fn (car in)) (mm (cdr in)))))
    (mm lst))
> (mymap invsqr '(5 2 hello 9 1))
*** ERROR IN invsqr, (console)@1.25 -- (Argument 1) NUMBER expected
(expt 'hello 2)
1> ,i
#<procedure #2 invsqr> =
(lambda (x) (/ 1 (expt x 2)))
1> ,e
x = 'hello
1> ,b
0  invsqr                    (console)@1:25          (expt x 2)
1  #<procedure #4>           (console)@6:17          (fn (car in))
2  #<procedure #4>           (console)@6:31          (mm (cdr in))
3  #<procedure #4>           (console)@6:31          (mm (cdr in))
4  (interaction)             (console)@8:1           (mymap invsqr '(5 2 hel...
1> ,+
1  #<procedure #4>           (console)@6.17          (fn (car in))
1\1> (pp #4)
(lambda (in) (if (null? in) '() (cons (fn (car in)) (mm (cdr in)))))
1\1> ,e
in = '(hello 9 1)
mm = (lambda (in) (if (null? in) '() (cons (fn (car in)) (mm (cdr in)))))
fn = invsqr
lst = '(5 2 hello 9 1)
1\1> ,(e mm)
mm = (lambda (in) (if (null? in) '() (cons (fn (car in)) (mm (cdr in)))))
fn = invsqr
lst = '(5 2 hello 9 1)
1\1> fn
#<procedure #2 invsqr>
1\1> (pp fn)
(lambda (x) (/ 1 (expt x 2)))
1\1> ,+
2  #<procedure #4>           (console)@6.31          (mm (cdr in))
1\2> ,e
in = '(2 hello 9 1)
mm = (lambda (in) (if (null? in) '() (cons (fn (car in)) (mm (cdr in)))))
fn = invsqr
lst = '(5 2 hello 9 1)
1\2> ,(c (list 3 4 5))
(1/25 1/4 3 4 5)
> ,q

5.4 Procedures related to debugging

The help procedure displays the section of the Gambit manual with the definition of the procedure or special form subject, which must be a procedure or symbol. For example the call (help gensym) will show the section documenting the gensym procedure and the call (help 'time) will show the section documenting the time special form. When the subject is absent, the documentation of the help procedure is shown. The help procedure returns the void object.

The parameter object help-browser is bound to a string naming the external program that is used by the help procedure to view the documentation. Initially it is bound to the empty string. In normal circumstances when help-browser is bound to an empty string the help procedure runs the script ~~bin/gambdoc.bat which searches for a suitable web browser to open the documentation in HTML format. Unless the system was built with the command ‘configure --enable-help-browser=...’, the text-only browser ‘lynx’ (see http://lynx.isc.org/) will be used by default if it is available. We highly recommend that you install this browser if you are interested in viewing the documentation within the console in which the REPL is running. You can exit ‘lynx’ conveniently by typing an end of file (usually <^D>).

For example:

> (help-browser "firefox") ; use firefox instead of lynx
> (help 'gensym)
> (help gensym) ; OK because gensym is a procedure
> (help 'time)
> (help time) ; not OK because time is a special form
*** ERROR IN (console)@5.7 -- Macro name can't be used as a variable: time
>

The apropos procedure writes to the port port a report of all the global variables whose name contains substring, a string or symbol. If substring is not specified the report contains all the global variables. If it is not specified, port defaults to the interaction channel (i.e. the output will appear at the REPL). The apropos procedure returns the void object.

The global variables are grouped into namespaces. The empty namespace, if it is relevant, is last. This reduces the likelihood it will scroll off the screen if there are several global variables in other namespaces, which are typically less interesting.

Note that with the apropos procedure it is possible to reveal the existence of procedures of the runtime system and modules that are not intended to be called by user code. These procedures often avoid type checking their arguments or must be called in a specific context, so calling them incorrectly may crash the system. On the other hand it also allows discovering the existence of certain functionalities that may have gone unnoticed.

For example:

> (apropos "cons")
"##" namespace:
  10^-constants, cons, cons*, cons*-aux, console-port,
  constant-expression-value, constant-expression?,
  cprc-quasi-cons, deconstruct-call,
  define-type-construct-constant, degen-quasi-cons,
  gen-quasi-cons, quasi-cons, stdio/console-repl-channel,
  void-constant?, xcons
empty namespace:
  cons, cons*, console-port, xcons
> (import (srfi 69))
> (apropos "table?")
"##" namespace:
  gc-hash-table?, mutable?, readtable?, table?
"srfi/69#" namespace:
  hash-table?
empty namespace:
  readtable?, table?
> (apropos "srfi/69#")
"srfi/69#" namespace:
  ||, alist->hash-table, hash, hash-by-identity,
  hash-table->alist, hash-table-copy, hash-table-delete!,
  hash-table-equivalence-function, hash-table-exists?,
  hash-table-fold, hash-table-hash-function,
  hash-table-keys, hash-table-merge!, hash-table-ref,
  hash-table-ref/default, hash-table-set!, hash-table-size,
  hash-table-update!, hash-table-update!/default,
  hash-table-values, hash-table-walk, hash-table?,
  make-hash-table, string-ci-hash, string-hash

The REPL keeps a history of the last few results printed by the REPL. The call (repl-result-history-ref i) returns the ith previous result (the last for i=0, the next to last for i=1, etc). By default the REPL result history remembers up to 3 results. The maximal length of the history can be set to n between 0 and 10 by a call to (repl-result-history-max-length-set! n).

For convenience the reader defines an abbreviation for calling repl-result-history-ref. Tokens formed by a sequence of one or more hash signs, such as ‘#’, ‘##’, etc, are expanded by the reader into the list (repl-result-history-ref i), where i is the number of hash signs minus 1. In other words, ‘#’ will return the last result printed by the REPL, ‘##’ will return the next to last, etc.

For example:

> (map (lambda (x) (* x x)) '(1 2 3))
(1 4 9)
> (reverse #)
(9 4 1)
> (append # ##)
(9 4 1 1 4 9)
> 1
1
> 1
1
> (+ # ##)
2
> (+ # ##)
3
> (+ # ##)
5
> ####
*** ERROR IN (console)@9.1 -- (Argument 1) Out of range
(repl-result-history-ref 3)
1>

The trace procedure starts tracing calls to the specified procedures. When a traced procedure is called, a line containing the procedure and its arguments is displayed (using the procedure call expression syntax). The line is indented with a sequence of vertical bars which indicate the nesting depth of the procedure’s continuation. After the vertical bars is a greater-than sign which indicates that the evaluation of the call is starting.

When a traced procedure returns a result, it is displayed with the same indentation as the call but without the greater-than sign. This makes it easy to match calls and results (the result of a given call is the value at the same indentation as the greater-than sign). If a traced procedure P1 performs a tail call to a traced procedure P2, then P2 will use the same indentation as P1. This makes it easy to spot tail calls. The special handling for tail calls is needed to preserve the space complexity of the program (i.e. tail calls are implemented as required by Scheme even when they involve traced procedures).

