這份教學聚焦 Python 內建
csv模組 的正確用法與最佳實務，並補充少量pandas的進階技巧。涵蓋讀寫、方言（dialect）、引號/跳脫、編碼、巨量資料處理、常見陷阱與範例食譜。
- CSV（Comma-Separated Values）是一種以分隔符號（預設逗號
,）分隔欄位的純文字格式。 - 各平台換行符可能不同（
\n,\r\n）。 - 以 Python 內建
csv模組 讀寫時，開檔務必加上newline=''，以避免多出空白行（Windows 常見）。
import csv
from pathlib import Path
p = Path('data.csv')
with p.open('w', newline='', encoding='utf-8') as f:
writer = csv.writer(f)
writer.writerow(['id', 'name', 'score'])
writer.writerow([1, 'Alice', 95])逐列回傳 list（每列一個 list）。
import csv
with open('data.csv', 'r', newline='', encoding='utf-8') as f:
reader = csv.reader(f) # 預設 delimiter=','
for row in reader:
print(row)常用參數：
delimiter=','：分隔符（可改成\t處理 TSV）。quotechar='"'、escapechar='\\'：引號與跳脫。skipinitialspace=True：忽略分隔後的首個空白。
每列回傳 dict（以首列或指定欄名當 key）。
import csv
with open('data.csv', 'r', newline='', encoding='utf-8') as f:
reader = csv.DictReader(f) # 會讀取首列作為欄名
for row in reader:
print(row['name'], row.get('score'))若檔案沒有標題列，可自訂欄名：
with open('no_header.csv', 'r', newline='', encoding='utf-8') as f:
fieldnames = ['id', 'name', 'score']
reader = csv.DictReader(f, fieldnames=fieldnames)
next(reader) # 視情況略過第一列（若第一列是描述或無用資料）csv 讀進來都是字串，需自行轉型：
with open('data.csv', 'r', newline='', encoding='utf-8') as f:
reader = csv.DictReader(f)
rows = []
for r in reader:
r['id'] = int(r['id'])
r['score'] = float(r['score'])
rows.append(r)import csv
data = [
['id', 'name', 'note'],
[1, 'Alice', 'He said: "Hi"'], # 內含引號
[2, 'Bob', 'Line1\nLine2'] # 內含換行
]
with open('out.csv', 'w', newline='', encoding='utf-8') as f:
writer = csv.writer(f, quoting=csv.QUOTE_MINIMAL)
writer.writerows(data)常用參數：
-
delimiter=','：分隔符。 -
lineterminator='\n'：列結尾（預設依平台）。 -
quoting=：引號策略：csv.QUOTE_MINIMAL（預設，必要時加引號）csv.QUOTE_ALL（全欄位加引號）csv.QUOTE_NONNUMERIC（非數值加引號，讀取時自動轉為 float）csv.QUOTE_NONE（不加引號，通常需搭配escapechar）
-
quotechar='"'、escapechar='\\'
import csv
rows = [
{"id": 1, "name": "Alice", "score": 95},
{"id": 2, "name": "Bob", "score": 88},
]
with open('out_dict.csv', 'w', newline='', encoding='utf-8') as f:
fieldnames = ['id', 'name', 'score']
writer = csv.DictWriter(f, fieldnames=fieldnames)
writer.writeheader()
writer.writerows(rows)import csv
csv.register_dialect(
'my_tsv',
delimiter='\t',
quoting=csv.QUOTE_MINIMAL,
lineterminator='\n'
)
with open('data.tsv', 'r', newline='', encoding='utf-8') as f:
reader = csv.reader(f, dialect='my_tsv')import csv
with open('unknown.csv', 'r', encoding='utf-8', newline='') as f:
sample = f.read(2048)
f.seek(0)
dialect = csv.Sniffer().sniff(sample)
has_header = csv.Sniffer().has_header(sample)
reader = csv.reader(f, dialect)
if has_header:
headers = next(reader)- 一般建議使用
utf-8。 - 若要給部分 Excel 版本（特別是 Windows/舊版）可靠辨識，寫檔時可用
utf-8-sig以加入 BOM：
with open('excel_friendly.csv', 'w', newline='', encoding='utf-8-sig') as f:
