从numpy ndarray提取特定数据

Question

对熊猫来说，任何帮助都是非常感激的

def csv_reader(fileName):
    reqcols=['_id__$oid','payload','channel']
    io = pd.read_csv(fileName,sep=",",usecols=reqcols)
    print(io['payload'].values)
    return io  

io输出行['payload']：

{
    "destination_ip": "172.31.14.66",
    "date": "2014-10-19T01:32:36.669861",
    "classification": "Potentially Bad Traffic",
    "proto": "UDP",
    "source_ip": "172.31.0.2",
    "priority": "`2",
    "header": "1:2003195:5",
    "signature": "ET POLICY Unusual number of DNS No Such Name Responses ",
    "source_port": "53",
    "destination_port": "34638",
    "sensor": "5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"
}

我试图从ndarray对象中提取特定数据。从数据帧中提取数据的方法是什么

"destination_ip": "172.31.13.124",
"proto": "ICMP",
"source_ip": "201.158.32.1",
"date": "2014-09-28T14:49:43.391463",
"sensor": "139cfdf2-471e-11e4-9ee4-0a0b6e7c3e9e"

Answer 1

使用@jezrael的sample df

d = {'_id__$oid': ['542f8', '542f8', '542f8'], 'channel': ['snort_alert', 'snort_alert', 'snort_alert'], 'payload': ['{"destination_ip":"172.31.14.66","date": "2014-10-19T01:32:36.669861","classification":"Potentially Bad Traffic","proto":"UDP","source_ip":"172.31.0.2","priority":"2","header":"1:2003195:5","signature":"ET POLICY Unusual number of DNS No Such Name Responses ","source_port":"53","destination_port":"34638","sensor":"5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"}', '{"destination_ip":"172.31.14.66","date": "2014-10-19T01:32:36.669861","classification":"Potentially Bad Traffic","proto":"UDP","source_ip":"172.31.0.2","priority":"2","header":"1:2003195:5","signature":"ET POLICY Unusual number of DNS No Such Name Responses ","source_port":"53","destination_port":"34638","sensor":"5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"}', '{"destination_ip":"172.31.14.66","date": "2014-10-19T01:32:36.669861","classification":"Potentially Bad Traffic","proto":"UDP","source_ip":"172.31.0.2","priority":"2","header":"1:2003195:5","signature":"ET POLICY Unusual number of DNS No Such Name Responses ","source_port":"53","destination_port":"34638","sensor":"5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"}']}
df = pd.DataFrame(d)

解决方案

• 用矢量化的str.cat将所有payload粉碎在一起
• 用pd.read_json一次解析整个过程

---

cols = 'destination_ip proto source_ip date sensor'.split()
df.drop(
    'payload', 1
).join(
    pd.read_json('[{}]'.format(df.payload.str.cat(sep=',')))[cols]
)

Answer 2

我认为您首先需要将string的dicts重新表示转换为dictionaries在每一行中的json.loads或ast.literal_eval在payload列中，然后通过构造函数创建新的DataFrame，通过子集过滤列，如果需要，通过^{}添加原始列：

d = {'_id__$oid': ['542f8', '542f8', '542f8'], 'channel': ['snort_alert', 'snort_alert', 'snort_alert'], 'payload': ['{"destination_ip":"172.31.14.66","date": "2014-10-19T01:32:36.669861","classification":"Potentially Bad Traffic","proto":"UDP","source_ip":"172.31.0.2","priority":"2","header":"1:2003195:5","signature":"ET POLICY Unusual number of DNS No Such Name Responses ","source_port":"53","destination_port":"34638","sensor":"5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"}', '{"destination_ip":"172.31.14.66","date": "2014-10-19T01:32:36.669861","classification":"Potentially Bad Traffic","proto":"UDP","source_ip":"172.31.0.2","priority":"2","header":"1:2003195:5","signature":"ET POLICY Unusual number of DNS No Such Name Responses ","source_port":"53","destination_port":"34638","sensor":"5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"}', '{"destination_ip":"172.31.14.66","date": "2014-10-19T01:32:36.669861","classification":"Potentially Bad Traffic","proto":"UDP","source_ip":"172.31.0.2","priority":"2","header":"1:2003195:5","signature":"ET POLICY Unusual number of DNS No Such Name Responses ","source_port":"53","destination_port":"34638","sensor":"5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"}']}
reqcols=['_id__$oid','payload','channel']
df = pd.DataFrame(d)
print (df)
  _id__$oid      channel                                            payload
0     542f8  snort_alert  {"destination_ip":"172.31.14.66","date": "2014...
1     542f8  snort_alert  {"destination_ip":"172.31.14.66","date": "2014...
2     542f8  snort_alert  {"destination_ip":"172.31.14.66","date": "2014...

