pandas教程：Interacting with Web APIs API和数据库的交互

文章目录

[6.3 Interacting with Web APIs (网络相关的API交互)](#6.3 Interacting with Web APIs (网络相关的API交互))
[6.4 Interacting with Databases(与数据库的交互)](#6.4 Interacting with Databases(与数据库的交互))

6.3 Interacting with Web APIs (网络相关的API交互)

很多网站都有公开的API，通过JSON等格式提供数据流。有很多方法可以访问这些API，这里推荐一个易用的requests包。

找到github里pandas最新的30个issues，制作一个GET HTTP request, 通过使用requests包：

python 复制代码

import pandas as pd
import numpy as np

python 复制代码

import requests

python 复制代码

url = 'https://api.github.com/repos/pandas-dev/pandas/issues'

python 复制代码

resp = requests.get(url)

python 复制代码

resp

复制代码

<Response [200]>

response的json方法能返回一个dict，包含可以解析为python object的JSON：

python 复制代码

data = resp.json()
data[0]['title']

复制代码

'Optimize data type'

python 复制代码

data[0]

复制代码

{'assignee': None,
 'assignees': [],
 'author_association': 'NONE',
 'body': 'Hi guys, i\'m user of mysql\r\nwe have an "function" PROCEDURE ANALYSE\r\nhttps://dev.mysql.com/doc/refman/5.5/en/procedure-analyse.html\r\n\r\nit get all "dataframe" and show what\'s the best "dtype", could we do something like it in Pandas?\r\n\r\nthanks!',
 'closed_at': None,
 'comments': 1,
 'comments_url': 'https://api.github.com/repos/pandas-dev/pandas/issues/18272/comments',
 'created_at': '2017-11-13T22:51:32Z',
 'events_url': 'https://api.github.com/repos/pandas-dev/pandas/issues/18272/events',
 'html_url': 'https://github.com/pandas-dev/pandas/issues/18272',
 'id': 273606786,
 'labels': [],
 'labels_url': 'https://api.github.com/repos/pandas-dev/pandas/issues/18272/labels{/name}',
 'locked': False,
 'milestone': None,
 'number': 18272,
 'repository_url': 'https://api.github.com/repos/pandas-dev/pandas',
 'state': 'open',
 'title': 'Optimize data type',
 'updated_at': '2017-11-13T22:57:27Z',
 'url': 'https://api.github.com/repos/pandas-dev/pandas/issues/18272',
 'user': {'avatar_url': 'https://avatars0.githubusercontent.com/u/2468782?v=4',
  'events_url': 'https://api.github.com/users/rspadim/events{/privacy}',
  'followers_url': 'https://api.github.com/users/rspadim/followers',
  'following_url': 'https://api.github.com/users/rspadim/following{/other_user}',
  'gists_url': 'https://api.github.com/users/rspadim/gists{/gist_id}',
  'gravatar_id': '',
  'html_url': 'https://github.com/rspadim',
  'id': 2468782,
  'login': 'rspadim',
  'organizations_url': 'https://api.github.com/users/rspadim/orgs',
  'received_events_url': 'https://api.github.com/users/rspadim/received_events',
  'repos_url': 'https://api.github.com/users/rspadim/repos',
  'site_admin': False,
  'starred_url': 'https://api.github.com/users/rspadim/starred{/owner}{/repo}',
  'subscriptions_url': 'https://api.github.com/users/rspadim/subscriptions',
  'type': 'User',
  'url': 'https://api.github.com/users/rspadim'}}

data中的每一个元素都是一个dict，这个dict就是在github上找到的issue页面上的信息。我们可以把data传给DataFrame并提取感兴趣的部分：

python 复制代码

issues = pd.DataFrame(data, columns=['number', 'title', 
                                    'labels', 'state'])
issues