The untrace procedure stops tracing calls to the specified procedures. When no argument is passed to the trace procedure, the list of procedures currently being traced is returned. The void object is returned by the trace procedure when it is passed one or more arguments. When no argument is passed to the untrace procedure stops all tracing and returns the void object. A compiled procedure may be traced but only if it is bound to a global variable.

For example:

> (define (fact n) (if (< n 2) 1 (* n (fact (- n 1)))))
> (trace fact)
> (fact 5)
| > (fact 5)
| | > (fact 4)
| | | > (fact 3)
| | | | > (fact 2)
| | | | | > (fact 1)
| | | | | 1
| | | | 2
| | | 6
| | 24
| 120
120
> (trace -)
*** WARNING -- Rebinding global variable "-" to an interpreted procedure
> (define (fact-iter n r) (if (< n 2) r (fact-iter (- n 1) (* n r))))
> (trace fact-iter)
> (fact-iter 5 1)
| > (fact-iter 5 1)
| | > (- 5 1)
| | 4
| > (fact-iter 4 5)
| | > (- 4 1)
| | 3
| > (fact-iter 3 20)
| | > (- 3 1)
| | 2
| > (fact-iter 2 60)
| | > (- 2 1)
| | 1
| > (fact-iter 1 120)
| 120
120
> (trace)
(#<procedure #2 fact-iter> #<procedure #3 -> #<procedure #4 fact>)
> (untrace)
> (fact 5)
120

The step procedure enables single-stepping mode. After the call to step the computation will stop just before the interpreter executes the next evaluation step (as defined by step-level-set!). A nested REPL is then started. Note that because single-stepping is stopped by the REPL whenever the prompt is displayed it is pointless to enter (step) by itself. On the other hand entering (begin (step) expr) will evaluate expr in single-stepping mode.

The procedure step-level-set! sets the stepping level which determines the granularity of the evaluation steps when single-stepping is enabled. The stepping level level must be an exact integer in the range 0 to 7. At a level of 0, the interpreter ignores single-stepping mode. At higher levels the interpreter stops the computation just before it performs the following operations, depending on the stepping level:

1. procedure call
2. delay special form and operations at lower levels
3. lambda special form and operations at lower levels
4. define special form and operations at lower levels
5. set! special form and operations at lower levels
6. variable reference and operations at lower levels
7. constant reference and operations at lower levels

The default stepping level is 7.

For example:

> (define (fact n) (if (< n 2) 1 (* n (fact (- n 1)))))
> (step-level-set! 1)
> (begin (step) (fact 5))
*** STOPPED IN (console)@3.15
1> ,s
| > (fact 5)
*** STOPPED IN fact, (console)@1.22
1> ,s
| | > (< n 2)
| | #f
*** STOPPED IN fact, (console)@1.43
1> ,s
| | > (- n 1)
| | 4
*** STOPPED IN fact, (console)@1.37
1> ,s
| | > (fact (- n 1))
*** STOPPED IN fact, (console)@1.22
1> ,s
| | | > (< n 2)
| | | #f
*** STOPPED IN fact, (console)@1.43
1> ,s
| | | > (- n 1)
| | | 3
*** STOPPED IN fact, (console)@1.37
1> ,l
| | | > (fact (- n 1))
*** STOPPED IN fact, (console)@1.22
1> ,l
| | > (* n (fact (- n 1)))
| | 24
*** STOPPED IN fact, (console)@1.32
1> ,l
| > (* n (fact (- n 1)))
| 120
120

The break procedure places a breakpoint on each of the specified procedures. When a procedure is called that has a breakpoint, the interpreter will enable single-stepping mode (as if step had been called). This typically causes the computation to stop soon inside the procedure if the stepping level is high enough.

The unbreak procedure removes the breakpoints on the specified procedures. With no argument, break returns the list of procedures currently containing breakpoints. The void object is returned by break if it is passed one or more arguments. With no argument unbreak removes all the breakpoints and returns the void object. A breakpoint can be placed on a compiled procedure but only if it is bound to a global variable.

For example:

> (define (double x) (+ x x))
> (define (triple y) (- (double (double y)) y))
> (define (f z) (* (triple z) 10))
> (break double)
> (break -)
*** WARNING -- Rebinding global variable "-" to an interpreted procedure
> (f 5)
*** STOPPED IN double, (console)@1.21
1> ,b
0  double                    (console)@1:21          +
1  triple                    (console)@2:31          (double y)
2  f                         (console)@3:18          (triple z)
3  (interaction)             (console)@6:1           (f 5)
1> ,e
x = 5
1> ,c
*** STOPPED IN double, (console)@1.21
1> ,c
*** STOPPED IN f, (console)@3.29
1> ,c
150
> (break)
(#<procedure #3 -> #<procedure #4 double>)
> (unbreak)
> (f 5)
150

[Note: this procedure is DEPRECATED and will be removed in a future version of Gambit. Use the ‘proper-tail-calls’ declaration instead.]

The parameter object generate-proper-tail-calls is bound to a boolean value controlling how the interpreter handles tail calls. When it is bound to #f the interpreter will treat tail calls like nontail calls, that is a new continuation will be created for the call. This setting is useful for debugging, because when a primitive signals an error the location information will point to the call site of the primitive even if this primitive was called with a tail call. The initial value of this parameter object is #t, which means that a tail call will reuse the continuation of the calling function.

This parameter object only affects code that is subsequently processed by load or eval, or entered at the REPL.

For example:

> (generate-proper-tail-calls)
#t
> (let loop ((i 1)) (if (< i 10) (loop (* i 2)) oops))
*** ERROR IN #<procedure #2>, (console)@2.47 -- Unbound variable: oops
1> ,b
0  #<procedure #2>           (console)@2:47          oops
1  (interaction)             (console)@2:1           ((letrec ((loop (lambda...
1> ,t
> (generate-proper-tail-calls #f)
> (let loop ((i 1)) (if (< i 10) (loop (* i 2)) oops))
*** ERROR IN #<procedure #3>, (console)@6.47 -- Unbound variable: oops
1> ,b
0  #<procedure #3>           (console)@6:47          oops
1  #<procedure #3>           (console)@6:32          (loop (* i 2))
2  #<procedure #3>           (console)@6:32          (loop (* i 2))
3  #<procedure #3>           (console)@6:32          (loop (* i 2))
4  #<procedure #3>           (console)@6:32          (loop (* i 2))
5  (interaction)             (console)@6:1           ((letrec ((loop (lambda...

[Note: this procedure is DEPRECATED and will be removed in a future version of Gambit. Use the parameter object repl-display-environment? instead.]

This procedure sets a flag that controls the automatic display of the environment by the REPL. If display? is true, the environment is displayed by the REPL before the prompt. The default setting is not to display the environment.

The parameter object repl-display-environment? is bound to a boolean value that controls the automatic display of the environment by the REPL. If display? is true, the environment is displayed by the REPL before the prompt. This is particularly useful in single-stepping mode. The default setting is not to display the environment.

The parameter object display-dynamic-environment? is bound to a boolean value that controls wether the dynamic environment is displayed when the environment is displayed. The default setting is not to display the dynamic environment.