writer = csv.writer(f)
writer.writerow(['編號', '姓名'])
writer.writerow([1, '王小明'])- 若來源是 CP950/Big5 等本地編碼，讀取時需指定正確
encoding，並處理錯字：
with open('legacy_big5.csv', 'r', newline='', encoding='cp950', errors='replace') as f:
reader = csv.reader(f)- 逐行處理：避免一次載入全部（特別是
readlines()），改用迭代器。 - 欄位過濾/映射：在讀取迴圈中即時轉型與過濾，減少中間結構。
- 分塊寫出：處理 N 筆就寫一次，避免巨大記憶體佔用。
- 壓縮檔：可直接用
gzip/bz2搭配檔案物件：
import csv, gzip
with gzip.open('data.csv.gz', 'rt', newline='', encoding='utf-8') as f:
reader = csv.DictReader(f)
for row in reader:
...- Windows 寫檔多出空行：忘了
newline=''。 - 欄位內含逗號/引號/換行：需設定
quoting（預設即可處理），或搭配escapechar。 - 型別全是字串：自行轉型（
int,float,datetime）。 - 雜訊標頭/註解列：讀取時先
next(reader)跳過或預處理。 - 分隔符不是逗號：指定
delimiter或用Sniffer。
import csv
with open('sales.csv', 'r', newline='', encoding='utf-8') as fin, \
open('high_value.csv', 'w', newline='', encoding='utf-8') as fout:
reader = csv.DictReader(fin)
fieldnames = reader.fieldnames
writer = csv.DictWriter(fout, fieldnames=fieldnames)
writer.writeheader()
for r in reader:
if float(r['amount']) >= 1000:
writer.writerow(r)import csv
with open('in.csv', 'r', newline='', encoding='utf-8') as f:
rows = list(csv.reader(f))
transposed = list(zip(*rows))
with open('out_transposed.csv', 'w', newline='', encoding='utf-8') as f:
writer = csv.writer(f)
writer.writerows(transposed)import csv
with open('scores.csv', 'r', newline='', encoding='utf-8') as fin, \
open('scores_plus.csv', 'w', newline='', encoding='utf-8') as fout:
reader = csv.DictReader(fin)
fieldnames = reader.fieldnames + ['passed']
writer = csv.DictWriter(fout, fieldnames=fieldnames)
writer.writeheader()
for r in reader:
r['passed'] = 'Y' if float(r['score']) >= 60 else 'N'
writer.writerow(r)import csv, glob
files = glob.glob('parts/part_*.csv')
with open('merged.csv', 'w', newline='', encoding='utf-8') as fout:
writer = None
for path in files:
with open(path, 'r', newline='', encoding='utf-8') as fin:
reader = csv.DictReader(fin)
if writer is None:
writer = csv.DictWriter(fout, fieldnames=reader.fieldnames)
writer.writeheader()
for row in reader:
writer.writerow(row)import csv
with open('data.tsv', 'r', newline='', encoding='utf-8') as f:
reader = csv.reader(f, delimiter='\t')
for row in reader:
...
with open('pipe.csv', 'w', newline='', encoding='utf-8') as f:
writer = csv.writer(f, delimiter='|')
writer.writerow(['a', 'b', 'c'])- 讀：
import pandas as pd
df = pd.read_csv('data.csv', encoding='utf-8')
# 常用參數：usecols, dtype, parse_dates, na_values, chunksize- 寫：
df.to_csv('out.csv', index=False, encoding='utf-8-sig')- 大量資料：
chunksize分塊讀取、指定dtype降低記憶體。
- 讀寫 CSV 時 一定用
newline=''；編碼建議utf-8/utf-8-sig（對 Excel 友善）。 - 欄位內含逗號、引號、換行時，調整
quoting/escapechar；或保留預設讓模組自動處理。 DictReader/DictWriter能以欄名取值，更可讀。- 巨量檔案：逐行處理、分塊寫出、必要時搭配壓縮與
pandas。 - 善用
Sniffer與 Dialect 對付未知或非標準 CSV。