With python/pandas:

import pandas
file1 = pandas.read_csv('file1')
file2 = pandas.read_csv('file1')
print(file1)
rs_list = file1['rs_id'].tolist()

# Get rows of file2 corresponding to rs ids in file1:
file2[file2['rs_id'].isin(rs_list)]

Alternatively:

file2.query('rs_id in rs_list')

In general, avoid parsing tabular files manually and access them through dictionaries. It is much more versatile to use pandas dataframes in these cases :-).

Here's my take on it. It seems what you're doing now is identifying all of the annotations and storing them in memory and then writing to the disk, if all you're doing is a simple join you can skip this and write the symbol as you make your output file:

import csv
import sys

def __main__():
    example_csv_1 = sys.argv[1]
    example_csv_2 = sys.argv[2]
    csv_1_id_col  = int(sys.argv[3])
    csv_2_id_col  = int(sys.argv[4])
    csv_2_sym_col = int(sys.argv[5])
    outfile       = sys.argv[6]

    with open(example_csv_2, 'rb') as fi:
        data = list(csv.reader(fi, delimiter=','))

    symbols = dict()

    for val in data:
        try:
            symbols[val[csv_2_id_col]] == None
            print "Error: Duplicate symbols found for ", val[csv_2_id_col]
            sys.exit(1)

        except KeyError:
            symbols[val[csv_2_id_col]] = val[csv_2_sym_col]

    with open(example_csv_1, 'rb') as fi:
        data = list(csv.reader(fi, delimiter=','))

    with open(outfile, 'w') as fi:
        for d in data:
            id = d[csv_1_id_col]
            try:
                sym = symbols[id]
            except KeyError:
                print "Error no gene symbol found for ", id
                sys.exit(1)

            fi.write(','.join(d + [sym]))
            fi.write('\n')

    return 0

__main__()

csv1:

a,stuff
b,stuff
c,stuff
d,stuff

csv2:

a,symA
b,symB
c,symC
d,symD

output:

a,stuff,symA
b,stuff,symB
c,stuff,symC
d,stuff,symD

Again, if all you want in the end is a csv file with annotated with gene symbols, there's no need to make it any more complex than it has to. Just grab the symbols as you write your output file. This will save time since you don't have to iterate through your large CSV file twice. It will also save memory since you don't have any overhead associated with storing the gene symbol for each row (big CSV file) in memory.

This script assumes that each id is associated with a single symbol. If there are multiple symbols per ID it can be easily updated. If you know there aren't any genes with missing symbols (big CSV) or genes with multiple symbols (small CSV), you can omit the checks I have in there and gain a bit of performance.