#! /usr/bin/env python
"""Generate C code from an ASDL description."""
# TO DO
# handle fields that have a type but no name
import os, sys, traceback
import asdl
TABSIZE = 8
MAX_COL = 80
def get_c_type(name):
"""Return a string for the C name of the type.
This function special cases the default types provided by asdl:
identifier, string, int, bool.
"""
# XXX ack! need to figure out where Id is useful and where string
if isinstance(name, asdl.Id):
name = name.value
if name in asdl.builtin_types:
return name
else:
return "%s_ty" % name
def reflow_lines(s, depth):
"""Reflow the line s indented depth tabs.
Return a sequence of lines where no line extends beyond MAX_COL
when properly indented. The first line is properly indented based
exclusively on depth * TABSIZE. All following lines -- these are
the reflowed lines generated by this function -- start at the same
column as the first character beyond the opening { in the first
line.
"""
size = MAX_COL - depth * TABSIZE
if len(s) < size:
return [s]
lines = []
cur = s
padding = ""
while len(cur) > size:
i = cur.rfind(' ', 0, size)
# XXX this should be fixed for real
if i == -1 and 'GeneratorExp' in cur:
i = size + 3
assert i != -1, "Impossible line %d to reflow: %s" % (size, `s`)
lines.append(padding + cur[:i])
if len(lines) == 1:
# find new size based on brace
j = cur.find('{', 0, i)
if j >= 0:
j += 2 # account for the brace and the space after it
size -= j
padding = " " * j
else:
j = cur.find('(', 0, i)
if j >= 0:
j += 1 # account for the paren (no space after it)
size -= j
padding = " " * j
cur = cur[i+1:]
else:
lines.append(padding + cur)
return lines
def is_simple(sum):
"""Return True if a sum is a simple.
A sum is simple if its types have no fields, e.g.
unaryop = Invert | Not | UAdd | USub
"""
for t in sum.types:
if t.fields:
return False
return True
class EmitVisitor(asdl.VisitorBase):
"""Visit that emits lines"""
emit("struct _%(name)s {")
emit("enum _%(name)s_kind kind;", depth + 1)
emit("union {", depth + 1)
for t in sum.types:
self.visit(t, depth + 2)
emit("} v;", depth + 1)
for field in sum.attributes:
# rudimentary attribute handling
type = str(field.type)
assert type in asdl.builtin_types, type
emit("%s %s;" % (type, field.name), depth + 1);
emit("};")
emit("")
def visitConstructor(self, cons, depth):
if cons.fields:
self.emit("struct {", depth)
for f in cons.fields:
self.visit(f, depth + 1)
self.emit("} %s;" % cons.name, depth)
self.emit("", depth)
else:
# XXX not sure what I want here, nothing is probably fine
pass
def visitField(self, field, depth):
# XXX need to lookup field.type, because it might be something
# like a builtin...
ctype = get_c_type(field.type)
name = field.name
if field.seq:
if field.type.value in ('cmpop',):
self.emit("asdl_int_seq *%(name)s;" % locals(), depth)
else:
self.emit("asdl_seq *%(name)s;" % locals(), depth)
else:
self.emit("%(ctype)s %(name)s;" % locals(), depth)
def visitProduct(self, product, name, depth):
self.emit("struct _%(name)s {" % locals(), depth)
for f in product.fields:
self.visit(f, depth + 1)
self.emit("};", depth)
self.emit("", depth)
class PrototypeVisitor(EmitVisitor):
"""Generate function prototypes for the .h file"""
def visitModule(self, mod):
for dfn in mod.dfns:
self.visit(dfn)
def visitSum(self, sum, name):
if is_simple(sum):
pass # XXX
else:
for t in sum.types:
self.visit(t, name, sum.attributes)
def get_args(self, fields):
"""Return list of C argument into, one for each field.
Argument info is 3-tuple of a C type, variable name, and flag
that is true if type can be NULL.
