UP | HOME

Spice up with Nix: Scripts with magical dependencies

Published on 2021-08-09

I wrote this cool script which shows you a list of letters and their frequency in English.

Check it!

#!/usr/bin/env python3

import requests                # Required to communicate via HTTP
import pandas                  # Required to mangle with spreadsheets
from bs4 import BeautifulSoup  # Required to parse HTML

# Let's download the Letter frequency page from Wikipedia
r = requests.get('https://en.wikipedia.org/wiki/Letter_frequency')
assert r.status_code == 200

# Use BeautifulSoup to parse HTML
s = BeautifulSoup(r.text, 'html.parser')

# Find the last table in the page
res = str(s.findAll('table', {'class': 'wikitable'})[-1])

# Let's convert that HTML table to a Pandas DataFrame
res = pandas.read_html(res)[0]

# We're only interested in English and the first 25 letters.
res = res[['Letter', 'English']][0:26]

# Properly handle the percentage in the table
res['English'] = res['English'].str.rstrip('%').astype('float') / 100.0

# Sort it
res = res.sort_values(by='English', ascending=False)

# Show it to the user
print(res)

Cool, right?

Oh, you don’t have Python on your machine?

Maybe, you don’t have the libraries it requires like requests, pandas, and BeautifulSoup.

Ah, it doesn’t run on your machine because you only have a different version of those libraries?

Instant runtime environments for scripts with Nix

If you don’t have Nix installed on your machine, install it now!

It won’t break your system, I promise!

As of writing, Nix runs well on macOS and Linux.

Nix can be used to setup a temporary runtime environment with all of the dependencies which your script need.

Simply use nix-shell as an interpreter in the shebang (#!) of your script.

Using the script I’ve written above instead of using Python which is installed on the machine:

#!/usr/bin/env python3

Replace with nix-shell and then give a list of packages to nix-shell.

#! /usr/bin/env nix-shell
#! nix-shell -i python -p python38 python38Packages.requests python38Packages.beautifulsoup4 python38Packages.pandas
#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/refs/tags/21.05.tar.gz

Let’s break it down on what each line means!

#! /usr/bin/env nix-shell

The first line simply tells your shell to use nix-shell as the interpreter for this script, nothing new here. Not that different how Python scripts are executed!

#! nix-shell -i python -p python38 python38Packages.requests python38Packages.beautifulsoup4 python38Packages.pandas

This line seems scary but no worries. We are simply passing arguments to nix-shell.

  • -i python tells Nix which interpreter to use. In this case, we’re using python.
  • -p python38 python38Packages... tells Nix which packages are required for this script.

It’s not that different to a recipe.

Instead of only containing instructions on what to do, with Nix, we are able to give a list of ingredients and what to use in order to perform the instructions!

#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/refs/tags/21.05.tar.gz

This line looks scary but no worries, it’s simple.

We’re telling Nix which version of the Nix packages collection (nixpkgs) to use. In this case, we are using version 21.05.

This is how we can pin our dependencies to ensure the person who are running the script are running the exact version of the dependencies that I works on my machine.

This also ensures that even if you read this 5 or even 10 years in the future, the script will still run with the same set of dependencies.

So, here’s the final script after we spice it up with Nix:

#! /usr/bin/env nix-shell
#! nix-shell -i python -p python38 python38Packages.requests python38Packages.beautifulsoup4 python38Packages.pandas
#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/refs/tags/21.05.tar.gz

import requests                # Required to communicate via HTTP
import pandas                  # Required to mangle with spreadsheets
from bs4 import BeautifulSoup  # Required to parse HTML

# Let's download the Letter frequency page from Wikipedia
r = requests.get('https://en.wikipedia.org/wiki/Letter_frequency')
assert r.status_code == 200

# Use BeautifulSoup to parse HTML
s = BeautifulSoup(r.text, 'html.parser')

# Find the last table in the page
res = str(s.findAll('table', {'class': 'wikitable'})[-1])

# Let's convert that HTML table to a Pandas DataFrame
res = pandas.read_html(res)[0]

# We're only interested in English and the first 25 letters.
res = res[['Letter', 'English']][0:26]

# Properly handle the percentage in the table
res['English'] = res['English'].str.rstrip('%').astype('float') / 100.0

# Sort it
res = res.sort_values(by='English', ascending=False)

# Show it to the user
print(res)

Only by changing that single line to three lines, this script will now run on any machine with Nix installed and it will automagically resolve the dependencies magically!

