Is there a way to detect anonymous enumerations using libclang without relying on the text in the spelling name?
The python bindings to libclang include functionality to detect whether C/C++ structs or unions are anonymous using clang.cindex.Cursor.is_anonymous, which ends up calling clang_Cursor_isAnonymous.
The following sample demonstrates the issue.
import sys
from clang.cindex import *
def nodeinfo(n):
return (n.kind, n.is_anonymous(), n.spelling, n.type.spelling)
idx = Index.create()
# translation unit parsed correctly
tu = idx.parse(sys.argv[1], ['-std=c++11'])
assert(len(tu.diagnostics) == 0)
for n in tu.cursor.walk_preorder():
if n.kind == CursorKind.STRUCT_DECL and n.is_anonymous():
print nodeinfo(n)
if n.kind == CursorKind.UNION_DECL and n.is_anonymous():
print nodeinfo(n)
if n.kind == CursorKind.ENUM_DECL:
if n.is_anonymous():
print nodeinfo(n)
else:
print 'INCORRECT', nodeinfo(n)
Which when run on sample.cpp
enum
{
VAL = 1
};
struct s
{
struct {};
union
{
int x;
float y;
};
};
Gives:
INCORRECT (CursorKind.ENUM_DECL, False, '', '(anonymous enum at sample1.cpp:1:1)')
(CursorKind.STRUCT_DECL, True, '', 's::(anonymous struct at sample1.cpp:8:5)')
(CursorKind.UNION_DECL, True, '', 's::(anonymous union at sample1.cpp:9:5)')