First commit

Completed interpreter, through chapter 4
2 months ago · 204736eceb
commit 204736eceb
1 changed files with 266 additions and 0 deletions
--- a/pl_comp.py
+++ b/pl_comp.py
@ -0,0 +1,266 @@
+class LoopBreak(Exception):
+    def __init__(self):
+        super().__init__('`break` outside a loop')
+
+
+class LoopContinue(Exception):
+    def __init__(self):
+        super().__init__('`continue` outside a loop')
+
+
+class FuncReturn(Exception):
+    def __init__(self, val):
+        super().__init__('`return` outside a function')
+        self.val = val
+
+
+def parse_expr(s: str, idx: int):
+    idx = skip_space(s, idx)
+    if s[idx] == '(':
+        # a list
+        idx += 1
+        l = []
+        while True:
+            idx = skip_space(s, idx)
+            if idx >= len(s):
+                raise Exception('unbalanced parenthesis')
+            if s[idx] == ')':
+                idx += 1
+                break
+
+            idx, v = parse_expr(s, idx)
+            l.append(v)
+        return idx, l
+    elif s[idx] == ')':
+        raise Exception('bad parenthesis')
+    else:
+        # an atom
+        start = idx
+        while idx < len(s) and (not s[idx].isspace()) and s[idx] not in '()':
+            idx += 1
+        if start == idx:
+            raise Exception('empty program')
+        return idx, parse_atom(s[start:idx])
+
+
+def skip_space(s: str, idx: int):
+    while True:
+        save = idx
+        while idx < len(s) and s[idx].isspace():
+            idx += 1
+        if idx < len(s) and s[idx] == ';':
+            idx += 1
+            while idx < len(s) and s[idx] != '\n':
+                idx += 1
+        if idx == save:
+            break
+    return idx
+
+
+def parse_atom(s):
+    # TODO: implement this
+    import json
+    try:
+        return ['val', json.loads(s)]
+    except json.JSONDecodeError:
+        return s
+
+
+def pl_parse(s):
+    idx, node = parse_expr(s, 0)
+    idx = skip_space(s, idx)
+    if idx < len(s):
+        raise ValueError('trailing garbage')
+    return node
+
+
+def name_lookup(env, key):
+    while env:
+        current, env = env
+        if key in current:
+            return current
+    raise ValueError('undefined name')
+
+
+def pl_eval(env, node):
+    if not isinstance(node, list):
+        assert isinstance(node, str)
+        return name_lookup(env, node)[node]
+
+    if len(node) == 0:
+        raise ValueError('empty list')
+
+    if len(node) == 2 and node[0] == 'val':
+        return node[1]
+
+    if node[0] == 'var' and len(node) == 3:
+        _, name, val = node
+        scope, _ = env
+        if name in scope:
+            raise ValueError('duplicated name')
+        val = pl_eval(env, val)
+        scope[name] = val
+        return val
+
+    if node[0] == 'set' and len(node) == 3:
+        _, name, val = node
+        scope = name_lookup(env, name)
+        val = pl_eval(env, val)
+        scope[name] = val
+        return val
+
+    if node[0] in ('do', 'then', 'else') and len(node) > 1:
+        new_env = (dict(), env)
+        # TODO: Why is Smith using the same variable name for the array items
+        # and their evaluations? Try using a different variable name for the
+        # latter.