$ head -n3 cool-script.py
#! /usr/bin/env nix-shell
#! nix-shell -i python -p python38 python38Packages.requests python38Packages.beautifulsoup4 python38Packages.pandas
#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/refs/tags/21.05.tar.gz
$ ./cool-script.py
these 85 paths will be fetched (132.87 MiB download, 693.09 MiB unpacked):
  /nix/store/0b90d74742wl9vpy0vpd89k6cks9xvd9-bash-interactive-4.4-p23-man
  ...
  /nix/store/zsh6dmqlp15yzj3mgq3zv212waz12lbk-python3.8-urllib3-1.26.4
copying path '/nix/store/lp6ccpip5ygq0k813gyb9m4iashzk2r0-bash-interactive-4.4-p23-doc' from 'https://hydra.iohk.io'...
...
copying path '/nix/store/4dlhsl4kxp9p632mbv1rcq9kjc0y6zdy-stdenv-linux' from 'https://hydra.iohk.io'...
   Letter  English
4       e  0.12702
19      t  0.09056
0       a  0.08167
14      o  0.07507
8       i  0.06966
13      n  0.06749
18      s  0.06327
7       h  0.06094
17      r  0.05987
3       d  0.04253
11      l  0.04025
2       c  0.02782
20      u  0.02758
12      m  0.02406
22      w  0.02360
5       f  0.02228
6       g  0.02015
24      y  0.01974
15      p  0.01929
1       b  0.01492
21      v  0.00978
10      k  0.00772
9       j  0.00153
23      x  0.00150
16      q  0.00095
25      z  0.00074

As you can see, Nix will magically prepare a temporary environment with the dependencies for your script.

Wait, does this mean Nix has installed those dependencies and Python?

$ python3
Python 3.8.9 (default, Apr  2 2021, 11:20:07)
[GCC 10.3.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import requests
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'requests'
>>> import pandas
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'pandas'
>>> import bs4
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'bs4'
>>>

As you can see, it didn’t alter your system. Your system is safe.

Only Python? Nope.

This is not limited to only Python. You can use it with everything.

You want to use it in a shell script which require programs to be present on the system? No problem.

#! /usr/bin/env nix-shell
#! nix-shell -i bash -p gnat mktemp
#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/refs/tags/21.05.tar.gz

# Make a temp folder
WORKDIR=$(mktemp -d)

# Go to the temp directory
pushd "${WORKDIR}"

# Write some code
cat > hello.adb <<EOF
with Text_IO; use Text_IO;
procedure hello is
begin
   Put_Line("Hello, world!");
end hello;
EOF

# Compile it with GNAT
gnatmake hello.adb

# Run it
./hello

# Go back to the previous directory
popd

# Copy the executable
cp "${WORKDIR}/hello" ./hello

# Remove the temp folder
rm -rf "${WORKDIR}"
$ ./cool.sh
these 9 paths will be fetched (46.63 MiB download, 226.41 MiB unpacked):
  /nix/store/1977rsramab0ks7bk3fxjd91kf1ddxy1-libmpc-1.2.1
  ...
  /nix/store/v07f9gbkizyfcz98ssyhbq5jr6x823lm-gmp-6.2.1
copying path '/nix/store/by026wm1s9phidfb96l5gcxyylly9j5y-expand-response-params' from 'https://hydra.iohk.io'...
...
copying path '/nix/store/a7y4ng9g8jn1vm5p6l34nbg8ns5ml7f2-gnat-wrapper-9.3.0' from 'https://cache.nixos.org'...
/run/user/1000/tmp.Q6gesiZ73k ~
gcc -c -I/nix/store/9b5w72pds7z3mj2bplx97yn9vpz5x0zl-gnat-9.3.0/lib/gcc/x86_64-unknown-linux-gnu/9.3.0/adainclude -I/nix/store/9b5w72pds7z3mj2bplx97yn9vpz5x0zl-gnat-9.3.0/lib/gcc/x86_64-unknown-linux-gnu/9.3.0/adalib hello.adb
/nix/store/a7y4ng9g8jn1vm5p6l34nbg8ns5ml7f2-gnat-wrapper-9.3.0/bin/gnatbind -I/nix/store/9b5w72pds7z3mj2bplx97yn9vpz5x0zl-gnat-9.3.0/lib/gcc/x86_64-unknown-linux-gnu/9.3.0/adainclude -I/nix/store/9b5w72pds7z3mj2bplx97yn9vpz5x0zl-gnat-9.3.0/lib/gcc/x86_64-unknown-linux-gnu/9.3.0/adalib -x hello.ali
/nix/store/a7y4ng9g8jn1vm5p6l34nbg8ns5ml7f2-gnat-wrapper-9.3.0/bin/gnatlink hello.ali
Hello, world!
~
$ ls ./hello
./hello
$ ./hello
Hello, world!

Did Nix just prepare an environment with the GNU Ada Toolchain and then run a script which compile an Ada program which runs?

Yes, it did.

