将具有不同键的字典列表转换为数据框

Question

How do I get a list of dictionaries converted into a dataframe whose columns are 'Event', 'Id', 'Name' ?

sample = [{'event': 'up', '53118': 'Harry'},
                  {'event': 'up', '51880': 'Smith'}, 
                  {'event': 'down', '51659': 'Joe'}, 
                  {'52983': 'Sam', 'event': 'up'}, 
                  {'event': 'down', '52917': 'Roger'},
                  {'event': 'up', '314615': 'Julie'},
                  {'event': 'left', '276298': 'Andrew'},
                  {'event': 'right', '457249': 'Carlos'}, 
                  {'event': 'down', '391485': 'Jason'},
                  {'event': 'right', '53191': 'Taylor'}, 
                  {'51248': 'Benjy', 'event': 'down'}]

pd.DataFrame(sample) would return;

Is there a pythonic panda-ic way to convert it to this form?

Event   Id      Name
up     53118    Harry
up     51880    Smith
down   51659    Joe

Answer 1

pd.melt can get you most of the way, starting from your df = pd.DataFrame(sample) :

In [74]: m = pd.melt(df, id_vars="event", var_name="Id", value_name="Name").dropna()

In [75]: m
Out[75]: 
     event      Id    Name
6     left  276298  Andrew
16      up  314615   Julie
30    down  391485   Jason
40   right  457249  Carlos
54    down   51248   Benjy
57    down   51659     Joe
67      up   51880   Smith
81    down   52917   Roger
91      up   52983     Sam
99      up   53118   Harry
119  right   53191  Taylor

And then you can do some cleanup ( reset_index(drop=True) , rename(columns={"event": "Event"}) , convert Id to integers, etc.)

Since @eumiro makes a good point, we could also implement @MattDMo's suggestion easily enough:

In [90]: sample = [dict(event=d.pop("event"), id=min(d), name=min(d.values())) for d in sample]

In [91]: pd.DataFrame(sample)
Out[91]: 
    event      id    name
0      up   53118   Harry
1      up   51880   Smith
2    down   51659     Joe
3      up   52983     Sam
4    down   52917   Roger
5      up  314615   Julie
6    left  276298  Andrew
7   right  457249  Carlos
8    down  391485   Jason
9   right   53191  Taylor
10   down   51248   Benjy

Here I've taken advantage of the fact that once we pop event there's only one element in the dictionary left, but a more manual loop would work as easily.

Answer 2

You need to adjust your dicts, so that instead of having:

{'event': 'up', '53118': 'Harry'}

you have:

{'event': 'up', 'id': '53118', 'name': 'Harry'}

resulting in:

In [23]: df = pd.DataFrame(sample)

In [24]: df
Out[24]: 
    event      id    name
0      up   53118   Harry
1      up   51880   Smith
2    down   51659     Joe
3      up   52983     Sam
4    down   52917   Roger
5      up  314615   Julie
6    left  276298  Andrew
7   right  457249  Carlos
8    down  391485   Jason
9   right   53191  Taylor
10   down   51248   Benjy