Skip to content
ClickHouse Docs
ClickHouse DocsClickHouse Docs
49.6kGet Started

DataStore クイックスタート

View as MarkdownEdit this pageReport an issueSet up docs MCP

DataStore を数分で使い始められます。このガイドでは、インストール、pandas からの移行、基本的な使い方を紹介します。

インストール

pip を使用して chDB をインストールします。

pip install "chdb>=4.0"

オプションの依存関係について:

# For pandas DataFrame support
pip install "chdb[pandas]>=4.0"

# For PyArrow support
pip install "chdb[arrow]>=4.0"

# All optional dependencies
pip install "chdb[all]>=4.0"

インストールの確認

import chdb

print(chdb.__version__)  # Should print 4.x.x or higher

from chdb import datastore as pd
print("DataStore ready!")

Pandas からのワンライナーでの移行

DataStore を使い始めるには、インポート文を変更するのが最も簡単です。

# Before (pandas)
import pandas as pd

# After (DataStore)
from chdb import datastore as pd

これで完了です！既存の pandas コードは DataStore を使うようになり、SQL 最適化のメリットを得られます。

移行例

from pathlib import Path
Path("employees.csv").write_text("""\
name,age,city,salary,department,dept_id,status,email
Alice,28,NYC,75000,Engineering,1,active,alice@company.com
Bob,35,LA,85000,Engineering,1,active,bob@company.com
Charlie,52,NYC,95000,Product,2,active,charlie@company.com
Diana,32,SF,70000,Design,3,active,diana@company.com
Eve,23,LA,48000,Product,2,inactive,eve@company.com
""")

# Original pandas code
import pandas as pd

df = pd.read_csv("employees.csv")
result = (df[df['salary'] > 50000]
          .groupby('department')['salary']
          .agg(['mean', 'count'])
          .sort_values('mean', ascending=False))
print(result)

# DataStore version - just change the import!
from chdb import datastore as pd

df = pd.read_csv("employees.csv")
result = (df[df['salary'] > 50000]
          .groupby('department')['salary']
          .agg(['mean', 'count'])
          .sort_values('mean', ascending=False))
print(result)  # Same result, faster execution!

基本的な使い方

DataStore の作成

from chdb import datastore as pd

# From a dictionary
ds = pd.DataFrame({
    'name': ['Alice', 'Bob', 'Charlie'],
    'age': [25, 30, 35],
    'city': ['NYC', 'LA', 'NYC']
})

# From a pandas DataFrame
import pandas

pdf = pandas.DataFrame({'a': [1, 2, 3], 'b': [4, 5, 6]})
ds = pd.DataFrame(pdf)

# From a CSV file
ds = pd.read_csv("data.csv")

# From a Parquet file (recommended for large datasets)
ds = pd.read_parquet("data.parquet")

データのフィルタリング

from chdb import datastore as pd

ds = pd.read_csv("employees.csv")

# Single condition
senior = ds[ds['age'] > 30]

# Multiple conditions (AND)
senior_nyc = ds[(ds['age'] > 30) & (ds['city'] == 'NYC')]

# Multiple conditions (OR)
young_or_senior = ds[(ds['age'] < 25) | (ds['age'] > 50)]

# Using filter method (SQL-style)
result = ds.filter(ds['salary'] > 50000)

カラムの選択

# Pandas style
subset = ds[['name', 'age']]

# SQL style
subset = ds.select('name', 'age')

ソート

# Pandas style
sorted_ds = ds.sort_values('salary', ascending=False)

# SQL style
sorted_ds = ds.sort('salary', ascending=False)

グループ化と集計

from pathlib import Path
Path("sales.csv").write_text("""\
region,product,category,amount,quantity,price,date,order_id
East,Widget,Electronics,5200,10,120,2024-01-15,1001
West,Gadget,Electronics,800,5,160,2024-02-20,1002
East,Gizmo,Home,6500,3,100,2024-03-10,1003
North,Widget,Electronics,4500,6,150,2024-06-18,1004
West,Gadget,Electronics,2000,8,250,2024-09-14,1005
""")

from chdb import datastore as pd

ds = pd.read_csv("sales.csv")

# Group by single column
by_region = ds.groupby('region')['amount'].sum()

# Group by multiple columns
by_region_product = ds.groupby(['region', 'product']).agg({
    'amount': ['sum', 'mean'],
    'quantity': 'sum'
})

# Multiple aggregations
summary = ds.groupby('category').agg({
    'price': ['min', 'max', 'mean'],
    'quantity': 'sum'
})

