如何在CSV文件中查找為主鍵候選集設置的列？

Question

我有一個CSV文件（未規范化，例如，實際文件最多100列）：

   ID, CUST_NAME, CLIENT_NAME, PAYMENT_NUM, START_DATE, END_DATE
    1,     CUST1,     CLIENT1,          10, 2018-04-01, 2018-04-02
    2,     CUST1,     CLIENT1,          10, 2018-04-01, 2018-05-30
    3,     CUST1,     CLIENT1,         101, 2018-04-02, 2018-04-03
    4,     CUST2,     CLIENT1,         102, 2018-04-02, 2018-04-03

如何找到所有可能用作主鍵的列集。

所需的輸出：

  1) ID
  2) PAYMENT_NUM,START_DATE,END_DATE
  3) CUST_NAME, CLIENT_NAME, PAYMENT_NUM,START_DATE,END_DATE

我可以用Java做到，但可能是Python / Pandas已經提供了快速解決方案

Answer 1

熊貓和itertools將為您提供所需的東西。

import pandas
from itertools import chain, combinations

def key_options(items):
    return chain.from_iterable(combinations(items, r) for r in range(1, len(items)+1) )

df = pandas.read_csv('test.csv');

# iterate over all combos of headings, excluding ID for brevity
for candidate in key_options(list(df)[1:]):
    deduped = df.drop_duplicates(candidate)

    if len(deduped.index) == len(df.index):
        print ','.join(candidate)

這將為您提供輸出：

PAYMENT_NUM, END_DATE
CUST_NAME, CLIENT_NAME, END_DATE
CUST_NAME, PAYMENT_NUM, END_DATE
CLIENT_NAME, PAYMENT_NUM, END_DATE
PAYMENT_NUM, START_DATE, END_DATE
CUST_NAME, CLIENT_NAME, PAYMENT_NUM, END_DATE
CUST_NAME, CLIENT_NAME, START_DATE, END_DATE
CUST_NAME, PAYMENT_NUM, START_DATE, END_DATE
CLIENT_NAME, PAYMENT_NUM, START_DATE, END_DATE
CUST_NAME, CLIENT_NAME, PAYMENT_NUM, START_DATE, END_DATE

Answer 2

這是通過itertools.combinations一種方法。 對於每組列，它都通過刪除重復項並檢查數據框的大小是否發生變化來工作。

這將產生44種不同的列組合。

from itertools import combinations, chain

full_list = chain.from_iterable(combinations(df, i) for i in range(1, len(df.columns)+1))

n = len(df.index)

res = []
for cols in full_list:
    cols = list(cols)
    if len(df[cols].drop_duplicates().index) == n:
        res.append(cols)

print(len(res))  # 44