This procedure pretty-prints obj on the port port. If it is not specified, port defaults to the current output-port.

For example:

> (pretty-print
    (let* ((x '(1 2 3 4)) (y (list x x x))) (list y y y)))
(((1 2 3 4) (1 2 3 4) (1 2 3 4))
 ((1 2 3 4) (1 2 3 4) (1 2 3 4))
 ((1 2 3 4) (1 2 3 4) (1 2 3 4)))

This procedure pretty-prints obj on the port port. When obj is a procedure created by the interpreter or a procedure created by code compiled with the declaration ‘debug-source’, the procedure’s source code is displayed. If it is not specified, port defaults to the interaction channel (i.e. the output will appear at the REPL).

For example:

> (define (f g) (+ (time (g 100)) (time (g 1000))))
> (pp f)
(lambda (g)
  (+ (##time (lambda () (g 100)) '(g 100))
     (##time (lambda () (g 1000)) '(g 1000))))

This procedure controls the generation of reports during garbage collections. If the argument is true, a brief report of memory usage is generated after every garbage collection. It contains: the time taken for this garbage collection, the amount of memory allocated in megabytes since the program was started, the size of the heap in megabytes, the heap memory in megabytes occupied by live data, the proportion of the heap occupied by live data, and the number of bytes occupied by movable and nonmovable objects.

5.5 Console line-editing

The console implements a simple Scheme-friendly line-editing user-interface that is enabled by default. It offers parentheses balancing, a history of previous commands, symbol completion, and several emacs-compatible keyboard commands. The user’s input is displayed in a bold font and the output produced by the system is in a plain font. The history of previous commands is saved in the file ‘~/.gambit_history’. It is restored when a REPL is started.

Symbol completion is triggered with the tab key. When the cursor is after a sequence of characters that can form a symbol, typing the tab key will search the symbol table for the first symbol (in alphabetical order) that begins with that sequence and insert that symbol. Typing the tab key in succession will cycle through all symbols with that prefix. When all possible symbols have been shown or there are no possible completions, the text reverts to the uncompleted symbol and the bell is rung.

Here are the keyboard commands available (where the ‘M-’ prefix means the escape key is typed and the ‘C-’ prefix means the control key is pressed):

C-d

Generate an end-of-file when the line is empty, otherwise delete character at cursor.

delete or backspace

Delete character before cursor.

M-C-d

Delete word forward and keep a copy of this text on the clipboard.

M-delete

Delete word backward and keep a copy of this text on the clipboard.

M-backspace

Delete S-expression backward and keep a copy of this text on the clipboard.

C-a

Move cursor to beginning of line.

C-e

Move cursor to end of line.

C-b or left-arrow

Move cursor left one character.

M-b

Move cursor left one word.

M-C-b or M-left-arrow

Move cursor left one S-expression.

C-f or right-arrow

Move cursor right one character.

M-f

Move cursor right one word.

M-C-f or M-right-arrow

Move cursor right one S-expression.

C-p or M-p or up-arrow

Move to previous line in history.

C-n or M-n or down-arrow

Move to next line in history.

C-t

Transpose character at cursor with previous character.

M-t

Transpose word after cursor with previous word.

M-C-t

Transpose S-expression after cursor with previous S-expression.

C-l

Clear console and redraw line being edited.

C-nul

Set the mark to the cursor.

C-w

Delete the text between the cursor and the mark and keep a copy of this text on the clipboard.

C-k

Delete the text from the cursor to the end of the line and keep a copy of this text on the clipboard.

C-y

Paste the text that is on the clipboard.

F8

Same as typing ‘#||#,c;’ (REPL command to continue the computation).

F9

Same as typing ‘#||#,-;’ (REPL command to move to newer frame).

F10

Same as typing ‘#||#,+;’ (REPL command to move to older frame).

F11

Same as typing ‘#||#,s;’ (REPL command to step the computation).

F12

Same as typing ‘#||#,l;’ (REPL command to leap the computation).

On macOS, depending on your configuration, you may have to press the fn key to access the function key F12 and the option key to access the other function keys.

On Microsoft Windows the clipboard is the system clipboard. This allows text to be copied and pasted between the program and other applications. On other operating systems the clipboard is internal to the program (it is not integrated with the operating system).

5.6 Emacs interface

Gambit comes with the Emacs package ‘gambit.el’ which provides a nice environment for running Gambit from within the Emacs editor. This package filters the standard output of the Gambit process and when it intercepts a location information (in the format ‘stream@line.column’ where stream is either ‘(stdin)’ when the expression was obtained from standard input, ‘(console)’ when the expression was obtained from the console, or a string naming a file) it opens a window to highlight the corresponding expression.

To use this package, make sure the file ‘gambit.el’ is accessible from your load-path and that the following lines are in your ‘.emacs’ file:

(autoload 'gambit-inferior-mode "gambit" "Hook Gambit mode into cmuscheme.")
(autoload 'gambit-mode "gambit" "Hook Gambit mode into scheme.")
(add-hook 'inferior-scheme-mode-hook (function gambit-inferior-mode))
(add-hook 'scheme-mode-hook (function gambit-mode))
(setq scheme-program-name "gsi -:debug=-")

Alternatively, if you don’t mind always loading this package, you can simply add this line to your ‘.emacs’ file:

(require 'gambit)

You can then start an inferior Gambit process by typing ‘M-x run-scheme’. The commands provided in ‘cmuscheme’ mode will be available in the Gambit interaction buffer (i.e. ‘*scheme*’) and in buffers attached to Scheme source files. Here is a list of the most useful commands (for a complete list type ‘C-h m’ in the Gambit interaction buffer):

C-x C-e

Evaluate the expression which is before the cursor (the expression will be copied to the Gambit interaction buffer).

C-c C-z

Switch to Gambit interaction buffer.

C-c C-l

Load a file (file attached to current buffer is default) using (load file).

C-c C-k

Compile a file (file attached to current buffer is default) using (compile-file file).

The file ‘gambit.el’ provides these additional commands:

F8 or C-c c

Continue the computation (same as typing ‘#||#,c;’ to the REPL).

F9 or C-c ]

Move to newer frame (same as typing ‘#||#,-;’ to the REPL).

F10 or C-c [

Move to older frame (same as typing ‘#||#,+;’ to the REPL).

F11 or C-c s

Step the computation (same as typing ‘#||#,s;’ to the REPL).

F12 or C-c l

Leap the computation (same as typing ‘#||#,l;’ to the REPL).

C-c _

Removes the last window that was opened to highlight an expression.

The two keystroke version of these commands can be shortened to ‘M-c’, ‘M-[’, ‘M-]’, ‘M-s’, ‘M-l’, and ‘M-_’ respectively by adding this line to your ‘.emacs’ file:

(setq gambit-repl-command-prefix "\e")

This is more convenient to type than the two keystroke ‘C-c’ based sequences but the purist may not like this because it does not follow normal Emacs conventions.

