如何根据字段合并两个 CSV 文件并在每条记录上保留相同数量的属性？

我正在尝试根据每个文件中的特定字段合并两个 CSV 文件。

文件1.csv

id,attr1,attr2,attr3
1,True,7,"Purple"
2,False,19.8,"Cucumber"
3,False,-0.5,"A string with a comma, because it has one"
4,True,2,"Nope"
5,True,4.0,"Tuesday"
6,False,1,"Failure"

文件2.csv

id,attr4,attr5,attr6
2,"python",500000.12,False
5,"program",3,True
3,"Another string",-5,False

这是我正在使用的代码：

import csv
from collections import OrderedDict

with open('file2.csv','r') as f2:
    reader = csv.reader(f2)
    fields2 = next(reader,None) # Skip headers
    dict2 = {row[0]: row[1:] for row in reader}

with open('file1.csv','r') as f1:
    reader = csv.reader(f1)
    fields1 = next(reader,None) # Skip headers
    dict1 = OrderedDict((row[0], row[1:]) for row in reader)

result = OrderedDict()
for d in (dict1, dict2):
    for key, value in d.iteritems():
        result.setdefault(key, []).extend(value)

with open('merged.csv', 'wb') as f:
    w = csv.writer(f)
    for key, value in result.iteritems():
        w.writerow([key] + value)

我得到这样的输出，它适当地合并，但所有行的属性数量不同：

1,True,7,Purple
2,False,19.8,Cucumber,python,500000.12,False
3,False,-0.5,"A string with a comma, because it has one",Another string,-5,False
4,True,2,Nope
5,True,4.0,Tuesday,program,3,True
6,False,1,Failure

file2不会有每个记录id in file1。我希望输出有空字段file2在合并的文件中。例如，id1 看起来像这样：

1,True,7,Purple,,,

如何将空字段添加到没有数据的记录中file2这样合并后的 CSV 中的所有记录都具有相同数量的属性吗？

如果我们不使用pandas，我会重构为类似的东西

import csv
from collections import OrderedDict

filenames = "file1.csv", "file2.csv"
data = OrderedDict()
fieldnames = []
for filename in filenames:
    with open(filename, "rb") as fp: # python 2
        reader = csv.DictReader(fp)
        fieldnames.extend(reader.fieldnames)
        for row in reader:
            data.setdefault(row["id"], {}).update(row)

fieldnames = list(OrderedDict.fromkeys(fieldnames))
with open("merged.csv", "wb") as fp:
    writer = csv.writer(fp)
    writer.writerow(fieldnames)
    for row in data.itervalues():
        writer.writerow([row.get(field, '') for field in fieldnames])

这使

id,attr1,attr2,attr3,attr4,attr5,attr6
1,True,7,Purple,,,
2,False,19.8,Cucumber,python,500000.12,False
3,False,-0.5,"A string with a comma, because it has one",Another string,-5,False
4,True,2,Nope,,,
5,True,4.0,Tuesday,program,3,True
6,False,1,Failure,,,

为了比较，将pandas相当于

df1 = pd.read_csv("file1.csv")
df2 = pd.read_csv("file2.csv")
merged = df1.merge(df2, on="id", how="outer").fillna("")
merged.to_csv("merged.csv", index=False)

在我看来，这要简单得多，意味着您可以花更多的时间处理数据，而减少重新发明轮子的时间。