"""
args = []
unnamed = {}
for f in fields:
if f.name is None:
name = f.type
c = unnamed[name] = unnamed.get(name, 0) + 1
if c > 1:
name = "name%d" % (c - 1)
else:
name = f.name
# XXX should extend get_c_type() to handle this
if f.seq:
if f.type.value in ('cmpop',):
ctype = "asdl_int_seq *"
else:
ctype = "asdl_seq *"
else:
ctype = get_c_type(f.type)
args.append((ctype, name, f.opt or f.seq))
return args
def prototype(self, sum, name):
ctype = get_c_type(name)
self.emit("static int marshal_write_%s(PyObject **, int *, %s);"
% (name, ctype), 0)
visitProduct = visitSum = prototype
class PyTypesDeclareVisitor(PickleVisitor):
def visitProduct(self, prod, name):
self.emit("static PyTypeObject *%s_type;" % name, 0)
self.emit("static PyObject* ast2obj_%s(void*);" % name, 0)
if prod.fields:
self.emit("static char *%s_fields[]={" % name,0)
for f in prod.fields:
self.emit('"%s",' % f.name, 1)
self.emit("};", 0)
def visitSum(self, sum, name):
self.emit("static PyTypeObject *%s_type;" % name, 0)
if sum.attributes:
self.emit("static char *%s_attributes[] = {" % name, 0)
for a in sum.attributes:
self.emit('"%s",' % a.name, 1)
self.emit("};", 0)
ptype = "void*"
if is_simple(sum):
ptype = get_c_type(name)
tnames = []
for t in sum.types:
tnames.append(str(t.name)+"_singleton")
tnames = ", *".join(tnames)
self.emit("static PyObject *%s;" % tnames, 0)
self.emit("static PyObject* ast2obj_%s(%s);" % (name, ptype), 0)
for t in sum.types:
self.visitConstructor(t, name)
def visitConstructor(self, cons, name):
self.emit("static PyTypeObject *%s_type;" % cons.name, 0)
if cons.fields:
self.emit("static char *%s_fields[]={" % cons.name, 0)
for t in cons.fields:
self.emit('"%s",' % t.name, 1)
self.emit("};",0)
class PyTypesVisitor(PickleVisitor):
def visitModule(self, mod):
self.emit("""
static PyTypeObject* make_type(char *type, PyTypeObject* base, char**fields, int num_fields)
{
PyObject *fnames, *result;
int i;
if (num_fields) {
fnames = PyTuple_New(num_fields);
if (!fnames) return NULL;
} else {
fnames = Py_None;
Py_INCREF(Py_None);
}
for(i=0; i < num_fields; i++) {
PyObject *field = PyString_FromString(fields[i]);
if (!field) {
Py_DECREF(fnames);
return NULL;
}
PyTuple_SET_ITEM(fnames, i, field);
}
result = PyObject_CallFunction((PyObject*)&PyType_Type, "s(O){sOss}",
type, base, "_fields", fnames, "__module__", "_ast");
Py_DECREF(fnames);
return (PyTypeObject*)result;
}
static int add_attributes(PyTypeObject* type, char**attrs, int num_fields)
{
int i, result;
PyObject *s, *l = PyList_New(num_fields);
if (!l) return 0;
for(i = 0; i < num_fields; i++) {
s = PyString_FromString(attrs[i]);
if (!s) {
Py_DECREF(l);
return 0;
}
PyList_SET_ITEM(l, i, s);
}
result = PyObject_SetAttrString((PyObject*)type, "_attributes", l) >= 0;
Py_DECREF(l);
return result;
}
static PyObject* ast2obj_list(asdl_seq *seq, PyObject* (*func)(void*))
{
int i, n = asdl_seq_LEN(seq);
PyObject *result = PyList_New(n);
PyObject *value;
if (!result)
return NULL;
for (i = 0; i < n; i++) {
value = func(asdl_seq_GET(seq, i));
if (!value) {
Py_DECREF(result);
return NULL;
}
PyList_SET_ITEM(result, i, value);
}
return result;
}
static PyObject* ast2obj_object(void *o)
{
if (!o)
o = Py_None;
Py_INCREF((PyObject*)o);
return (PyObject*)o;
}
#define ast2obj_identifier ast2obj_object
#define ast2obj_string ast2obj_object
static PyObject* ast2obj_bool(bool b)
{
return PyBool_FromLong(b);
}
def find_sequence(fields, doing_specialization):
"""Return True if any field uses a sequence."""