Here is what a typical ‘.emacs’ file will look like:

(setq load-path ; add directory containing gambit.el
  (cons "/usr/local/Gambit/share/emacs/site-lisp"
        load-path))
(setq scheme-program-name "/tmp/gsi -:debug=-") ; if gsi not in executable path
(setq gambit-highlight-color "gray") ; if you don't like the default
(setq gambit-repl-command-prefix "\e") ; if you want M-c, M-s, etc
(require 'gambit)

5.7 GUIDE

The implementation and documentation for GUIDE, the Gambit Universal IDE, are not yet complete.

6. Scheme extensions

6.1 Extensions to standard procedures

These procedures do nothing.

The procedure call-with-current-continuation is bound to the global variables call-with-current-continuation and call/cc.

6.2 Extensions to standard special forms

• lambda-formals = ( formal-argument-list ) | r4rs-lambda-formals
• define-formals = formal-argument-list | r4rs-define-formals
• formal-argument-list = dsssl-formal-argument-list | rest-at-end-formal-argument-list
• dsssl-formal-argument-list = reqs opts rest keys
• rest-at-end-formal-argument-list = reqs opts keys rest | reqs opts keys . rest-formal-argument
• reqs = required-formal-argument*
• required-formal-argument = variable
• opts = #!optional optional-formal-argument* | empty
• optional-formal-argument = variable | ( variable initializer )
• rest = #!rest rest-formal-argument | empty
• rest-formal-argument = variable
• keys = #!key keyword-formal-argument* | empty
• keyword-formal-argument = variable | ( variable initializer )
• initializer = expression
• r4rs-lambda-formals = ( variable* ) | ( variable+ . variable ) | variable
• r4rs-define-formals = variable* | variable* . variable

These forms are extended versions of the lambda and define special forms of standard Scheme. They allow the use of optional formal arguments, either positional or named, and support the syntax and semantics of the DSSSL standard.

When the procedure introduced by a lambda (or define) is applied to a list of actual arguments, the formal and actual arguments are processed as specified in the R4RS if the lambda-formals (or define-formals) is a r4rs-lambda-formals (or r4rs-define-formals).

If the formal-argument-list matches dsssl-formal-argument-list or extended-formal-argument-list they are processed as follows:

1. Variables in required-formal-arguments are bound to successive actual arguments starting with the first actual argument. It shall be an error if there are fewer actual arguments than required-formal-arguments.
2. Next variables in optional-formal-arguments are bound to remaining actual arguments. If there are fewer remaining actual arguments than optional-formal-arguments, then the variables are bound to the result of evaluating initializer, if one was specified, and otherwise to #f. The initializer is evaluated in an environment in which all previous formal arguments have been bound.
3. If #!key does not appear in the formal-argument-list and there is no rest-formal-argument then it shall be an error if there are any remaining actual arguments.
4. If #!key does not appear in the formal-argument-list and there is a rest-formal-argument then the rest-formal-argument is bound to a list of all remaining actual arguments.
5. If #!key appears in the formal-argument-list and there is no rest-formal-argument then there shall be an even number of remaining actual arguments. These are interpreted as a series of pairs, where the first member of each pair is a keyword specifying the argument name, and the second is the corresponding value. It shall be an error if the first member of a pair is not a keyword. It shall be an error if the argument name is not the same as a variable in a keyword-formal-argument. If the same argument name occurs more than once in the list of actual arguments, then the first value is used. If there is no actual argument for a particular keyword-formal-argument, then the variable is bound to the result of evaluating initializer if one was specified, and otherwise to #f. The initializer is evaluated in an environment in which all previous formal arguments have been bound.
6. If #!key appears in the formal-argument-list and there is a rest-formal-argument before the #!key then there may be an even or odd number of remaining actual arguments and the rest-formal-argument is bound to a list of all remaining actual arguments. Then, these remaining actual arguments are scanned from left to right in pairs, stopping at the first pair whose first element is not a keyword. Each pair whose first element is a keyword matching the name of a keyword-formal-argument gives the value (i.e. the second element of the pair) of the corresponding formal argument. If the same argument name occurs more than once in the list of actual arguments, then the first value is used. If there is no actual argument for a particular keyword-formal-argument, then the variable is bound to the result of evaluating initializer if one was specified, and otherwise to #f. The initializer is evaluated in an environment in which all previous formal arguments have been bound.
7. If #!key appears in the formal-argument-list and there is a rest-formal-argument after the #!key then there may be an even or odd number of remaining actual arguments. The remaining actual arguments are scanned from left to right in pairs, stopping at the first pair whose first element is not a keyword. Each pair shall have as its first element a keyword matching the name of a keyword-formal-argument; the second element gives the value of the corresponding formal argument. If the same argument name occurs more than once in the list of actual arguments, then the first value is used. If there is no actual argument for a particular keyword-formal-argument, then the variable is bound to the result of evaluating initializer if one was specified, and otherwise to #f. The initializer is evaluated in an environment in which all previous formal arguments have been bound. Finally, the rest-formal-argument is bound to the list of the actual arguments that were not scanned (i.e. after the last keyword/value pair).

In all cases it is an error for a variable to appear more than once in a formal-argument-list.

Note that this specification is compatible with the DSSSL language standard (i.e. a correct DSSSL program will have the same semantics when run with Gambit).

It is unspecified whether variables receive their value by binding or by assignment. Currently the compiler and interpreter use different methods, which can lead to different semantics if call-with-current-continuation is used in an initializer. Note that this is irrelevant for DSSSL programs because call-with-current-continuation does not exist in DSSSL.

For example:

> ((lambda (#!rest x) x) 1 2 3)
(1 2 3)
> (define (f a #!optional b) (list a b))
> (define (g a #!optional (b a) #!key (k (* a b))) (list a b k))
> (define (h1 a #!rest r #!key k) (list a k r))
> (define (h2 a #!key k #!rest r) (list a k r))
> (f 1)
(1 #f)
> (f 1 2)
(1 2)
> (g 3)
(3 3 9)
> (g 3 4)
(3 4 12)
> (g 3 4 k: 5)
(3 4 5)
> (g 3 4 k: 5 k: 6)
(3 4 5)
> (h1 7)
(7 #f ())
> (h1 7 k: 8 9)
(7 8 (k: 8 9))
> (h1 7 k: 8 z: 9)
(7 8 (k: 8 z: 9))
> (h2 7)
(7 #f ())
> (h2 7 k: 8 9)
(7 8 (9))
> (h2 7 k: 8 z: 9)
*** ERROR IN (console)@17.1 -- Unknown keyword argument passed to procedure
(h2 7 k: 8 z: 9)

6.3 Miscellaneous extensions

This procedure is the vector analog of the substring procedure. It returns a newly allocated vector formed from the elements of the vector vector beginning with index start (inclusive) and ending with index end (exclusive).

For example:

> (subvector '#(a b c d e f) 3 5)
#(d e)

This procedure is like the procedure subvector except the parameter start defaults to 0 and the parameter end defaults to the length of the vector vector. Note that the elements are not recursively copied.