+        for val in node[1:]:
+            val = pl_eval(new_env, val)
+        return val
+
+    import operator
+    binops = {
+        '+': operator.add,
+        '-': operator.sub,
+        '*': operator.mul,
+        '/': operator.truediv,
+        'eq': operator.eq,
+        'ne': operator.ne,
+        'ge': operator.ge,
+        'gt': operator.gt,
+        'le': operator.le,
+        'lt': operator.lt,
+        'and': operator.and_,
+        'or': operator.or_,
+    }
+    if len(node) == 3 and node[0] in binops:
+        op = binops[node[0]]
+        return op(pl_eval(env, node[1]), pl_eval(env, node[2]))
+
+    unops = {
+        '-': operator.neg,
+        'not': operator.not_,
+    }
+    if len(node) == 2 and node[0] in unops:
+        op = unops[node[0]]
+        return op(pl_eval(env, node[1]))
+
+    if len(node) in (3, 4) and node[0] in ('?', 'if'):
+        _, cond, yes, *no = node
+        no = no[0] if no else ['val', None]
+        new_env = (dict(), env)
+        if pl_eval(new_env, cond):
+            return pl_eval(new_env, yes)
+        else:
+            return pl_eval(new_env, no)
+
+    if node[0] == 'loop' and len(node) == 3:
+        _, cond, body = node
+        ret = None
+        while True:
+            new_env = (dict(), env)
+            if not pl_eval(new_env, cond):
+                break
+            try:
+                ret = pl_eval(new_env, body)
+            except LoopBreak:
+                break
+            except LoopContinue:
+                continue
+        return ret
+
+    if node[0] == 'break' and len(node) == 1:
+        raise LoopBreak
+    if node[0] == 'continue' and len(node) == 1:
+        raise LoopContinue
+
+    # function definition
+    if node[0] == 'def' and len(node) == 4:
+        _, name, args, body = node
+        # sanity checks
+        for arg_name in args:
+            if not isinstance(arg_name, str):
+                raise ValueError('bad argument name')
+        if len(args) != len(set(args)):
+            raise ValueError('duplicated arguments')
+        # add the function to the scope
+        dct, _ = env
+        key = (name, len(args))
+        if key in dct:
+            raise ValueError('duplicated function')
+        dct[key] = (args, body, env)
+        return
+
+    # function call
+    if node[0] == 'call' and len(node) >= 2:
+        _, name, *args = node
+        key = (name, len(args))
+        fargs, fbody, fenv = name_lookup(env, key)[key]
+        # args
+        new_env = dict()
+        for arg_name, arg_val in zip(fargs, args):
+            new_env[arg_name] = pl_eval(env, arg_val)
+        # call
+        try:
+            return pl_eval((new_env, fenv), fbody)
+        except FuncReturn as ret:
+            return ret.val
+
+    # return
+    if node[0] == 'return' and len(node) == 1:
+        raise FuncReturn(None)
+    if node[0] == 'return' and len(node) == 2:
+        _, val = node
+        raise FuncReturn(pl_eval(env, val))
+
+    if node[0] == 'print':
+        return print(*(pl_eval(env, val) for val in node[1:]))
+
+    raise ValueError('unknown expression')
+
+
+def pl_parse_prog(s):
+    return pl_parse('(do ' + s + ')')
+
+
+def test_eval():
+    def f(s):
+        return pl_eval(dict(), pl_parse_prog(s))
+    assert f('1') == 1
+    assert f('(+ 1 3)') == 4
+    assert f('(? (lt 1 3) "yes" "no")') == "yes"
+    assert f('(print 1 2 3)') is None
+    assert f('''
+        ;; first scope
+        (var a 1)
+        (var b (+ a 1))
+        ;; a=1, b=2
+        (do 
+            ;; new scope
+            (var a (+ b 5)) ;; name collision
+            (set b (+ a 10))
+        )
+        ;; a=1, b=17
+        (* a b)
+    ''') == 17
+    assert f('''
+        (def gauss (n)
+             (if (le n 0)
+                 (then 0)
+                 (else (+ n (call gauss (- n 1))))))
+        (call gauss 5)
+    ''') == 5 + 4 + 3 + 2 + 1
+    assert f('''
+        (def gauss (n) (do
+             (var r 0)
+             (loop (gt n 0) (do
+                 (set r (+ r n))
+                 (set n (- n 1))
+             ))
+             (return r)
+        ))
+        (call gauss 5)
+    ''') == 5 + 4 + 3 + 2 + 1
+
+
+test_eval()  # Run this file as python `/path/to/pl_comp.py`; should work