for f in fields:
if f.seq:
if not doing_specialization:
return True
if str(f.type) not in _SPECIALIZED_SEQUENCES:
return True
return False
def has_sequence(types, doing_specialization):
for t in types:
if find_sequence(t.fields, doing_specialization):
return True
return False
class StaticVisitor(PickleVisitor):
CODE = '''Very simple, always emit this static code. Overide CODE'''
def set(self, field, value, depth):
if field.seq:
# XXX should really check for is_simple, but that requires a symbol table
if field.type.value == "cmpop":
# While the sequence elements are stored as void*,
# ast2obj_cmpop expects an enum
self.emit("{", depth)
self.emit("int i, n = asdl_seq_LEN(%s);" % value, depth+1)
self.emit("value = PyList_New(n);", depth+1)
self.emit("if (!value) goto failed;", depth+1)
self.emit("for(i = 0; i < n; i++)", depth+1)
# This cannot fail, so no need for error handling
self.emit("PyList_SET_ITEM(value, i, ast2obj_cmpop((cmpop_ty)asdl_seq_GET(%s, i)));" % value,
depth+2, reflow=False)
self.emit("}", depth)
else:
self.emit("value = ast2obj_list(%s, ast2obj_%s);" % (value, field.type), depth)
else:
ctype = get_c_type(field.type)
self.emit("value = ast2obj_%s(%s);" % (field.type, value), depth, reflow=False)
class ChainOfVisitors:
def __init__(self, *visitors):
self.visitors = visitors
def visit(self, object):
for v in self.visitors:
v.visit(object)
v.emit("", 0)
def main(srcfile):
argv0 = sys.argv[0]
components = argv0.split(os.sep)
argv0 = os.sep.join(components[-2:])
auto_gen_msg = '/* File automatically generated by %s */\n' % argv0
mod = asdl.parse(srcfile)
if not asdl.check(mod):
sys.exit(1)
if INC_DIR:
p = "%s/%s-ast.h" % (INC_DIR, mod.name)
f = open(p, "wb")
print >> f, auto_gen_msg
print >> f, '#include "asdl.h"\n'
c = ChainOfVisitors(TypeDefVisitor(f),
StructVisitor(f),
PrototypeVisitor(f),
)
c.visit(mod)
print >>f, "PyObject* PyAST_mod2obj(mod_ty t);"
f.close()
if SRC_DIR:
p = os.path.join(SRC_DIR, str(mod.name) + "-ast.c")
f = open(p, "wb")
print >> f, auto_gen_msg
print >> f, '#include "Python.h"'
print >> f, '#include "%s-ast.h"' % mod.name
print >> f
print >>f, "static PyTypeObject* AST_type;"
v = ChainOfVisitors(
PyTypesDeclareVisitor(f),
PyTypesVisitor(f),
FunctionVisitor(f),
ObjVisitor(f),
ASTModuleVisitor(f),
PartingShots(f),
)
v.visit(mod)
f.close()
if __name__ == "__main__":
import sys
import getopt
INC_DIR = ''
SRC_DIR = ''
opts, args = getopt.getopt(sys.argv[1:], "h:c:")
if len(opts) != 1:
print "Must specify exactly one output file"
sys.exit(1)
for o, v in opts:
if o == '-h':
INC_DIR = v
if o == '-c':
SRC_DIR = v
if len(args) != 1:
print "Must specify single input file"
sys.exit(1)
main(args[0])