For example:

> (define v1 '#(a b c d e f))
> (define v2 (vector-copy v1))
> v2
#(a b c d e f)
> (eq? v1 v2)
#f
> (vector-copy v1 3)
#(d e f)
> (vector-copy v1 3 5)
#(d e)

This procedure mutates the vector dest-vector. It copies the elements of the vector vector beginning with index start (inclusive) and ending with index end (exclusive) to the vector dest-vector at index dest-start. The parameters start and end default respectively to 0 and the length of the vector vector. It is an error to copy more elements than will fit in the tail of the vector dest-vector starting at index dest-start. Note that the elements are not recursively copied.

For example:

> (define v1 (vector 10 11 12 13 14 15))
> (define v2 (vector 20 21 22 23))
> (vector-copy! v1 1 v2)
> v1
#(10 20 21 22 23 15)
> (vector-copy! v1 1 v2 3)
> v1
#(10 23 21 22 23 15)
> (vector-copy! v1 1 v2 1 3)
> v1
#(10 21 22 22 23 15)

This procedure is the vector analog of the string-append procedure. It returns a newly allocated vector whose elements form the concatenation of the given vectors.

For example:

> (define v '#(1 2 3))
> (vector-append v v v)
#(1 2 3 1 2 3 1 2 3)

This procedure returns a newly allocated vector whose elements form the concatenation of all the vectors in the list lst. If the optional vector separator argument is specified, it will be added between all the elements of lst. Without the separator argument the result is the same as (apply vector-append lst).

For example:

> (define v '#(1 2 3))
> (vector-concatenate (list v v v))
#(1 2 3 1 2 3 1 2 3)
> (vector-concatenate (list v v v) '#(88 99))
#(1 2 3 88 99 1 2 3 88 99 1 2 3)

This procedure is like vector-fill!, but fills a selected part of the given vector. It sets the elements of the vector vector, beginning with index start (inclusive) and ending with index end (exclusive) to fill. The value returned is unspecified.

For example:

> (define v (vector 'a 'b 'c 'd 'e 'f))
> (subvector-fill! v 3 5 'x)
> v
#(a b c x x f)

This procedure replaces part of the contents of vector dst-vector with part of the contents of vector src-vector. It copies elements from src-vector, beginning with index src-start (inclusive) and ending with index src-end (exclusive) to dst-vector beginning with index dst-start (inclusive). The value returned is unspecified.

For example:

> (define v1 '#(1 2 3 4 5 6))
> (define v2 (vector 'a 'b 'c 'd 'e 'f))
> (subvector-move! v1 3 5 v2 1)
> v2
#(a 4 5 d e f)

This procedure shortens the vector vector so that its new size is k. The value returned is unspecified.

For example:

> (define v (vector 'a 'b 'c 'd 'e 'f))
> v
#(a b c d e f)
> (vector-shrink! v 3)
> v
#(a b c)

The procedure vector-cas! performs an atomic compare-and-swap operation on the element of vector vector at index k. If the element’s value is eq? to old-value then the element is changed to new-value, otherwise the value does not change. Regardless what happened, the element’s value prior to any change is returned. It is thus possible to detect a change by an explicit eq? test of the result.

For example:

> (define v (vector 'a))
> (eq? 'foo (vector-cas! v 0 'b 'foo))
#f
> v
#(a)
> (eq? 'a (vector-cas! v 0 'b 'a))
#t
> v
#(b)

The procedure vector-inc! performs an atomic incrementation on the element of vector vector at index k, which must be a fixnum. The parameter step defaults to 1 and it is the fixnum value that is added (with wraparound) to the element. The procedure returns the value of the element prior to the incrementation.

For example:

> (define v (vector 100))
> (vector-inc! v 0)
100
> (vector-inc! v 0)
101
> (vector-inc! v 0 5)
102
> v
#(107)

The procedure vector-set returns a new copy of the vector vector with the element at index k replaced with obj.

For example:

> (define v1 (vector 10 11 12 13))
> (define v2 (vector-set v1 2 99))
> v2
#(10 11 99 13)
> (eq? v1 v2)
#f

The procedure string-set returns a new copy of the string string with the character at index k replaced with char.

For example:

> (define s1 (string #\a #\b #\c #\d))
> (define s2 (string-set s1 2 #\.))
> s2
"ab.d"
> (eq? s1 s2)
#f

This procedure returns a newly allocated string which is the concatenation of all the strings in the list lst. If the optional string separator argument is specified, it will be added between all the elements of lst. Without the separator argument the result is the same as (apply string-append lst).

For example:

> (define s "abc")
> (string-concatenate (list s s s))
"abcabcabc"
> (string-concatenate (list s s s) ",")
"abc,abc,abc"

This procedure is like string-fill!, but fills a selected part of the given string. It sets the elements of the string string, beginning with index start (inclusive) and ending with index end (exclusive) to fill. The value returned is unspecified.

For example:

> (define s (string #\a #\b #\c #\d #\e #\f))
> (substring-fill! s 3 5 #\x)
> s
"abcxxf"

This procedure replaces part of the contents of string dst-string with part of the contents of string src-string. It copies elements from src-string, beginning with index src-start (inclusive) and ending with index src-end (exclusive) to dst-string beginning with index dst-start (inclusive). The value returned is unspecified.

For example:

> (define s1 "123456")
> (define s2 (string #\a #\b #\c #\d #\e #\f))
> (substring-move! s1 3 5 s2 1)
> s2
"a45def"

This procedure shortens the string string so that its new size is k. The value returned is unspecified.

For example:

> (define s (string #\a #\b #\c #\d #\e #\f))
> s
"abcdef"
> (string-shrink! s 3)
> s
"abc"

These procedures implement the box data type. A box is a cell containing a single mutable field. The lexical syntax of a box containing the object obj is #&obj (see section Box syntax).

The procedure box returns a new box object whose content is initialized to obj. The procedure box? returns #t if obj is a box, and otherwise returns #f. The procedure unbox returns the content of the box box. The procedure set-box! changes the content of the box box to obj. The procedure set-box! returns an unspecified value.

For example:

> (define b (box 0))
> b
#&0
> (define (inc!) (set-box! b (+ (unbox b) 1)))
> (inc!)
> b
#&1
> (unbox b)
1

These procedures implement the keyword data type. Keywords are similar to symbols but are self evaluating and distinct from the symbol data type. The lexical syntax of keywords is specified in Keyword syntax.

The procedure keyword? returns #t if obj is a keyword, and otherwise returns #f. The procedure keyword->string returns the name of keyword as a string. The procedure string->keyword returns the keyword whose name is string.

For example:

> (keyword? 'color)
#f
> (keyword? color:)
#t
> (keyword->string color:)
"color"
> (string->keyword "color")
color:

This procedure returns a new uninterned symbol. Uninterned symbols are guaranteed to be distinct from the symbols generated by the procedures read and string->symbol. The symbol prefix is the prefix used to generate the new symbol’s name. If it is not specified, the prefix defaults to ‘g’.

For example:

> (gensym)
#:g0
> (gensym)
#:g1
> (gensym 'star-trek-)
#:star-trek-2

The procedure string->uninterned-symbol returns a new uninterned symbol whose name is name and hash is hash. The name must be a string and the hash must be a nonnegative fixnum.

The procedure uninterned-symbol? returns #t when obj is a symbol that is uninterned and #f otherwise.

