How to sort (by whatever key) BibTex entries in `.bib` file

I think what you're after is something like these:

Method #1: using head & tail

$ (head -n 2 sample.txt; tail -n +3 sample.txt | sort -t' ' -nk2) > a.tmp && mv a.tmp sample.txt

Nome     Note
------------
Mehdi    0
Shnou    5
Others   10
Sunday   20

This takes the first line of the text file, then tails everything after the first 2 lines which is then sorted.

Method #2: just using head

$ (head -n 2; sort -t' ' -nk2) < sample.txt > a.tmp && mv a.tmp sample.txt

Nome     Note
------------
Mehdi    0
Shnou    5
Others   10
Sunday   20

Takes the text file as input, displays just the first line, sort the rest.

It's typically not a good idea to edit files in place. It's possible, but better to use an intermediate file.

Method #3: Doing #2 without an intermediate file

Stealing the idea from @StephaneChazelas you could do the following using the "1<>" notation to open a file for reading & writing, and the improvements he suggested with the sort command.

$ (head -n 2; sort -nk2) < sample.txt 1<> sample.txt

Nome     Note
------------
Mehdi    0
Shnou    5
Others   10
Sunday   20

The following Python script does the desired filtering.

#!/usr/bin/python
import re

# Bibliography entries to retrieve
# Multiple pattern compilation from: http://stackoverflow.com/a/11693340/147021
pattern_strings = ['Author2010', 'Author2012',]
pattern_string = '|'.join(pattern_strings)
patterns = re.compile(pattern_string)


with open('bibliography.bib', 'r') as bib_file:
    keep_printing = False
    for line in bib_file:
        if patterns.findall(line):
            # Beginning of an entry
            keep_printing = True

        if line.strip() == '}':
            if keep_printing:
                print line
                # End of an entry -- should be the one which began earlier
                keep_printing = False

        if keep_printing:
            # The intermediate lines
            print line,

Personally, I prefer moving to a scripting language when the filtering logic becomes complex. That, perhaps, has an advantage on the readability factor at least.