python - 如何将 csv 的嵌套 dict 格式的“_source”列展平为数据框
问题描述
我有一个包含 500 多行的 csv,其中一列“_source”存储为 JSON。我想将其提取到熊猫数据框中。我需要每个键都是它自己的列。#我有一个 1 mb 的在线社交媒体数据 Json 文件,我需要将字典和键值转换为它们自己的单独列。社交媒体数据来自 Facebook、Twitter/网络爬网...等。大约有 528 行单独的帖子/推文/文本,每行在字典中都有许多字典。我在下面附上了我的 Jupyter 笔记本中的几个步骤,以提供更完整的理解。需要将字典中字典的所有键值对转换为数据框内的列非常感谢,这将是一个巨大的帮助!!!我已经尝试通过这样做将其更改为数据框
source = pd.DataFrame.from_dict(source, orient='columns')
它返回类似这样的东西......我认为它可能会解压缩字典,但它没有。
#source.head()
#_source
#0 {'sub_organization_id': 'default', 'uid': 'aba...
#1 {'sub_organization_id': 'default', 'uid': 'ab0...
#2 {'sub_organization_id': 'default', 'uid': 'ac0...
下面是形状
#source.shape (528, 1)
下面是实际的“_source”行的样子。有许多字典和键:值对,其中每个键都需要是自己的列。谢谢!出于隐私原因,实际链接已被更改/加扰。
{'sub_organization_id': 'default',
'uid': 'ac0fafe9ba98327f2d0c72ddc365ffb76336czsa13280b',
'project_veid': 'default',
'campaign_id': 'default',
'organization_id': 'default',
'meta': {'rule_matcher': [{'atribs': {'website': 'github.com/res',
'source': 'Explicit',
'version': '1.1',
'type': 'crawl'},
'results': [{'rule_type': 'hashtag',
'rule_tag': 'Far',
'description': None,
'project_veid': 'A7180EA-7078-0C7F-ED5D-86AD7',
'campaign_id': '2A6DA0C-365BB-67DD-B05830920',
'value': '#Far',
'organization_id': None,
'sub_organization_id': None,
'appid': 'ray',
'project_id': 'CDE2F42-5B87-C594-C900E578C',
'rule_id': '1838',
'node_id': None,
'metadata': {'campaign_title': 'AF',
'project_title': 'AF '}}]}],
'render': [{'attribs': {'website': 'github.com/res',
'version': '1.0',
'type': 'Page Render'},
'results': [{'render_status': 'success',
'path': 'https://east.amanaws.com/rays-ime-store/renders/b/b/70f7dffb8b276f2977f8a13415f82c.jpeg',
'image_hash': 'bb7674b8ea3fc05bfd027a19815f82c',
'url': 'https://discooprdapp.com/',
'load_time': 32}]}]},
'norm_attribs': {'website': 'github.com/res',
'version': '1.1',
'type': 'crawl'},
'project_id': 'default',
'system_timestamp': '2019-02-22T19:04:53.569623',
'doc': {'appid': 'subtter',
'links': [],
'response_url': 'https://discooprdapp.com',
'url': 'https://discooprdapp.com/',
'status_code': 200,
'status_msg': 'OK',
'encoding': 'utf-8',
'attrs': {'uid': '2ab8f2651cb32261b911c990a8b'},
'timestamp': '2019-02-22T19:04:53.963',
'crawlid': '7fd95-785-4dd259-fcc-8752f'},
'type': 'crawl',
'norm': {'body': '\n',
'domain': 'discordapp.com',
'author': 'crawl',
'url': 'https://discooprdapp.com',
'timestamp': '2019-02-22T19:04:53.961283+00:00',
'id': '7fc5-685-4dd9-cc-8762f'}}
在您发布之前,请确保实际代码适用于附加的数据。谢谢!
我尝试了以下代码,但它没有工作,有一个我无法弄清楚的语法错误。
pd.io.json.json_normalize(source_data.[_source].apply(json.loads))
pd.io.json.json_normalize(source_data.[_source].apply(json.loads))
^
SyntaxError: invalid syntax
谁能帮助我,谁就是圣人!
解决方案
不久前我不得不做类似的事情。基本上,我使用了一个完全展平 json 的函数来识别将转换为列的键,然后遍历 json 以重建一行并将每一行附加到“结果”数据帧中。因此,使用您提供的数据,它创建了 52 列行并查看它,看起来它将所有键都包含在它自己的列中。任何嵌套的东西,例如:'meta': {'rule_matcher':[{'atribs': {'website': ...]}
然后应该有一个列名meta.rule_matcher.atribs.website
,其中'.'
