[Python] Reading and Writing CSV/TSV Files with Pandas (read_csv / to_csv)

2026年1月4日

In practical data analysis, text files like CSV (Comma-Separated Values) and TSV (Tab-Separated Values) are the most widely used formats for exchanging data. Pandas provides features to load these files into a DataFrame with very simple code, as well as to save DataFrames as files.

This article explains flexible data reading methods using the read_csv function and file output control using the to_csv method, provided with specific code examples.

Reading CSV/TSV Files (read_csv)

Pandas’ pd.read_csv() is a function for loading data from local files or URLs. While it assumes comma-separated files by default, you can handle tab-separated files (TSV) or files without headers by changing the arguments.

Key Arguments

filepath_or_buffer: The file path or URL.
sep: The delimiter character. The default is a comma ,. Specify \t for TSV.
header: The position of the header row. Specify None if there is no header.
dtype: Specifies the data type for each column using a dictionary. This is effective for saving memory and preventing type misidentification.

Writing to CSV/TSV Files (to_csv)

To save the contents of a DataFrame to a file, use the df.to_csv() method.

Key Arguments

path_or_buf: The output file path.
index: Whether to output row labels (index) to the file. Setting this to False removes the index column.
sep: The delimiter character.
index_label: Specifies the column name when outputting the index.

Implementation Sample Code

The following code demonstrates the procedure of creating a dummy CSV file for verification, reading it with various settings, and then outputting the processed data in a different format. Here, we use a system’s “user access log” as the subject matter.

Python

import pandas as pd
import numpy as np
import os

def process_log_data():
    """
    CSV/TSVファイルの読み書き（入出力）を実演する関数
    """
    
    # ---------------------------------------------------------
    # 1. 準備: 動作確認用のサンプルファイルを作成
    # ---------------------------------------------------------
    csv_filename = "sample_access_log.csv"
    tsv_filename = "sample_no_header.tsv"

    # 標準的なCSVデータ（ヘッダーあり）
    csv_content = """user_id,access_time,response_time
1001,2026-05-01 10:00:00,0.45
1002,2026-05-01 10:05:00,1.20
1003,2026-05-01 10:10:00,0.08"""

    # TSVデータ（ヘッダーなし、タブ区切り）
    tsv_content = "2001\tLogin_Success\n2002\tLogin_Failed\n2003\tLogout"

    with open(csv_filename, "w", encoding="utf-8") as f:
        f.write(csv_content)
    
    with open(tsv_filename, "w", encoding="utf-8") as f:
        f.write(tsv_content)

    print(f"--- 準備完了: {csv_filename}, {tsv_filename} を作成しました ---\n")


    # ---------------------------------------------------------
    # 2. ファイルの読み込み (read_csv)
    # ---------------------------------------------------------
    print("=== CSVファイルの読み込み ===")
    # 基本的な読み込み（型指定あり）
    # user_idは整数、response_timeは浮動小数点数として読み込む
    df_csv = pd.read_csv(
        csv_filename,
        dtype={
            "user_id": np.int64,
            "response_time": np.float64
        }
    )
    print(df_csv)
    print(f"dtypes:\n{df_csv.dtypes}\n")

    print("=== TSVファイルの読み込み（ヘッダーなし） ===")
    # sep="\t" でタブ区切りを指定
    # header=None でヘッダー行がないことを明示
    # names引数で列名を後付けで付与
    df_tsv = pd.read_csv(
        tsv_filename,
        sep="\t",
        header=None,
        names=["session_id", "action_type"]
    )
    print(df_tsv)
    print("\n")


    # ---------------------------------------------------------
    # 3. ファイルへの書き込み (to_csv)
    # ---------------------------------------------------------
    print("=== CSVファイルへの出力 ===")
    output_csv = "processed_log.csv"
    
    # index=False にすることで、Pandasの行番号(0, 1, 2...)を出力しない
    df_csv.to_csv(output_csv, index=False)
    print(f"保存完了: {output_csv} (インデックスなし)")

    print("=== TSVファイルへの出力 ===")
    output_tsv = "processed_log.tsv"
    
    # sep="\t" でTSV形式として保存
    # index=True でインデックスも保存し、index_labelでその列名を設定
    df_tsv.to_csv(
        output_tsv,
        sep="\t",
        index=True,
        index_label="row_number"
    )
    print(f"保存完了: {output_tsv} (インデックスあり, ラベル='row_number')")

    # (参考) 作成したファイルを削除する場合は以下を有効化してください
    # os.remove(csv_filename)
    # os.remove(tsv_filename)
    # os.remove(output_csv)
    # os.remove(output_tsv)

if __name__ == "__main__":
    process_log_data()

Execution Result

Plaintext

--- 準備完了: sample_access_log.csv, sample_no_header.tsv を作成しました ---

=== CSVファイルの読み込み ===
   user_id          access_time  response_time
0     1001  2026-05-01 10:00:00           0.45
1     1002  2026-05-01 10:05:00           1.20
2     1003  2026-05-01 10:10:00           0.08
dtypes:
user_id            int64
access_time       object
response_time    float64
dtype: object

=== TSVファイルの読み込み（ヘッダーなし） ===
   session_id   action_type
0        2001  Login_Success
1        2002  Login_Failed 
2        2003        Logout


=== CSVファイルへの出力 ===
保存完了: processed_log.csv (インデックスなし)
=== TSVファイルへの出力 ===
保存完了: processed_log.tsv (インデックスあり, ラベル='row_number')

Notes and Supplements

Delimiter: When handling formats other than CSV (such as tabs or pipes |), always specify the sep argument.
Index Output: The default for to_csv is index=True. To prevent unnecessary row numbers from being saved, it is recommended to specify index=False for general data storage.
Encoding: When handling files containing non-ASCII characters (e.g., Japanese), you may need to explicitly specify encoding="utf-8" or other encodings appropriate for your environment (e.g., pd.read_csv("file.csv", encoding="utf-8")).