robert-strandh / Bootstrap-Common-Lisp

An implementation of Common Lisp with the sole purpose of bootstrapping other implementations.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Welcome to Bootstrap Common Lisp, or BOCL for short.

BOCL is (meant to one day be) a fully conforming Common Lisp
implementation with a twist.  It is meant only for bootstrapping other
Common Lisp implementations on platforms where, for some reason, no
Common Lisp implementation is already available.  It is not meant for
writing end-user applications.

This main purpose of BOCL has some consequences to the way it is
implemented.  In particular, there is no attempt to make it fast.
Instead, the emphasis is on simplicity, maintainability, and
correctness.  Specifically:

  * We use the Boehm-Weiser conservative garbage collector.

  * We take advantage of all situations in which the Common Lisp
    standard states that some behavior is undefined or implementation
    defined, in that we then do whatever is the easiest solution for
    the situation.

  * The type FIXNUM does not have a corresponding class.  All integers
    are represented the same way.

  * All objects have the same basic representation as a header
    containing a pointer to the class of the object and a pointer to a
    vector (not a Common Lisp vector) with the objects-specific
    contents.

  * The only specialized arrays are those mandated by the Common Lisp
    standard, and even those arrays do not use any special
    representation.  Thus an array of single bits will still have a
    full BOCL object for each element, but that object is either the
    integer 0 or the integer 1.

  * More generally, we make no attempt to save space.

  * Hash tables are internally implemented as lists for simplicity.

  * There is no attempt to handle tail-call merging.

  * All functions are called with three arguments: the static
    environment, the dynamic environment, and a Common Lisp list of
    arguments.

Currently, several free (or, more generally, FLOSS) implementations of
Common Lisp are built from sources in some other language.  That
language is typically C, but sometimes something else, like C++.
There are probably several reasons for these implementations to be
written this way, including historical (there was no widespread
existing Common Lisp implementation when they were written), limited
knowledge (the author did not know how to write a Common Lisp
implementation in Common Lisp), and more.

The process of building a Common Lisp system from sources written in
some other language is a painful one, especially if one of the main
objectives is for it to be fast, safe, and helpful to the programmer.
Large parts of the system must be written in a language that is not
very well adapted to expressing the semantics of Common Lisp.  Even
though large parts of the system can be written in Common Lisp,
because of the way the system is built, many system modules must be
written in a subset of Common Lisp, making also that part painful to
write.  For instance, it is common that CLOS not be available until
very late in the bootstrapping process.

The existence of BOCL will relieve most of the pain of maintainers of
Common Lisp systems by concentrating that pain into BOCL.  These
maintainers can now migrate their code from the existing
implementation language to Common Lisp, relying on BOCL to exist.  At
the same time, there is considerably less pain because of the
simplifications to BOCL listed above.

BOCL is (or rather, will be) implemented in ISO C.  We (will) do our
utmost to avoid using constructs with undefined semantics, as the C
standard indicates.  Only readily available shell tools are used in
order to maximize portability.

As an exception to the goal of simplicity, BOCL will contain a full
implementation of the CLOS meta-object protocol (MOP) so as to allow
implementers of other Common Lisp systems to take advantage of the MOP
in their own implementations.

In addition to being a bootstrapping tool for Common Lisp
implementers, we think that BOCL will be an excellent illustration of
the semantics of Common Lisp.  To make it easier for people interested
in that topic, we think it is important that the BOCL code be highly
idiomatic, uniform, and well documented.

BOCL will not happen unless it is contributed to by others, so if you
have knowledge of how to implement Common Lisp in C or interest in
learning how to do it, please consider contributing.

About

An implementation of Common Lisp with the sole purpose of bootstrapping other implementations.


Languages

Language:C 89.2%Language:C++ 7.2%Language:Makefile 3.6%