For example:

> (uninterned-symbol? (gensym))
#t
> (string->uninterned-symbol "foo")
#:foo:
> (uninterned-symbol? (string->uninterned-symbol "foo"))
#t
> (uninterned-symbol? 'hello)
#f
> (uninterned-symbol? 123)
#f

The procedure string->uninterned-keyword returns a new uninterned keyword whose name is name and hash is hash. The name must be a string and the hash must be a nonnegative fixnum.

The procedure uninterned-keyword? returns #t when obj is a keyword that is uninterned and #f otherwise.

For example:

> (string->uninterned-keyword "foo")
#:foo:
> (uninterned-keyword? (string->uninterned-keyword "foo"))
#t
> (uninterned-keyword? hello:)
#f
> (uninterned-keyword? 123)
#f

This procedure returns obj.

This procedure returns the void object. The read-eval-print loop prints nothing when the result is the void object.

The first parameter is a datum representing an expression. The eval procedure evaluates this expression in the global interaction environment and returns the result. If present, the second parameter is ignored (it is provided for compatibility with R5RS).

For example:

> (eval '(+ 1 2))
3
> ((eval 'car) '(1 2))
1
> (eval '(define x 5))
> x
5

Define name as a macro special form which expands into body. This form can only appear where a define form is acceptable. Macros are lexically scoped. The scope of a local macro definition extends from the definition to the end of the body of the surrounding binding construct. Macros defined at the top level of a Scheme module are only visible in that module. To have access to the macro definitions contained in a file, that file must be included either directly using the include special form or indirectly with the import special form. Macros which are visible from the REPL are also visible during the compilation of Scheme source files.

For example:

(define-macro (unless test . body)
  `(if ,test #f (begin ,@body)))

(define-macro (push var #!optional val)
  `(set! ,var (cons ,val ,var)))

To examine the code into which a macro expands you can use the compiler’s ‘-expansion’ option or the pp procedure. For example:

> (define-macro (push var #!optional val)
    `(set! ,var (cons ,val ,var)))
> (pp (lambda () (push stack 1) (push stack) (push stack 3)))
(lambda ()
  (set! stack (cons 1 stack))
  (set! stack (cons #f stack))
  (set! stack (cons 3 stack)))

Define name as a macro special form whose expansion is specified by expander. This form is available only when the runtime option ‘-:s’ is used. This option causes the loading of the ~~lib/syntax-case support library, which is the Hieb and Dybvig portable syntax-case implementation which has been ported to the Gambit interpreter and compiler. Note that this implementation of syntax-case does not support special forms that are specific to Gambit.

For example:

$ gsi -:s
Gambit v4.9.4

> (define-syntax unless
    (syntax-rules ()
      ((unless test body ...)
       (if test #f (begin body ...)))))
> (let ((test 111)) (unless (= 1 2) (list test test)))
(111 111)
> (pp (lambda () (let ((test 111)) (unless (= 1 2) (list test test)))))
(lambda () ((lambda (%%test14) (if (= 1 2) #f (list %%test14 %%test14))) 111))
> (unless #f (pp xxx))
*** ERROR IN (console)@7.16 -- Unbound variable: xxx

This procedure can only be executed during the phase of the Scheme code’s processing (compilation) that corresponds to macro expansion. Calls to this procedure are typically contained in macro definitions but they can also be contained in procedures that are called from a macro definition’s body directly or indirectly.

The result returned by the compilation-target procedure gives an indication of the target language of the compilation. This can be used to write macros that depend on the type of compilation and the target language.

When the result is the symbol T the macro expansion is in the context of compiling to the target language T, e.g. C, js, etc. When the result is a single element list (T) the macro expansion is for the interpreter which itself was compiled for the target language T, e.g. (C), (js), etc.

For example:

$ cat ct.scm
(define (level-0)
  (string-append "0: " (object->string (compilation-target))))

(define-macro (test)

  (define (level-1)
    (string-append "1: " (object->string (compilation-target))))

  (define-macro (level-2)
    (string-append "2: " (object->string (compilation-target))))

  `(list ,(level-1) ,(level-2)))

(pp (test))

(pp (level-0)) ;; run time exception
$ gsi ct.scm
("1: (C)" "2: (C)")
*** ERROR IN level-0, "ct.scm"@2.40 -- Not in compilation context
(compilation-target)
$ gsc -target js -exe ct.scm
$ ./ct
("1: js" "2: (C)")
*** ERROR IN level-0 -- Not in compilation context
(compilation-target)

Regardless of whether ‘ct.scm’ is being processed by the interpreter or the compiler, the body of the level-0 procedure is not in a compilation context and in the body of the level-2 macro the compilation target is (C) indicating that the macro expansion is being done for interpretation.

During the execution of the level-1 procedure, the compilation target will correspond to what is processing ‘ct.scm’ (interpreter or compiler).

Note that the compilation target can also be tested by the cond-expand special form.

The cond-expand expression type provides a way to statically expand different expressions depending on the presence or absence of a set of features. A ce-clause takes the following form:

(feature-requirement expression …)

The last clause can be an “else clause,” which has the form

(else expression)

A feature-requirement takes one of the following forms:

• feature-identifier
• (library library-name)
• (and feature-requirement …)
• (or feature-requirement …)
• (not feature-requirement)
• (compilation-target target …)

The runtime system maintains a list of feature identifiers which are present, as well as a list of libraries which can be imported. The value of a feature-requirement is determined by replacing each feature-identifier and (library library-name) on the runtime system’s lists with #t. Similarly, #t replaces each (compilation-target target …) for which one of the target matches the expansion time value of (compilation-target), with a target of (_) matching any single element list (i.e. the interpreter). All other feature-identifier, (library library-name), and (compilation-target target …) are replaced with #f. The resulting expression is then evaluated as a Scheme boolean expression under the normal interpretation of and, or, and not.

A cond-expand is then expanded by evaluating the feature-requirements of successive ce-clauses in order until one of them returns #t. When a true clause is found, the corresponding expressions are expanded to a begin, and the remaining clauses are ignored. If none of the feature-requirements evaluate to #t, then if there is an else clause, its expressions are included. Otherwise, an expansion time error is raised. Unlike cond, cond-expand does not depend on the value of any variables.

The feature identifier gambit is always true when the cond-expand is expanded by the Gambit interpreter or compiler.

For example:

> (cond-expand (foobar 111) (gambit 222) (else 333))
222
> (cond-expand ((compilation-target js) 111) (else 222))
222
> (cond-expand ((compilation-target (_)) 111) (else 222))
111

The define-cond-expand-feature form can be used to add the feature identifiers feature-identifier … to the list of features maintained by the runtime system. These features are usable for the expansion of following cond-expand forms in the same file of source code, and the processing of other files and REPL interactions.

For example:

> (cond-expand (foobar 111) (gambit 222) (else 333))
222
> (define-cond-expand-feature foobar)
> (cond-expand (foobar 111) (gambit 222) (else 333))
111

This form introduces declarations to be used by the compiler (currently the interpreter ignores the declarations). This form can only appear where a define form is acceptable. Declarations are lexically scoped in the same way as macros. The following declarations are accepted by the compiler:

(dialect)

Use the given dialect’s semantics. dialect can be: ‘ieee-scheme’, ‘r4rs-scheme’, ‘r5rs-scheme’ or ‘gambit-scheme’.

