URLとStringIOからの読み込み

ファイルをダウンロードせずにデータを読み込む

import pandas as pd

url = 'https://raw.githubusercontent.com/mwaskom/seaborn-data/master/tips.csv'
df = pd.read_csv(url)
print(df.shape)   # (244, 7)
print(df.head(2))

URL読み込みの内部動作

import pandas as pd

# Compressed CSV from URL -- auto-decompressed
url = 'https://example.com/data.csv.gz'
df = pd.read_csv(url, compression='infer')
print(df.shape)

認証用のリクエストヘッダーを追加する

import pandas as pd
import requests
import io

headers = {'Authorization': 'Bearer mytoken123'}
response = requests.get('https://api.example.com/export.csv',
                        headers=headers)
df = pd.read_csv(io.StringIO(response.text))
print(df.shape)

io.StringIOとは

import pandas as pd
import io

csv_text = 'name,score\nAlice,90\nBob,75\nCarol,88'
df = pd.read_csv(io.StringIO(csv_text))
print(df)
#     name  score
# 0  Alice     90
# 1    Bob     75
# 2  Carol     88

バイナリデータ用のio.BytesIO

import pandas as pd
import requests
import io

# Download an Excel file without saving to disk
response = requests.get('https://example.com/report.xlsx')
buffer = io.BytesIO(response.content)
df = pd.read_excel(buffer, sheet_name='Data')
print(df.shape)

StringIOを使ったテスト

import pandas as pd
import io

def test_clean_names():
    raw = 'name,age\n Alice ,30\nBob ,25'
    df = pd.read_csv(io.StringIO(raw))
    df['name'] = df['name'].str.strip()
    assert df['name'].tolist() == ['Alice', 'Bob']
    print('Test passed')

test_clean_names()

メモリ上のCSV用にStringIOへ書き込む

import pandas as pd
import io

df = pd.DataFrame({'x': [1, 2], 'y': [3, 4]})
buffer = io.StringIO()
df.to_csv(buffer, index=False)
csv_string = buffer.getvalue()
print(repr(csv_string))
# 'x,y\n1,3\n2,4\n'

Requests-Cacheでリモートデータをキャッシュする

import pandas as pd
import os

LOCAL = 'data/tips_cache.csv'
URL = 'https://raw.githubusercontent.com/mwaskom/seaborn-data/master/tips.csv'

if os.path.exists(LOCAL):
    df = pd.read_csv(LOCAL)
else:
    df = pd.read_csv(URL)
    df.to_csv(LOCAL, index=False)
print('Loaded:', df.shape)

大容量のリモートファイルをストリーミングする

import requests
import io
import pandas as pd

def stream_csv(url, chunksize=10000):
    with requests.get(url, stream=True) as r:
        content = r.content.decode('utf-8')
    for chunk in pd.read_csv(io.StringIO(content), chunksize=chunksize):
        yield chunk

複数のURLを並列に読み込む

import pandas as pd
from concurrent.futures import ThreadPoolExecutor

urls = [
    'https://example.com/data_jan.csv',
    'https://example.com/data_feb.csv',
]

with ThreadPoolExecutor(max_workers=4) as ex:
    dfs = list(ex.map(pd.read_csv, urls))
combined = pd.concat(dfs, ignore_index=True)
print(combined.shape)

リモートURLに関するセキュリティ上の考慮事項

import pandas as pd
import requests
import io

url = 'https://trusted-source.example.com/data.csv'
response = requests.get(url, timeout=30)  # 30-second timeout
response.raise_for_status()              # raise on HTTP error
df = pd.read_csv(io.StringIO(response.text), nrows=1000)
print(df.shape)

クイックチェック

レッスンのまとめ