Question

我在Suse Enterprise Linux 11 w / python 2.7.9上使用pandas 0.18。我有两张桌子，A和B.

A包含以下列和类型：

>>> print a.dtypes
cid     object
bid     int64
li      object
lit     int64
x1      float64
y1      float64
x2      float64
y2      float64
hit_num object

B包含以下列和类型：

>>> print b.dtypes
    cid     object
    li      object
    x1      float64
    y1      float64
    x2      float64
    y2      float64
    hit_num object

现在这里是A：

的示例数据集

cid,bid,li,lit,x1,y1,x2,y2,hit_num
id1,0,m0,1,6775.5711,6102.5771,6775.6051,6102.7731,
id1,0,m0,2,6775.5311,6103.0631,6775.5531,6103.2051,
id1,0,m0,3,6775.6231,6103.0631,6775.6451,6103.2051,
id1,0,m0,4,6775.1631,6103.6571,6775.1971,6103.7451,

现在这里是B的示例数据集：

cid,li,x1,y1,x2,y2,hit_num    
id1,m0,6775.1631,6103.6571,6775.1971,6103.7451,hello
id1,m0,6775.6231,6103.0631,6775.6451,6103.2051,world
id1,m0,6775.5311,6103.0631,6775.5531,6103.2051,gotta
id1,m0,6775.5711,6102.5771,6775.6051,6102.7731,go

我做 A.update（B）。所以我期待B [hit_num]通过对齐cid，lid，x1，y1，x2，y2列来更新A [hit_num]。

所以我希望这样的事情（除非我对update（）的理解是错误的？）：

cid,bid,li,lit,x1,y1,x2,y2,hit_num
id1,0,m0,1,6775.5711,6102.5771,6775.6051,6102.7731,0.018,0.02,0.0269,go
id1,0,m0,2,6775.5311,6103.0631,6775.5531,6103.2051,0.018,0.02,0.0269,gotta
id1,0,m0,3,6775.6231,6103.0631,6775.6451,6103.2051,0.018,0.02,0.0269,world
id1,0,m0,4,6775.1631,6103.6571,6775.1971,6103.7451,0.018,0.02,0.0269,hello

然而，我最终得到了以下内容。 “点亮”列（以粗体突出显示）似乎搞砸了，并且存在重复的“1”条目。这不存在于A.我想知道为什么会这样。我创建了一个小例子并尝试重现该问题，但未成功。我在那里得到预期的结果。

但是，在我正在运行回归的较大表中，我看到了这种行为。我打印了表A，表B和A.update（B），我看到了下面的内容。我之间没有调用任何其他数据帧操作。即，伪代码：

print v['les_tables']['foo']
print overlay_tables['foo']
v['les_tables']['foo'].update(overlay_tables['foo'])
print v['les_tables']['foo']

我不完全确定更新如何工作，但我认为它使用某种类型的相等运算符来匹配列？如果是这样，x1，y1，x2，y2是否会引起任何问题？我有什么想法吗？

我已经确认要对齐的列在两个A / B中都是相同的名称/类型（参见上面的A.dtypes / B.dtypes）。

CID，出价，李，点燃，X1，Y1，X2，Y2，hit_num id1,0，M0， 1 下，6775.5711,6102.5771,6775.6051,6102.7731,0.018,0.02,0.0269，去 id1,0，M0， 3 下，6775.5311,6103.0631,6775.5531,6103.2051,0.018,0.02,0.0269，总得 id1,0，M0， 2 下，6775.6231,6103.0631,6775.6451,6103.2051,0.018,0.02,0.0269，世界 id1,0，M0， 1 下，6775.1631,6103.6571,6775.1971,6103.7451,0.018,0.02,0.0269，你好

Answer 1

试试这个：

In [73]: df = (A.set_index(['cid','li','x1','y1','x2','y2'])
   ....:        .drop(['hit_num'], axis=1)
   ....:        .join(B.set_index(['cid','li','x1','y1','x2','y2']))
   ....:        .reset_index()
   ....:      )

In [74]: df
Out[74]:
   cid  li         x1         y1         x2         y2  bid  lit hit_num
0  id1  m0  6775.5711  6102.5771  6775.6051  6102.7731    0    1      go
1  id1  m0  6775.5311  6103.0631  6775.5531  6103.2051    0    2   gotta
2  id1  m0  6775.6231  6103.0631  6775.6451  6103.2051    0    3   world
3  id1  m0  6775.1631  6103.6571  6775.1971  6103.7451    0    4   hello

Pandas：Update（）数据帧问题

1 个答案: