熊猫Python更新数据框A的值（如果在数据框B中找到）

Question

I have two dataframes, Users and Devices. The users dataframe is a list of all interactions by user_id & timestamp. However, if someone uses our app as a guest, their user_id gets set to a device_id. If these guests eventually become members, we map their user_id to their device_id in the devices dataframe.

So we have for Users

user_id                                 timestamp
user13123                               2019-02-17
user224234                              2019-02-17
user32134234                            2019-02-17
00029AD9-X5X5-999N-807F-73F0EAE4A98B    2019-02-17

Where the final row is a guest user, with device id stored as user_id

Then for Devices

device_id                               user_id
00029AD9-X5X5-999N-807F-73F0EAE4A98B    user3423
37029BD9-D5D5-435D-837F-73F0EAE4A98B    user34423
...

Which is a simple mapping between device_ids and known user_ids

So what I want to do is check if Users.user_id matches to a Devices.device_id, and if so set Users.user_id to Devices.user_id. Basically, I want to update any old guest interactions to use the user_id if we have this information in Devices.

Messed around with it for a while and it was getting more and more convoluted, and feels like something that could be solved pretty cleanly in pandas. Any help is much appreciated.

Thanks!

Answer 1

Dataframes

In [32]: users
Out[32]:
                                user_id   timestamp
0                             user13123  2019-02-17
1                            user224234  2019-02-17
2                          user32134234  2019-02-17
3  00029AD9-D5D5-435D-807F-73F0EAE4A98B  2019-02-17

In []: devices
Out[]:
                              device_id    user_id
0  00029AD9-D5D5-435D-807F-73F0EAE4A98B   user3423
1  37029BD9-D5D5-435D-837F-73F0EAE4A98B  user34423

Compute a filter

All users for which user_id matches a device_id

In []: filtr = users.user_id.isin(devices.device_id)

In []: filtr
Out[]:
0    False
1    False
2    False
3     True
Name: user_id, dtype: bool

Substitute values

All filtered users' user_id are replaced with matching device's user_id right in dataframe users .

In []: users.loc[filtr, "user_id"] = users[filtr].user_id.map(devices.set_index("device_id").user_id)

In []: users
Out[]:
        user_id   timestamp
0     user13123  2019-02-17
1    user224234  2019-02-17
2  user32134234  2019-02-17
3      user3423  2019-02-17

---

Using np.where

Just another variation.

users.loc[:, 'user_id'] = pd.np.where(users.user_id.isin(devices.device_id),
                                      users.user_id.map(devices.set_index('device_id').user_id),
                                      users.user_id)

---

These solutions expect that only one user_id exists for each device_id .

Answer 2

left merge users with devices , fillna on the left-joined column user_id (it is name as user_id_y ). Finally, assign this back to column users.user_id

In [59]: users['user_id'] = users.merge(devices, how='left', left_on='user_id', right_on='device_id')['user_id_y'].fillna(users.user_id)

In [60]: users
Out[60]:
    timestamp       user_id
0  2019-02-17     user13123
1  2019-02-17    user224234
2  2019-02-17  user32134234
3  2019-02-17      user3423

Answer 3

This is a loop that will run though all the entries checking of a user_id matched a device_id and if so it will update the Users dataframe with the correct id.

for i in range(len(Users.index)):
    for p in range(len(Devices.index)):
        if(Users.loc[i,"user_id"] == Devices.loc[p,"device_id"]):
             # Fixed part of the code, check old version.
             Users.loc[i,"user_id"] = Devices.loc[p,"user_id"]

Answer 4

This solution finds the list of matching IDs then loops through once and updates the user_ids using it as the index.

devices = pd.DataFrame({'device_id':{0:'00029AD9-X5X5-999N-807F-73F0EAE4A98B',1:'37029BD9-D5D5-435D-837F-73F0EAE4A98B'},'user_id':{0:'user3423',1:'user34423'}})
users = pd.DataFrame({'user_id':{0:'user13123',1:'user224234',2:'user32134234',3:'00029AD9-X5X5-999N-807F-73F0EAE4A98B'},'timestamp':{0:'2019-02-17',1:'2019-02-17',2:'2019-02-17',3:'2019-02-17'}})

matching_ids = list(set(users.user_id).intersection(set(devices.device_id)))
for id in matching_ids:
    users.loc[users.user_id == id, 'user_id'] = devices.set_index('device_id').at[id, 'user_id']