Sampling of the dateset under several conditions Python

[Mystic]

A date is given at the entrance where id is the user, url is the page reviewed, timestamp is the time of the page.

    id  url timestamp
0   a   page_1  2021-10-09 15:46:20
1   a   page_2  2021-10-09 15:47:20
2   a   page_3  2021-10-09 15:48:20
3   a   page_4  2021-10-09 15:49:20
4   a   page_2  2021-10-09 15:50:20
5   b   page_4  2021-10-09 15:18:20
6   b   page_3  2021-10-09 15:21:20
7   b   page_2  2021-10-09 15:22:20
8   b   page_1  2021-10-09 15:24:20
9   b   page_1  2021-10-09 15:26:20

Each user guaranteed visit page 2 Each user ' s home page should be selected until page 2. If the user has visited page 2 several times, it is necessary to include all pages before of the page 2 The original mass is very large, so I'd like to find a way faster than a diversion. Thank you.

[elysha]

My decision will only work if your data have the following format:

# файл input.csv
id,url,timestamp
a,page_1,2021-10-09 15:46:20
a,page_2,2021-10-09 15:47:20
a,page_3,2021-10-09 15:48:20
a,page_4,2021-10-09 15:49:20
a,page_2,2021-10-09 15:50:20
b,page_4,2021-10-09 15:18:20
b,page_3,2021-10-09 15:21:20
b,page_2,2021-10-09 15:22:20
b,page_1,2021-10-09 15:24:20
b,page_1,2021-10-09 15:26:20

Decision:

import pandas as pd
def all_visited_pages_until(page, user_id, df):

df_by_user = df[(df.id == user_id)]

df_by_user = df_by_user.sort_values(by="timestamp")

last_row_index_by_page = df_by_user.url.where(df_by_user.url == page).last_valid_index()
return df_by_user.loc[:last_row_index_by_page]

def main():

df = pd.read_csv("input.csv", sep=",")

print(all_visited_pages_until("page_2", "a", df))
if name == 'main':

main()

Programme withdrawal:

0    page_1

1    page_2

2    page_3

3    page_4

4    page_2

Name: url, dtype: object

On the way out, they got the object. Series♪ If necessary, it could be transformed into list

== sync, corrected by elderman ==

If all pages have a view page_i (sighs)i This number, respectively, may be replaced to optimize the storage of large data volumes page_i Keep it. i♪

To this end, before calling a function all_visited_pages_until enough to write df.url = df.url.apply(lambda url: int(url.replace("page_", "")))