What about something niche?

#! /usr/bin/env nix-shell
#! nix-shell -i runhaskell -p ghc
#! nix-shell -I nixpkgs=https://github.com/NixOS/nixpkgs/archive/refs/tags/21.05.tar.gz

nums :: [Int]
nums = [1780, 1693, 1830, 1756, 1858, 1868, 1968, 1809, 1996, 1962, 1800, 1974,
        1805, 1795,  170, 1684, 1659, 1713, 1848, 1749, 1717, 1734,  956, 1782,
        1834, 1785, 1786, 1994, 1652, 1669, 1812, 1954, 1984, 1665, 1987, 1562,
        2004, 2010, 1551,  961, 1854, 2005, 1883, 1965,  475, 1776, 1791,  262,
        1912, 1227, 1486, 1989, 1857,  825, 1683, 1991, 1875, 1982, 1654, 1767,
        1673, 1973, 1886, 1731, 1745, 1770, 1995, 1721, 1662, 1679, 1783, 1999,
        1889, 1746, 1902, 2003, 1698, 1794, 1798, 1951, 1953, 2007, 1899, 1658,
        1705,   62, 1819, 1708, 1666, 2006, 1763, 1732, 1613, 1841, 1747, 1489,
        1845, 2008, 1885, 2002, 1735, 1656, 1771, 1950, 1704, 1737, 1748, 1759,
        1802, 2000, 1955, 1738, 1761, 1765, 1853, 1900, 1709, 1979, 1911, 1775,
        1813, 1949, 1966, 1774, 1977, 1757, 1992, 2009, 1956, 1840, 1988, 1985,
        1993, 1718, 1976, 1078, 1997, 1897, 1792, 1790, 1801, 1871, 1727, 1700,
        1485,  942, 1686, 1859, 1676,  802, 1952, 1998, 1961, 1844, 1808, 1703,
        1980, 1766, 1963, 1849, 1670, 1716, 1957, 1660, 1816, 1762, 1829,  526,
         359, 2001, 1874, 1778, 1873, 1511, 1810, 1699, 1970, 1690, 1978, 1892,
        1691, 1781, 1777, 1975, 1967, 1694, 1969, 1959, 1910, 1826, 1672, 1655,
        1839, 1986, 1872, 1983, 1981, 1972, 1772, 1760]

solution :: [Int] -> Int
solution xs =
  head
    [ x * y * z
      | x <- xs,
        y <- xs,
        z <- xs,
        x + y + z == 2020
    ]
main :: IO ()
main = print $ solution nums
% ./day1-part2-aoc.hs
these 4 paths will be fetched (138.01 MiB download, 1870.79 MiB unpacked):
  /nix/store/1pfdb37qm067ib7dnj5arbhl5q2kam3l-libffi-3.3-dev
  /nix/store/55ysfrn085k1lsya0c2msm6npi7x6ihh-ghc-8.10.4-doc
  /nix/store/f2psw0phlmp7h7gk14rfsqdmjz4d1arb-gmp-6.2.1-dev
  /nix/store/mcm0wvsqcwqaxsp8fnwmknmpi0a196w6-ghc-8.10.4
copying path '/nix/store/55ysfrn085k1lsya0c2msm6npi7x6ihh-ghc-8.10.4-doc' from 'https://cache.nixos.org'...
copying path '/nix/store/f2psw0phlmp7h7gk14rfsqdmjz4d1arb-gmp-6.2.1-dev' from 'https://hydra.iohk.io'...
copying path '/nix/store/1pfdb37qm067ib7dnj5arbhl5q2kam3l-libffi-3.3-dev' from 'https://hydra.iohk.io'...
copying path '/nix/store/mcm0wvsqcwqaxsp8fnwmknmpi0a196w6-ghc-8.10.4' from 'https://cache.nixos.org'...
200878544

As you can see, it works for any language. Simply check if your favorite language and libraries are packaged: search.nixos.org.

Closing

This is the first in a series of article that I will be publishing more regularly.

It is based of a mega draft I have written about Nix which was written in the same style of the Docker article.

However, due to the massive differences in Nix vs other existing package managers, it has grown to a rather massive and daunting draft (1787 lines, 7265 words, 68436 characters).

Because it would be a massive article and a huge turn off for a lot of people due to its size, I will be breaking it down into smaller bite-sized posts like this one.

I will start with practical examples on how Nix can help in a lot of real-world situations. In fact, I have delivered a couple of real world projects on production with Nix and at work, we are now happily using Nix by default.

Once that’s done, we will dive head first and see how Nix works. I hope you’ll find it interesting and this has been one of the longest writing projects that I’ve done with so much research!

Thanks for reading and I hope you’ll be there on the next one!