如何基于字段合并两个CSV文件并在每个记录上保留相同数量的属性？

Question

我正试图合并两个CSV文件基于每个文件中的特定字段。

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，它可以适当地合并，但并非所有行都有相同数量的属性：

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不会为file1中的每个id都有记录。我希望输出在合并文件中有来自file2的空字段。例如，id1如下所示：

1,True,7,Purple,,,

如何将空字段添加到在file2中没有数据的记录中，以便合并CSV中的所有记录都具有相同数量的属性？

Answer 1

您可以使用^{}来执行此操作：

import pandas

csv1 = pandas.read_csv('filea1.csv')
csv2 = pandas.read_csv('file2.csv')
merged = csv1.merge(csv2, on='id')
merged.to_csv("output.csv", index=False)

我还没有测试过这个，但在我测试之前，它应该会让你走上正轨。这段代码很简单；首先导入pandas库，以便使用它。然后使用pandas.read_csv读取2个csv文件并使用merge方法合并它们。on参数指定哪个列应用作“键”。最后，合并的csv被写入output.csv。

Answer 2

如果我们不使用pandas，我将重构为

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])

它给予

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,,,

作为比较，等价的pandas类似于

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)

这对我来说简单得多，意味着你可以花更多的时间处理你的数据，更少的时间重新发明轮子。

Answer 3

使用dict of dict，然后更新它。像这样：

import csv
from collections import OrderedDict

with open('file2.csv','r') as f2:
    reader = csv.reader(f2)
    lines2 = list(reader)

with open('file1.csv','r') as f1:
    reader = csv.reader(f1)
    lines1 = list(reader)

dict1 = {row[0]: dict(zip(lines1[0][1:], row[1:])) for row in lines1[1:]}
dict2 = {row[0]: dict(zip(lines2[0][1:], row[1:])) for row in lines2[1:]}

#merge
updatedDict = OrderedDict()
mergedAttrs = OrderedDict.fromkeys(lines1[0][1:] + lines2[0][1:], "?")
for id, attrs in dict1.iteritems():
    d = mergedAttrs.copy()
    d.update(attrs)
    updatedDict[id] = d

for id, attrs in dict2.iteritems():
    updatedDict[id].update(attrs)

#out
with open('merged.csv', 'wb') as f:
    w = csv.writer(f)
    for id, rest in sorted(updatedDict.iteritems()):
        w.writerow([id] + rest.values())