DataStoreの結合

from pathlib import Path
Path("departments.csv").write_text("""\
dept_id,department_name
1,Engineering
2,Product
3,Design
""")

from chdb import datastore as pd

employees = pd.read_csv("employees.csv")
departments = pd.read_csv("departments.csv")

# Inner join
result = employees.join(departments, on='dept_id', how='inner')

# Left join
result = employees.join(departments, on='dept_id', how='left')

# Using merge (pandas style)
result = pd.merge(employees, departments, on='dept_id')

結果を取得する

DataStore は遅延評価を採用しているため、結果が必要になるまで操作は実行されません。

実行をトリガーする

# Automatic triggers
print(ds)           # Displaying results
len(ds)             # Getting row count
ds.columns          # Accessing properties
list(ds)            # Converting to list

# Explicit conversion
df = ds.to_df()     # Convert to pandas DataFrame
df = ds.to_pandas() # Same as to_df()

生成された SQL の表示

Querypython
# See what SQL DataStore will execute
query = ds.filter(ds['age'] > 25).groupby('city').agg({'salary': 'mean'})
print(query.to_sql())
Responsesql
SELECT city, AVG(salary) AS mean
FROM file('data.csv', 'CSVWithNames')
WHERE age > 25
GROUP BY city

さまざまなデータソースを扱う

ローカルファイル

from chdb import datastore as pd

# CSV
ds = pd.read_csv("data.csv")

# Parquet (best performance)
ds = pd.read_parquet("data.parquet")

# JSON
ds = pd.read_json("data.json")

Cloud ストレージ

from chdb.datastore import DataStore

# S3 (anonymous)
ds = DataStore.uri("s3://bucket/data.parquet?nosign=true")

# S3 (with credentials)
ds = DataStore.from_s3(
    "s3://bucket/data.parquet",
    access_key_id="KEY",
    secret_access_key="SECRET"
)

# HTTP/HTTPS
ds = DataStore.uri("https://example.com/data.csv")

データベース

from chdb.datastore import DataStore

# MySQL
ds = DataStore.from_mysql(
    host="localhost",
    database="mydb",
    table="users",
    user="root",
    password="pass"
)

# PostgreSQL
ds = DataStore.from_postgresql(
    host="localhost",
    database="mydb",
    table="users",
    user="postgres",
    password="pass"
)

# Using URI
ds = DataStore.uri("mysql://user:pass@localhost:3306/mydb/users")

String および DateTime の操作

文字列操作

# All pandas .str methods work
ds['name_upper'] = ds['name'].str.upper()
ds['name_len'] = ds['name'].str.len()
ds['has_a'] = ds['name'].str.contains('a')

DateTime の操作

# All pandas .dt methods work
ds['year'] = ds['date'].dt.year
ds['month'] = ds['date'].dt.month
ds['day_of_week'] = ds['date'].dt.dayofweek

ClickHouse 拡張機能

# URL parsing (not available in pandas!)
ds['domain'] = ds['url'].url.domain()

# JSON extraction
ds['user_name'] = ds['json_data'].json.get_string('name')

# IP address operations
ds['is_ipv4'] = ds['ip_addr'].ip.is_ipv4_string()

ベストプラクティス

1. 大容量ファイルにはParquetを使用する

# CSV - slower, reads entire file
ds = pd.read_csv("large_data.csv")

# Parquet - faster, columnar format, reads only needed columns
ds = pd.read_parquet("large_data.parquet")

2. 早い段階でフィルタする

# Good - filter first, then aggregate
result = (ds
    .filter(ds['date'] >= '2024-01-01')
    .groupby('category')['amount'].sum()
)

# Less optimal - aggregate first
result = ds.groupby('category')['amount'].sum()

3. 必要なカラムのみを選択する

# Good - select specific columns
result = ds.select('name', 'age', 'city').filter(ds['age'] > 25)

# Less optimal - work with all columns
result = ds.filter(ds['age'] > 25)

4. 複雑な処理にはSQLを使用する

# For complex queries, use SQL directly
ds = DataStore()
result = ds.sql("""
    SELECT category, 
           SUM(amount) as total,
           COUNT(*) as count,
           AVG(amount) as avg
    FROM file('sales.csv', 'CSVWithNames')
    WHERE date >= '2024-01-01'
    GROUP BY category
    HAVING total > 10000
    ORDER BY total DESC
    LIMIT 10
""")

次のステップ

  • DataStore を作成するための各種 Factory Methods について学ぶ
  • SQL スタイルの操作については Query Building を参照する
  • 文字列や datetime などを扱う Accessors を確認する
  • 最適化のヒントについては Performance Guide を参照する
Navigation