| | number | title | labels | state |
| 0 | 18272 | Optimize data type | [] | open |
| 1 | 18271 | BUG: Series.rank(pct=True).max() != 1 for a la... | [] | open |
| 2 | 18270 | (Series|DataFrame) datetimelike ops | [] | open |
| 3 | 18268 | DOC: update Series.combine/DataFrame.combine d... | [] | open |
| 4 | 18266 | DOC: updated .combine_first doc strings | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 5 | 18265 | Calling DataFrame.stack on an out-of-order col... | [] | open |
| 6 | 18264 | cleaned up imports | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 7 | 18263 | Tslibs offsets paramd | [] | open |
| 8 | 18262 | DEPR: let's deprecate | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 9 | 18258 | DEPR: deprecate (Sparse)Series.from_array | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 10 | 18255 | ENH/PERF: Add cache='infer' to to_datetime | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 11 | 18250 | Categorical.replace() unexpectedly returns non... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 12 | 18246 | pandas.MultiIndex.reorder_levels has no inplac... | [] | open |
| 13 | 18245 | TST: test tz-aware DatetimeIndex as separate m... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 14 | 18244 | RLS 0.21.1 | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 15 | 18243 | DEPR: deprecate .ftypes, get_ftype_counts | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 16 | 18242 | CLN: Remove days, seconds and microseconds pro... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 17 | 18241 | DEPS: drop 2.7 support | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 18 | 18238 | BUG: Fix filter method so that accepts byte an... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 19 | 18237 | Deprecate Series.asobject, Index.asobject, ren... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 20 | 18236 | df.plot() very slow compared to explicit matpl... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 21 | 18235 | Quarter.onOffset looks fishy | [] | open |
| 22 | 18231 | Reduce copying of input data on Series constru... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 23 | 18226 | Patch init to prevent passing invalid kwds | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 24 | 18222 | DataFrame.plot() produces incorrect legend lab... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 25 | 18220 | DataFrame.groupy renames columns when given a ... | [] | open |
| 26 | 18217 | Deprecate Index.summary | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 27 | 18216 | Pass kwargs from read_parquet() to the underly... | [{'url': 'https://api.github.com/repos/pandas-... | open |
| 28 | 18215 | DOC/DEPR: ensure that @deprecated functions ha... | [{'url': 'https://api.github.com/repos/pandas-... | open |

29	18213	Deprecate Series.from_array ?	[{'url': 'https://api.github.com/repos/pandas-...	open

6.4 Interacting with Databases(与数据库的交互)

如果在工作中，大部分数据并不会以text或excel的格式存储。最广泛使用的是SQL-based的关系型数据库（SQL Server，PostgreSQL，MySQL）。选择数据库通常取决于性能，数据整合性，实际应用的可扩展性。

读取SQL到DataFrame非常直观，pandas中有一些函数能简化这个过程。举个例子，这里创建一个SQLite数据库，通过使用python内建的sqlite3 driver：

python 复制代码

import sqlite3
import pandas as pd

python 复制代码

query = """
CREATE TABLE test
(a VARCHAR(20), b VARCHAR(20),
 c REAL,        d INTEGER
);"""

python 复制代码

con = sqlite3.connect('../examples/mydata.sqlite')

python 复制代码

con.execute(query)

复制代码

<sqlite3.Cursor at 0x1049931f0>

python 复制代码

con.commit()

然后我们插入几行数据：

python 复制代码

data = [('Atlanta', 'Georgia', 1.25, 6),
        ('Tallahassee', 'Florida', 2.6, 3),
        ('Sacramento', 'California', 1.7, 5)]

python 复制代码

stmt = "INSERT INTO test VALUES(?, ?, ?, ?)"

python 复制代码

con.executemany(stmt, data)

复制代码

<sqlite3.Cursor at 0x1049932d0>

python 复制代码

con.commit()

大部分python的SQL驱动(PyODBC, psycopg2, MySQLdb, pymssql, 等)返回a list of tuple，当从一个表格选择数据的时候：

python 复制代码

cursor = con.execute('select * from test')

python 复制代码

rows = cursor.fetchall()

python 复制代码

rows

复制代码

[('Atlanta', 'Georgia', 1.25, 6),
 ('Tallahassee', 'Florida', 2.6, 3),
 ('Sacramento', 'California', 1.7, 5)]

我们可以把list of tuples传递给DataFrame，但是我们也需要column names,包含cursor的description属性：

python 复制代码

cursor.description

复制代码

(('a', None, None, None, None, None, None),
 ('b', None, None, None, None, None, None),
 ('c', None, None, None, None, None, None),
 ('d', None, None, None, None, None, None))

python 复制代码

pd.DataFrame(rows, columns=[x[0] for x in cursor.description])

| | a | b | c | d |
| 0 | Atlanta | Georgia | 1.25 | 6 |
| 1 | Tallahassee | Florida | 2.60 | 3 |

2	Sacramento	California	1.70	5

我们不希望每次询问数据库的时候都重复以上步骤，这样对计算机很不好(逐步对计算机系统或文件做小改动导致大的损害)。SQLAlchemy计划是一个六星的Python SQL工具箱，它能抽象出不同SQL数据库之间的不同。pandas有一个read_sql函数，能让我们从SQLAlchemy connection从读取数据。这里我们用SQLAlchemy连接到同一个SQLite数据库，并从之前创建的表格读取数据：

python 复制代码

import sqlalchemy as sqla

python 复制代码

db = sqla.create_engine('sqlite:///../examples/mydata.sqlite')

python 复制代码

pd.read_sql('select * from test', db)

| | a | b | c | d |
| 0 | Atlanta | Georgia | 1.25 | 6 |
| 1 | Tallahassee | Florida | 2.60 | 3 |

2	Sacramento	California	1.70	5