熊猫：具有相同列表的不同索引性能

Question

last5_ls = range(-5, 0, 1)
last25_ls = range(-25, 0, 1)
data = data.groupby('stockid')
data_last5 = data.nth(last5_ls)
data_last25 = data.nth(last25_ls)
close25_dict = {}
close5_dict = {}

选项A：约1分钟

stockid_ls = data_last25.index.drop_duplicates().tolist()
for sid in stockid_ls:
    if len(data_last25.loc[sid, 'close']) >= 25:
        close25_dict[sid] = data_last25.loc[sid, 'close'].tolist()
        close5_dict[sid] = data_last5.loc[sid, 'close'].tolist()

选项B：约25秒

stockid_ls = data_last5.index.drop_duplicates().tolist() 
for sid in stockid_ls:
    if len(data_last25.loc[sid, 'close']) >= 25:
        close25_dict[sid] = data_last25.loc[sid, 'close'].tolist()
        close5_dict[sid] = data_last5.loc[sid, 'close'].tolist()

data_last5.index.drop_duplicates().tolist()和data_last25.index.drop_duplicates().tolist()的结果相同。

但是，如果我将它们用作循环列表，前者要快得多。为什么会这样？

data_last25的示例：

    close   transtime
stockid         
000001.SH   3145.7724   2017-02-07 11:25:00
000001.SH   3146.2715   2017-02-07 11:30:00
000001.SH   3145.4279   2017-02-07 13:05:00
000001.SH   3146.0700   2017-02-07 13:10:00
000001.SH   3145.5921   2017-02-07 13:15:00
000001.SH   3144.9807   2017-02-07 13:20:00
000001.SH   3141.3619   2017-02-07 13:25:00
000001.SH   3142.4082   2017-02-07 13:30:00
000001.SH   3141.7696   2017-02-07 13:35:00
000001.SH   3144.6115   2017-02-07 13:40:00
000001.SH   3144.9876   2017-02-07 13:45:00
000001.SH   3146.0163   2017-02-07 13:50:00
000001.SH   3145.1942   2017-02-07 13:55:00
000001.SH   3146.9230   2017-02-07 14:00:00
000001.SH   3147.7423   2017-02-07 14:05:00
000001.SH   3146.1607   2017-02-07 14:10:00
000001.SH   3145.6695   2017-02-07 14:15:00
000001.SH   3148.0311   2017-02-07 14:20:00
000001.SH   3150.0356   2017-02-07 14:25:00
000001.SH   3150.7101   2017-02-07 14:30:00
000001.SH   3149.4439   2017-02-07 14:35:00
000001.SH   3148.4014   2017-02-07 14:40:00
000001.SH   3147.6509   2017-02-07 14:45:00
000001.SH   3147.9084   2017-02-07 14:50:00
000001.SH   3150.6996   2017-02-07 14:55:00
000001.SZ   9.2900  2017-02-07 11:25:00
000001.SZ   9.2800  2017-02-07 11:30:00
000001.SZ   9.2800  2017-02-07 13:05:00
000001.SZ   9.2800  2017-02-07 13:10:00
000001.SZ   9.2800  2017-02-07 13:15:00
000001.SZ   9.2800  2017-02-07 13:20:00
000001.SZ   9.2800  2017-02-07 13:25:00
000001.SZ   9.2800  2017-02-07 13:30:00
000001.SZ   9.2800  2017-02-07 13:35:00
000001.SZ   9.2800  2017-02-07 13:40:00
000001.SZ   9.2800  2017-02-07 13:45:00
000001.SZ   9.2900  2017-02-07 13:50:00
000001.SZ   9.2800  2017-02-07 13:55:00
000001.SZ   9.2800  2017-02-07 14:00:00
000001.SZ   9.2800  2017-02-07 14:05:00
000001.SZ   9.2900  2017-02-07 14:10:00
000001.SZ   9.2900  2017-02-07 14:15:00
000001.SZ   9.2900  2017-02-07 14:20:00
000001.SZ   9.2900  2017-02-07 14:25:00
000001.SZ   9.3000  2017-02-07 14:30:00
000001.SZ   9.3000  2017-02-07 14:35:00
000001.SZ   9.2800  2017-02-07 14:40:00
000001.SZ   9.2900  2017-02-07 14:45:00
000001.SZ   9.2800  2017-02-07 14:50:00
000001.SZ   9.2900  2017-02-07 14:55:00

data_last5的示例：

    close   transtime
stockid         
000001.SH   3149.4439   2017-02-07 14:35:00
000001.SH   3148.4014   2017-02-07 14:40:00
000001.SH   3147.6509   2017-02-07 14:45:00
000001.SH   3147.9084   2017-02-07 14:50:00
000001.SH   3150.6996   2017-02-07 14:55:00
000001.SZ   9.3000  2017-02-07 14:35:00
000001.SZ   9.2800  2017-02-07 14:40:00
000001.SZ   9.2900  2017-02-07 14:45:00
000001.SZ   9.2800  2017-02-07 14:50:00
000001.SZ   9.2900  2017-02-07 14:55:00
000002.SH   3297.8762   2017-02-07 14:35:00
000002.SH   3296.7849   2017-02-07 14:40:00
000002.SH   3295.9991   2017-02-07 14:45:00
000002.SH   3296.2701   2017-02-07 14:50:00
000002.SH   3299.2052   2017-02-07 14:55:00