---

import json
import ast
df.payload = df.payload.apply(json.loads)
#another slowier solution
#df.payload = df.payload.apply(ast.literal_eval)

required = ["destination_ip", "proto", "source_ip", "date", "sensor"]
df1 = pd.DataFrame(df.payload.values.tolist())[required]
print (df1)
  destination_ip proto   source_ip                        date  \
0   172.31.14.66   UDP  172.31.0.2  2014-10-19T01:32:36.669861   
1   172.31.14.66   UDP  172.31.0.2  2014-10-19T01:32:36.669861   
2   172.31.14.66   UDP  172.31.0.2  2014-10-19T01:32:36.669861   

                                 sensor  
0  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  
1  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  
2  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  

df2 = pd.concat([df[['_id__$oid','channel']], df1], axis=1)
print (df2)
  _id__$oid      channel destination_ip proto   source_ip  \
0     542f8  snort_alert   172.31.14.66   UDP  172.31.0.2   
1     542f8  snort_alert   172.31.14.66   UDP  172.31.0.2   
2     542f8  snort_alert   172.31.14.66   UDP  172.31.0.2   

                         date                                sensor  
0  2014-10-19T01:32:36.669861  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  
1  2014-10-19T01:32:36.669861  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  
2  2014-10-19T01:32:36.669861  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  

时间安排：

#[30000 rows x 3 columns]
df = pd.concat([df]*10000).reset_index(drop=True)
print (df)

In [38]: %timeit pd.DataFrame(df.payload.apply(json.loads).values.tolist())[required]
1 loop, best of 3: 379 ms per loop

In [39]: %timeit pd.read_json('[{}]'.format(df.payload.str.cat(sep=',')))[required]
1 loop, best of 3: 528 ms per loop

In [40]: %timeit pd.DataFrame(df.payload.apply(ast.literal_eval).values.tolist())[required]
1 loop, best of 3: 1.98 s per loop

Answer 3

访问pandas中的列是相当直接的。只需传递所需列的列表：

代码：

columns = ["destination_ip", "proto", "source_ip", "date", "sensor"]
extracted_data = df[columns]

测试代码：

data = {
    "destination_ip": "172.31.14.66",
    "date": "2014-10-19T01:32:36.669861",
    "classification": "Potentially Bad Traffic",
    "proto": "UDP",
    "source_ip": "172.31.0.2",
    "priority": "`2",
    "header": "1:2003195:5",
    "signature": "ET POLICY Unusual number of DNS No Such Name Responses ",
    "source_port": "53",
    "destination_port": "34638",
    "sensor": "5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e"
}
df = pd.DataFrame([data, data])

columns = ["destination_ip", "proto", "source_ip", "date", "sensor"]
print(df[columns])

结果：

  destination_ip proto   source_ip                        date  \
0   172.31.14.66   UDP  172.31.0.2  2014-10-19T01:32:36.669861   
1   172.31.14.66   UDP  172.31.0.2  2014-10-19T01:32:36.669861   

                                 sensor  
0  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e  
1  5cda4a12-4730-11e4-9ee4-0a0b6e7c3e9e