表示那些嵌套的键
data_source = {'sub_organization_id': 'default',
'uid': 'ac0fafe9ba98327f2d0c72ddc365ffb76336czsa13280b',
'project_veid': 'default',
'campaign_id': 'default',
'organization_id': 'default',
'meta': {'rule_matcher': [{'atribs': {'website': 'github.com/res',
'source': 'Explicit',
'version': '1.1',
'type': 'crawl'},
'results': [{'rule_type': 'hashtag',
'rule_tag': 'Far',
'description': None,
'project_veid': 'A7180EA-7078-0C7F-ED5D-86AD7',
'campaign_id': '2A6DA0C-365BB-67DD-B05830920',
'value': '#Far',
'organization_id': None,
'sub_organization_id': None,
'appid': 'ray',
'project_id': 'CDE2F42-5B87-C594-C900E578C',
'rule_id': '1838',
'node_id': None,
'metadata': {'campaign_title': 'AF',
'project_title': 'AF '}}]}],
'render': [{'attribs': {'website': 'github.com/res',
'version': '1.0',
'type': 'Page Render'},
'results': [{'render_status': 'success',
'path': 'https://east.amanaws.com/rays-ime-store/renders/b/b/70f7dffb8b276f2977f8a13415f82c.jpeg',
'image_hash': 'bb7674b8ea3fc05bfd027a19815f82c',
'url': 'https://discooprdapp.com/',
'load_time': 32}]}]},
'norm_attribs': {'website': 'github.com/res',
'version': '1.1',
'type': 'crawl'},
'project_id': 'default',
'system_timestamp': '2019-02-22T19:04:53.569623',
'doc': {'appid': 'subtter',
'links': [],
'response_url': 'https://discooprdapp.com',
'url': 'https://discooprdapp.com/',
'status_code': 200,
'status_msg': 'OK',
'encoding': 'utf-8',
'attrs': {'uid': '2ab8f2651cb32261b911c990a8b'},
'timestamp': '2019-02-22T19:04:53.963',
'crawlid': '7fd95-785-4dd259-fcc-8752f'},
'type': 'crawl',
'norm': {'body': '\n',
'domain': 'discordapp.com',
'author': 'crawl',
'url': 'https://discooprdapp.com',
'timestamp': '2019-02-22T19:04:53.961283+00:00',
'id': '7fc5-685-4dd9-cc-8762f'}}
代码:
def flatten_json(y):
out = {}
def flatten(x, name=''):
if type(x) is dict:
for a in x:
flatten(x[a], name + a + '_')
elif type(x) is list:
i = 0
for a in x:
flatten(a, name + str(i) + '_')
i += 1
else:
out[name[:-1]] = x
flatten(y)
return out
flat = flatten_json(data_source)
import pandas as pd
import re
results = pd.DataFrame()
special_cols = []
columns_list = list(flat.keys())
for item in columns_list:
try:
row_idx = re.findall(r'\_(\d+)\_', item )[0]
except:
special_cols.append(item)
continue
column = re.findall(r'\_\d+\_(.*)', item )[0]
column = re.sub(r'\_\d+\_', '.', column)
row_idx = int(row_idx)
value = flat[item]
results.loc[row_idx, column] = value
for item in special_cols:
results[item] = flat[item]
输出:
print (results.to_string())
atribs_website atribs_source atribs_version atribs_type results.rule_type results.rule_tag results.description results.project_veid results.campaign_id results.value results.organization_id results.sub_organization_id results.appid results.project_id results.rule_id results.node_id results.metadata_campaign_title results.metadata_project_title attribs_website attribs_version attribs_type results.render_status results.path results.image_hash results.url results.load_time sub_organization_id uid project_veid campaign_id organization_id norm_attribs_website norm_attribs_version norm_attribs_type project_id system_timestamp doc_appid doc_response_url doc_url doc_status_code doc_status_msg doc_encoding doc_attrs_uid doc_timestamp doc_crawlid type norm_body norm_domain norm_author norm_url norm_timestamp norm_id
0 github.com/res Explicit 1.1 crawl hashtag Far NaN A7180EA-7078-0C7F-ED5D-86AD7 2A6DA0C-365BB-67DD-B05830920 #Far NaN NaN ray CDE2F42-5B87-C594-C900E578C 1838 NaN AF AF github.com/res 1.0 Page Render success https://east.amanaws.com/rays-ime-store/render... bb7674b8ea3fc05bfd027a19815f82c https://discooprdapp.com/ 32.0 default ac0fafe9ba98327f2d0c72ddc365ffb76336czsa13280b default default default github.com/res 1.1 crawl default 2019-02-22T19:04:53.569623 subtter https://discooprdapp.com https://discooprdapp.com/ 200 OK utf-8 2ab8f2651cb32261b911c990a8b 2019-02-22T19:04:53.963 7fd95-785-4dd259-fcc-8752f crawl \n discordapp.com crawl https://discooprdapp.com 2019-02-22T19:04:53.961283+00:00 7fc5-685-4dd9-cc-8762f
推荐阅读
- php - 使 PHP zipArchive 变慢
- mysql - 在特定行的众多列中查找最大值和最小值?
- java-8 - 用于下载 excel 的 Apachi poi 返回具有零字节的空文件异常
- c# - 是否可以将 Guid 与 Oracle 的实体框架核心包一起使用?
- c++ - 编译器错误:“对象的默认初始化...”使用 CodeLite
- java - 为什么我检查是否有括号等的方法不起作用?
- apache-spark - 如何从 org.apache.spark.sql.Column 中检索名称和类型?
- android - Android EditText,选择所有水平滚动到文本末尾。希望它滚动回文本的开头
- r - 将向量的元素与数据框中的条目进行比较
- c# - ssh.net sshclient 运行命令 grep