http://people.csail.mit.edu/jaffer/Schlep/scm2c.html | ||||||
Scm2c | ||||||
|
#include "schleprt.h"
in files generated by scm2c.
Compile schleprt.c and link with the program or library
generated by scm2c.
Usage: scm2c [-p stdc | nil | scm] FILE1.scm FILE2.scm ... [DIR/] Usage: scm2c [-p stdc | nil | scm] FILE1.c FILE2.c ... [DIR/] Translates Scheme files FILE1.scm, FILE2.scm, ... to DIR/FILEn.c, DIR/FILEn.h, and DIR/FILEn.txi. Usage: scm2c [-p stdc | nil | scm] FILE1.h FILE2.h ... [DIR/] Translates Scheme files FILE1.scm, FILE2.scm, ... to DIR/FILEn.h and DIR/FILEn.txi. Options: -p stdc FILE*.h will have ANSI prototypes -p nil FILE*.h will have () prototypes -p scm FILE*.h will have SCM conditional prototypes http://people.csail.mit.edu/jaffer/Schlep/scm2c
Scm2c is a Scheme to C translator for a subset of Scheme. Using Scheme files as source, scm2c produces texinfo documentation and formatted C code preserving comments; and type, function, and variable names as much as possible. The output from scm2c is human-readable and can form the base for further development in C; abandoning the original Scheme source.
Note that scm2c is a translator -- the C code it produces will be nearly as readable as the original Scheme source. An unrelated project, Hobbit compiles full R4RS Scheme to C functions for use with the SCM Scheme Implementation.
The Scheme subset supported by scm2c is listed at
http://people.csail.mit.edu/jaffer/Schlep/#Schlep Subset.
Scheme integers and real numbers, booleans, chars, and strings are
translated to the equivalent C types. SLIB byte vectors map to the
type byte*
in C code. Vectors of user defined (C) types
are supported.
The user incorporates custom C data types by coding their definitions and accessors in a C include file. For the purposes of running in Scheme, put analogous definitions in a file which is not translated by scm2c.
Characters in variable names are translated as follows:
`%' | => | `_Percent' |
`@' | => | `_At' |
`->' | => | `to_' |
`=' | => | `_equal' |
`>' | => | `_more' |
`<' | => | `_less' |
`?' | => | `_P' |
`:' | => | `_' |
`.' | => | `_' |
`-' | => | `_' |
Scm2c does not include support for a Scheme runtime. Generated C programs should not assume garbage collection or general Scheme values unless linked with a library providing these features.
Scheme identifier names are used to determine the C types of the corresponding variables and functions. This is described in the Declarations section.
Defmacros transforming Scheme code to Scheme code may be defined in
Scheme source files. If there are defmacros used in more than one
Scheme source file, they can be removed to a separate file, and that
file imported by a call to defmacro:load
in each file (before
a defmacro is invoked).
Imports defmacro
s from filename.
Is an alist associating defmacro names with their transformers. It
is fluid-let
during each file translatation, so that scm2c
macro definitions do not take effect in the Scheme top level.
(glob type)
. glob is a
match string; type an expression. Globs are matched to
variable names with trailing digits removed. Declarations made
later override those made earlier.
Declarations can be made by a call to declare-names
. If
no declarations have been made before schlep
is called,
schlep
will read a list of declarations from a file named
`scm2c.typ' if it exists in the same directory as the Scheme
file being translated.
If the type is a symbol or string, variables which match the glob string will be declared as types of that name. If the type is a list, then it will be interpreted as follows.
(ptr type)
* type
.
(array type)
type[]
.
Types can be nested. Procedure names ending with `!' are typed
to return void
. Procedure names ending with `?' are
translated to names with a trailing `_P'.
The Scheme files in the table at
http://people.csail.mit.edu/jaffer/CNS/benchmarks.html#PRNG
have examples of the use of declare-names
.
pragma.h
has no effect, but the strings are written to
filename.h during translation.
pragma.c
has no effect, but the strings are written to
filename.c during translation.
To work with the conditional prototypes, an include file loaded
before the Schlepped .h files should contain:
#include "schleprt.h"
#include "schleprt.h"
or its content may also be
needed if your code uses min
, max
, non-stack
allocations, or the diagnostic output routines dprintf
,
wdprintf
, or edprintf
. If you use the diagnostic
output routines, you must also define diagout. The file
"schleprt.c" does this; its entire content is:
#include <stdio.h> FILE *diagout;
Define number-lines? to #t for C code to be prefixed with commmented line numbers. The default value is #f.
The following procedures are called to translate a file. They should not be called from the file being translated.
Otherwise, the file named by the argument filename is translated to filename.c, filename,h, and filename.txi.
If filename.txi is empty after translation, then it is deleted.
declare-name!
) along with the counts of references to each
name pattern type declared.
I am a guest and not a member of the MIT Computer Science and Artificial Intelligence Laboratory. My actions and comments do not reflect in any way on MIT. | ||
Schlep | ||
agj @ alum.mit.edu | Go Figure! |