(strategy)

Select block compilation or separate compilation. In block compilation, the compiler assumes that global variables defined in the current file that are not mutated in the file will never be mutated. strategy can be: ‘block’ or ‘separate’.

([not] inline)

Allow (or disallow) inlining of user procedures.

([not] inline-primitives primitive…)

The given primitives should (or should not) be inlined if possible (all primitives if none specified).

(inlining-limit n)

Select the degree to which the compiler inlines user procedures. n is the upper-bound, in percent, on code expansion that will result from inlining. Thus, a value of 300 indicates that the size of the program will not grow by more than 300 percent (i.e. it will be at most 4 times the size of the original). A value of 0 disables inlining. The size of a program is the total number of subexpressions it contains (i.e. the size of an expression is one plus the size of its immediate subexpressions). The following conditions must hold for a procedure to be inlined: inlining the procedure must not cause the size of the call site to grow more than specified by the inlining limit, the site of definition (the define or lambda) and the call site must be declared as (inline), and the compiler must be able to find the definition of the procedure referred to at the call site (if the procedure is bound to a global variable, the definition site must have a (block) declaration). Note that inlining usually causes much less code expansion than specified by the inlining limit (an expansion around 10% is common for n=370).

(allocation-limit n)

Indicate the maximum size of objects allocated with make-vector, make-string, make-u8vector, etc. Knowing the maximum size allows the compiler to inline calls to these allocators for small allocations. This is only supported by the C target and only up to a size that is allowed for movable objects (typically on the order of 1-2 KB). When n is an exact nonnegative integer it is the upper-bound on the number of elements of the allocated objects. When n is #t a dynamic test of the size is done. When n is #f the allocation operation is not inlined.

([not] lambda-lift)

Lambda-lift (or don’t lambda-lift) locally defined procedures.

([not] constant-fold)

Allow (or disallow) constant-folding of primitive procedures.

([not] standard-bindings var…)

The given global variables are known (or not known) to be equal to the value defined for them in the dialect (all variables defined in the standard if none specified).

([not] extended-bindings var…)

The given global variables are known (or not known) to be equal to the value defined for them in the runtime system (all variables defined in the runtime if none specified).

([not] run-time-bindings var…)

The given global variables will be tested at run time to see if they are equal to the value defined for them in the runtime system (all variables defined in the runtime if none specified).

([not] safe)

Generate (or don’t generate) code that will prevent fatal errors at run time. Note that in ‘safe’ mode certain semantic errors will not be checked as long as they can’t crash the system. For example the primitive char=? may disregard the type of its arguments in ‘safe’ as well as ‘not safe’ mode.

([not] interrupts-enabled)

Generate (or don’t generate) interrupt checks. Interrupt checks are used to detect user interrupts and also to check for stack overflows. Interrupt checking should not be turned off casually.

([not] poll-on-return)

Generate (or don’t generate) interrupt checks on procedure returns (when interrupt checking is enabled). This declaration has no effect on the behavior of interrupt checking on procedure calls, which is needed to guarantee that stack overflows are handled properly.

([not] debug)

Enable (or disable) the generation of debugging information. The kind of debugging information that is generated depends on the declarations ‘debug-location’, ‘debug-source’, and ‘debug-environments’. If any of the command line options ‘-debug’, ‘-debug-location’, ‘-debug-source’ and ‘-debug-environments’ are present, the ‘debug’ declaration is initially enabled, otherwise it is initially disabled. When all kinds of debugging information are generated there is a substantial increase in the C compilation time and the size of the generated code. When compiling a 3000 line Scheme file it was observed that the total compilation time is 500% longer and the executable code is 150% bigger.

([not] debug-location)

Select (or deselect) source code location debugging information. When this declaration and the ‘debug’ declaration are in effect, run time error messages indicate the location of the error in the source code file. If any of the command line options ‘-debug-source’ and ‘-debug-environments’ are present and ‘-debug-location’ is absent, the ‘debug-location’ declaration is initially disabled, otherwise it is initially enabled. When compiling a 3000 line Scheme file it was observed that the total compilation time is 200% longer and the executable code is 60% bigger.

([not] debug-source)

Select (or deselect) source code debugging information. When this declaration and the ‘debug’ declaration are in effect, run time error messages indicate the source code, the backtraces are more precise, and the pp procedure will display the source code of compiled procedures. If any of the command line options ‘-debug-location’ and ‘-debug-environments’ are present and ‘-debug-source’ is absent, the ‘debug-source’ declaration is initially disabled, otherwise it is initially enabled. When compiling a 3000 line Scheme file it was observed that the total compilation time is 90% longer and the executable code is 90% bigger.

([not] debug-environments)

Select (or deselect) environment debugging information. When this declaration and the ‘debug’ declaration are in effect, the debugger will have access to the environments of the continuations. In other words the local variables defined in compiled procedures (and not optimized away by the compiler) will be shown by the ‘,e’ REPL command. If any of the command line options ‘-debug-location’ and ‘-debug-source’ are present and ‘-debug-environments’ is absent, the ‘debug-environments’ declaration is initially disabled, otherwise it is initially enabled. When compiling a 3000 line Scheme file it was observed that the total compilation time is 70% longer and the executable code is 40% bigger.

([not] proper-tail-calls)

Generate (or don’t generate) proper tail calls. When proper tail calls are turned off, tail calls are handled like non-tail calls, that is a continuation frame will be created for all calls regardless of their kind. This is useful for debugging because the caller of a procedure will be visible in the backtrace produced by the REPL’s ‘,b’ command even when the call is a tail call. Be advised that this does cause stack space to be consumed for tail calls which may cause the stack to overflow when performing long iterations with tail calls (whether they are expressed with a letrec, named let, do, or other form).

([not] generative-lambda)

Force (or don’t force) the creation of fresh closures when evaluating lambda-expressions. A fresh closure is always created when a lambda-expression has at least one free variable (that has not been eliminated by dead-code elimination or other compiler optimization) or when the generative-lambda declaration is turned on. When a lambda-expression has no free variables and the generative-lambda declaration is turned off, the value of the lambda-expression may be the same procedure (in the sense of eq?).

([not] optimize-dead-local-variables)

Remove (or preserve) the dead local variables in the environment. Preserving the dead local variables is useful for debugging because continuations will contain the dead variables. Thus, if the code is also compiled with the declaration ‘debug-environments’ the ‘,e’, ‘,ed’, ‘,be’, and ‘,bed’ REPL commands will display the dead variables. On the other hand, preserving the dead local variables may change the space complexity of the program (i.e. some of the data that would normally be reclaimed by the garbage collector will not be). Note that due to other compiler optimizations some dead local variables may be removed regardless of this declaration.

