在熊猫数据框上按总和分组

df：

a b class
0 0  +
0 1  +
1 0  -

预期输出：

group each column based on class value and find the count, 
as in taking column 'a' as example the count should be 2 
for class '+' and 1 for class '-'

代码：

grp = df.groupby(df.class, axis=1).sum()

得到我什至不理解的错误。对熊猫来说是新手，任何建议都很棒。