A small Scheme written in Python.

Originally written as a warmup exercise to prepare for this class.

smallscheme is not a complete R7RS Scheme (or R5RS for that matter) -- it implements everything needed to follow along in SICP up to page 102 (however, see "Limitations," below). smallscheme is very lightweight, is extendable, and can be installed on any Python 3 installation as follows:

pip install smallscheme

See tests.scm for many examples of the language in action.

$ smallscheme
scheme> (define (fact n) (if (< n 2) n (* n (fact (- n 1)))))
scheme> (fact 50)
30414093201713378043612608166064768844377641568960512000000000000
scheme> ^D
$

The REPL comes with arrow-key history, line editing, and other features provided by the Python Prompt Toolkit.

Example:

$  cat fact.scm
(define (fact n)
  (if (< n 2)
      n
      (* n (fact (- n 1)))))

(define f100 (fact 100))

(display f100)
(newline)
$  smallscheme fact.scm
933262154439441526816992388562667004907159682643816214685929638
952175999932299156089414639761565182862536979208272237582511852
10916864000000000000000000000000
$

smallscheme implements the basics of Scheme required to follow examples or work problems in SICP, including the following:

*
+
-
/
<
=
>
atan
car
cdr
cons
cos
display
newline
not
random
remainder
runtime
sin

There are also two simple functions used in the Scheme-language tests: is and test; test currently behaves like a progn or do in other lisps, in that it collects multiple forms to be evaluated and returns the result of the last evaluation. is is basically assert.

and
cond
define
if
lambda
let
or
quote

For explanation of these, and of Scheme in general, I recommend reading Structure and Intepretation of Computer Programs, in print or (free!) online.

smallscheme is easily extendable using Python functions.

To add a new function to smallscheme, create a function which accepts a list of arguments args. These will have to be converted into Python objects via the helper functions in smallscheme.dtypes. For example, to create a function inc which increments its argument,

from smallscheme.dtypes import *
from smallscheme.interop import scheme_fn

@scheme_fn
def inc(args):
    assert len(args) == 1, "inc only takes one argument"
    arg_number = args[0]
    assert typeof(arg_number) in ['int', 'float'], "inc expects a number"
    num_value = value(arg_number)
    return int_(num_value + 1)

The scheme_fn decorator handles registration of your Python function with the interpreter, which could also be registered explicitly, as follows:

from smallscheme.interop import register_fn
register_fn('inc', inc)

For function names which are valid in Scheme but not in Python (such as +), you'll need to use register_fn rather than scheme_fn.

At the current time, any builtin function (but not special forms) can be overridden using this mechanism. There's an example of this in the tests.

Once your Python functions have been registered, your code can then execute Scheme expressions, either by executing one or more Scheme source files, or by launching a REPL for the user:

import smallscheme

smallscheme.run_file("myprogram.scm")

# ... or ...:

smallscheme.repl()

... and the new function can then be called from Scheme:

scheme> (inc 1)
;;=>
2

For a full list of data type operators (for converting from Python to Scheme and vice-versa), look at dtypes.py. See also the sample program in the examples folder.

Not a production-ready language implementation -- error messages and performance in particular may not be the best. Features used in later parts of SICP may not be available yet. I have been implementing them roughly in the order they are introduced.

The status of this work should be considered very alpha. APIs and implementation details may change.

smallscheme uses Python lists to represent Scheme lists. As such it does not support dotted-pair notation or consing onto a non-list. This will require slight changes in a few of the examples in SICP. The pair representation of rational numbers in Chapter 2, for example, would need to be adapted to use lists of two numbers rather than a single cons pair.

pip install -r requirements.txt

Python tests are run with pytest.

Scheme tests are run with ./smallscheme/main.py -t tests.scm. These tests are particularly helpful in seeing what's been implemented so far.

• Original blog post introducing smallscheme
• Structure and Interpretation of Computer Programs Web site
• P. Norvig, (How to Write a (Lisp) Interpreter (in Python))
• l1, a simple Lisp written in Go by the author

MIT

Some code fragments used in tests are copied directly from SICP, which is licensed with the Creative Commons "Attribution-ShareAlike 4.0 International (CC BY-SA 4.0)" license.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.