([not] optimize-dead-definitions var…)

Remove (or preserve) the dead toplevel definitions of the given global variables (all global variables if none specified). A toplevel definition is dead if it is not referenced by toplevel expressions of the program or toplevel definitions that aren’t dead (regardless of the evaluation of its expression causing a side-effect). When a module is separately compiled and some of its definitions are only used by other modules, this declaration must be used with care to keep definitions that are used by other modules, for example if foo is referenced in another module the following declaration should be used: ‘(declare (not optimize-dead-definitions foo))’.

(number-type primitive…)

Numeric arguments and result of the specified primitives are known to be of the given type (all primitives if none specified). number-type can be: ‘generic’, ‘fixnum’, or ‘flonum’.

(mostly-number-type primitive…)

Numeric arguments and result of the specified primitives are expected to be most often of the given type (all primitives if none specified). mostly-number-type can be: ‘mostly-generic’, ‘mostly-fixnum’, ‘mostly-fixnum-flonum’, ‘mostly-flonum’, or ‘mostly-flonum-fixnum’.

The default declarations used by the compiler are equivalent to:

(declare
  (gambit-scheme)
  (separate)
  (inline)
  (inline-primitives)
  (inlining-limit 370)
  (allocation-limit #t)
  (constant-fold)
  (lambda-lift)
  (not standard-bindings)
  (not extended-bindings)
  (run-time-bindings)
  (safe)
  (interrupts-enabled)
  (not poll-on-return)
  (not debug)           ;; depends on debugging command line options
  (debug-location)      ;; depends on debugging command line options
  (debug-source)        ;; depends on debugging command line options
  (debug-environments)  ;; depends on debugging command line options
  (proper-tail-calls)
  (not generative-lambda)
  (optimize-dead-local-variables)
  (not optimize-dead-definitions)
  (generic)
  (mostly-fixnum-flonum)
)

These declarations are compatible with the semantics of R5RS Scheme and includes a few procedures from R6RS (mainly fixnum specific and flonum specific procedures). Typically used declarations that enhance performance, at the cost of violating the R5RS Scheme semantics, are: (standard-bindings), (block), (not safe) and (fixnum).

These procedures provide access to internal first-class continuations which are represented using continuation objects distinct from procedures.

The procedure continuation? returns #t when obj is a continuation object and #f otherwise.

The procedure continuation-capture is similar to the call/cc procedure but it represents the continuation with a continuation object. The proc parameter must be a procedure accepting a single argument. The procedure continuation-capture reifies its continuation and calls proc with the corresponding continuation object as its sole argument. Like for call/cc, the implicit continuation of the call to proc is the implicit continuation of the call to continuation-capture.

The procedure continuation-graft performs a procedure call to the procedure proc with arguments obj… and the implicit continuation corresponding to the continuation object cont. The current continuation of the call to procedure continuation-graft is ignored.

The procedure continuation-return invokes the implicit continuation corresponding to the continuation object cont with the result(s) obj…. This procedure can be easily defined in terms of continuation-graft:

(define (continuation-return cont . objs)
  (continuation-graft cont apply values objs))

For example:

> (define x #f)
> (define p (make-parameter 11))
> (pp (parameterize ((p 22))
        (cons 33 (continuation-capture
                  (lambda (c) (set! x c) 44)))))
(33 . 44)
> x
#<continuation #2>
> (continuation-return x 55)
(33 . 55)
> (continuation-graft x (lambda () (expt 2 10)))
(33 . 1024)
> (continuation-graft x expt 2 10)
(33 . 1024)
> (continuation-graft x (lambda () (p)))
(33 . 22)
> (define (map-sqrt1 lst)
    (call/cc
     (lambda (k)
       (map (lambda (x)
              (if (< x 0)
                  (k 'error)
                  (sqrt x)))
            lst))))
> (map-sqrt1 '(1 4 9))
(1 2 3)
> (map-sqrt1 '(1 -1 9))
error
> (define (map-sqrt2 lst)
    (continuation-capture
     (lambda (c)
       (map (lambda (x)
              (if (< x 0)
                  (continuation-return c 'error)
                  (sqrt x)))
            lst))))
> (map-sqrt2 '(1 4 9))
(1 2 3)
> (map-sqrt2 '(1 -1 9))
error

The procedure display-continuation-backtrace displays the frames of the continuation corresponding to the continuation object cont on the port port. If it is not specified, port defaults to the current output-port. The frames are displayed in the same format as the REPL’s ‘,b’ command.

The parameter display-env?, which defaults to #f, controls if the frames are displayed with its environment (the variables accessible and their bindings).

The parameter all-frames?, which defaults to #f, controls which frames are displayed. Some frames of ancillary importance, such as internal frames created by the interpreter, are not displayed when all-frames? is #f. Otherwise all frames are displayed.

The parameters max-head and max-tail, which default to 10 and 4 respectively, control how many frames are displayed at the head and tail of the continuation.

The parameter depth, which defaults to 0, causes the frame numbers to be offset by that value.

For example:

> (define x #f)
> (define (fib n)
    (if (< n 2)
        (continuation-capture
         (lambda (c) (set! x c) 1))
        (+ (fib (- n 1))
           (fib (- n 2)))))
> (fib 10)
89
> (display-continuation-backtrace x)
0  fib             (console)@7:12     (fib (- n 2))
1  fib             (console)@7:12     (fib (- n 2))
2  fib             (console)@7:12     (fib (- n 2))
3  fib             (console)@7:12     (fib (- n 2))
4  fib             (console)@7:12     (fib (- n 2))
5  (interaction)   (console)@8:1      (fib 10)
#f
> (display-continuation-backtrace x (current-output-port) #f #t)
0  fib             (console)@7:12     (fib (- n 2))
1  fib             (console)@6:9      (+ (fib (- n 1)) (fib (- ...
2  fib             (console)@7:12     (fib (- n 2))
3  fib             (console)@6:9      (+ (fib (- n 1)) (fib (- ...
4  fib             (console)@7:12     (fib (- n 2))
5  fib             (console)@6:9      (+ (fib (- n 1)) (fib (- ...
6  fib             (console)@7:12     (fib (- n 2))
7  fib             (console)@6:9      (+ (fib (- n 1)) (fib (- ...
8  fib             (console)@7:12     (fib (- n 2))
9  fib             (console)@6:9      (+ (fib (- n 1)) (fib (- ...
...
13 ##with-no-result-expected-toplevel
14 ##repl-debug
15 ##repl-debug-main
16 ##kernel-handlers
#f
> (display-continuation-backtrace x (current-output-port) #t #f)
0  fib             (console)@7:12     (fib (- n 2))
        n = 2
1  fib             (console)@7:12     (fib (- n 2))
        n = 4
2  fib             (console)@7:12     (fib (- n 2))
        n = 6
3  fib             (console)@7:12     (fib (- n 2))
        n = 8
4  fib             (console)@7:12     (fib (- n 2))
        n = 10
5  (interaction)   (console)@8:1      (fib 10)
#f
> (display-continuation-backtrace x (current-output-port) #f #f 2 1 100)
100 fib            (console)@7:12     (fib (- n 2))
101 fib            (console)@7:12     (fib (- n 2))
...
105 (interaction)  (console)@8:1      (fib 10)
#f

6.4 Undocumented extensions

The procedures in this section are not yet documented.