Skip to content
Advertisement

Joining two CSV files with common column in Python without Pandas

I wanted to inquire on how I could merge 2 csv files so that I can generate queries. Note that I am not allowed to use the “pandas” library.

As an example I have these 2 csv:

data.csv:

cod_pers, cod_enti, fec_venc
2317422,208,04/12/2022
2392115,210,04/02/2022
2086638,211,31/03/2022
2086638,212,03/13/2022

enti.csv:

cod_enti,cod_market
208,40
209,50
210,16
211,40
212,50

And what I’m looking for is to be able to join them through cod_enti and thus be able to evaluate the number of cod_pers that they have as cod_mercado = 40 in the last 15 days.

For this I understand that I can generate the reading of the csv files as follows:

import csv
import numpy as np
from time import strftime
from datetime import datetime, date, time, timedelta
from dateutil.relativedelta import relativedelta

#Read the CSV file
str2date = lambda x: datetime.strptime(x, '%d/%m/%Y')
data_datos = np.genfromtxt('datos.csv', delimiter=',', dtype=None, names=True, converters={'fec_venc':str2date}, encoding="UTF-8")
data_enti = np.genfromtxt('enti.csv', delimiter=',', dtype=None, names=True, encoding="UTF-8")

And later to be able to search by days with a method similar to this:

#definition of days
today = datetime.now()
yesterday = today - timedelta(days=15)

# Generate array of dates
values_on_date =[]
calc_date = data_datos['fec_venc']
for date_obt in calc_date:
    if (yesterday <= date_obt):
        values_on_date.append(date_obt)

tot_doc_mor_15d = len(values_on_date)
print(tot_doc_mor_15d)

But for this I must first join the 2 csv files to be able to generate the query.

I look forward to your comments and any kind of help, it is appreciated. Thank you!! :D

MODIFICATION:

I have added the following lines to my code:

from numpy.lib import recfunctions

merged_array = recfunctions.join_by('cod_enti', data_datos, data_enti)

And it works correctly, but when I want to enter csv with more data, it gives me the following error:

TypeError: invalid type promotion with structured datatype(s).

Advertisement

Answer

You can do this in pretty straight-forward approach with just the CSV module.

I create a map of each row in data.csv to its code_enti value. Then, for every row in enti.csv that has a matching code_enti, I update the row in the map:

import csv
import pprint

# Create a mapping of a data row to its cod_enti, e.g.:
# {208: {cod_pers:2317422, cod_enti:208, fec_venc:04/12/2022}, ...}
cod_enti_row_map = {}

with open("data.csv", newline="") as f:
    reader = csv.DictReader(f, skipinitialspace=True)  # because your header row has leading spaces
    for row in reader:
        cod_enti = row["cod_enti"]
        cod_enti_row_map[cod_enti] = row

print(f"Map before join")
pprint.pprint(cod_enti_row_map, width=100, sort_dicts=False)

# Now, update each row in the map with cod_market for the key, cod_enti
with open("enti.csv", newline="") as f:
    reader = csv.DictReader(f)
    for row in reader:
        cod_enti = row["cod_enti"]

        # skip cod_enti in enti.csv that is not in data.csv, like 209
        if cod_enti not in cod_enti_row_map:
            continue

        cod_enti_row_map[cod_enti].update(row)

print(f"Map after join")
pprint.pprint(cod_enti_row_map, width=100, sort_dicts=False)

Here’s what I get when I run that:

Map before join

{'208': {'cod_pers': '2317422', 'cod_enti': '208', 'fec_venc': '04/12/2022'},
 '210': {'cod_pers': '2392115', 'cod_enti': '210', 'fec_venc': '04/02/2022'},
 '211': {'cod_pers': '2086638', 'cod_enti': '211', 'fec_venc': '31/03/2022'},
 '212': {'cod_pers': '2086638', 'cod_enti': '212', 'fec_venc': '03/13/2022'}}

Map after join

{'208': {'cod_pers': '2317422', 'cod_enti': '208', 'fec_venc': '04/12/2022', 'cod_market': '40'},
 '210': {'cod_pers': '2392115', 'cod_enti': '210', 'fec_venc': '04/02/2022', 'cod_market': '16'},
 '211': {'cod_pers': '2086638', 'cod_enti': '211', 'fec_venc': '31/03/2022', 'cod_market': '40'},
 '212': {'cod_pers': '2086638', 'cod_enti': '212', 'fec_venc': '03/13/2022', 'cod_market': '50'}}

From there, you can extract the rows into a normal list and do all your filtering by key-value, or however else.

User contributions licensed under: CC BY-SA
6 People found this is helpful
Advertisement