Hi,
in process of ripping out the exec statement, I stumbled over the
following function in symtable.c (line 468ff):
------------------------------------------------------------------------------------
/* Check for illegal statements in unoptimized namespaces */
static int
check_unoptimized(const PySTEntryObject* ste) {
char buf[300];
const char* trailer;
if (ste->ste_type != FunctionBlock || !ste->ste_unoptimized
|| !(ste->ste_free || ste->ste_child_free))
return 1;
trailer = (ste->ste_child_free ?
"contains a nested function with free variables" :
"is a nested function");
switch (ste->ste_unoptimized) {
case OPT_TOPLEVEL: /* exec / import * at top-level is fine */
case OPT_EXEC: /* qualified exec is fine */
return 1;
case OPT_IMPORT_STAR:
PyOS_snprintf(buf, sizeof(buf),
"import * is not allowed in function '%.100s' "
"because it is %s",
PyString_AS_STRING(ste->ste_name), trailer);
break;
case OPT_BARE_EXEC:
PyOS_snprintf(buf, sizeof(buf),
"unqualified exec is not allowed in function "
"'%.100s' it %s",
PyString_AS_STRING(ste->ste_name), trailer);
break;
default:
PyOS_snprintf(buf, sizeof(buf),
"function '%.100s' uses import * and bare exec, "
"which are illegal because it %s",
PyString_AS_STRING(ste->ste_name), trailer);
break;
}
PyErr_SetString(PyExc_SyntaxError, buf);
PyErr_SyntaxLocation(ste->ste_table->st_filename,
ste->ste_opt_lineno);
return 0;
}
--------------------------------------------------------------------------------------
Of course, this check can't be made at compile time if exec() is a function.
(You can even outsmart it currently by giving explicit None arguments to the
exec statement)
So my question is: is this check required, and can it be done at execution time
instead?
Comparing the exec code to execfile(), only this can be the cause for the
extra precaution:
(from Python/ceval.c, function exec_statement)
if (plain)
PyFrame_LocalsToFast(f, 0);
Georg
_______________________________________________
Python-3000 mailing list
[email protected]
http://mail.python.org/mailman/listinfo/python-3000
Unsubscribe:
http://mail.python.org/mailman/options/python-3000/archive%40mail-archive.com
