The beauty of automating little tedious things

In which our protagonist reinvents the bycicle, and he is very proud of himself

Somi András | August 4, 2017

Here’s an example why I love coding. Yesterday when I woke up I didn’t know anything about the existence of git hooks and what they are capabale of. By the time I got to bed I managed to automate the whole process of adding header images to my posts and credit them to the photographers. Whoa!

Blog like a (lazy) pro

This blog is a Jekyll static site hosted on GitHub Pages, which provides a fairly straightforward workflow for writing. I just scribble down the new post in Markdown (most efficient way to write structured text!), add some details like the title and subtitle to the top (the so called ‘front-matter’), commit the changes to the local git repo, and git push to GitHub. All the rest is handled by GitHub, automatically building the site from the Jekyll assets and publishing it literally in seconds.

It’s good for me but also good for you, as the result is a static site, where everything is pre-rendered, no database queries, elaborate CMS engines and whatnots. Clean, fast and easy.

There was only one hiccup in this flow (I mean, until yesterday), the header image. I love Unsplash which offers wide range of amazing high-res photos for free. But before ‘The Day I Found Git Hooks’ I had to download, resize and save the chosen image, then add the path and author credits to the post, commit the photo and the changes, and push it to GitHub.

Very inefficient to do all this manually, and not the most elegant way either as Unsplash, unlike other photo providers, encourages hotlinking.

“Unlike most APIs, we prefer for the image URLs returned by the API to be directly used or embedded in your applications (generally referred to as hotlinking). By using our CDN and embedding the photo URLs in your application, we can better track photo views and pass those stats on to the photographer, providing them with context for how popular their photo is and how it’s being used.”

Finding a solution that is good enough

So in theory all I need is the proper url of the photo which I can include into my page’s markup and load the image. ’Something needs to be done!’ - I said to myself in a dramatic voice. But how do I do it in an automated fashion with me putting the least amount of work into the photos?

First, I looked into Jekyll plugins, but with little searching it turned out GitHub Pages runs the Jekyll build with the --safe tag, disabling every custom plugin. Then I found git hooks. (I was actually looking for compiling .less files into .css during build, and found my hints here, but that’s another story.)

As I learned from the docs git runs scripts from the .git/hooks directory at certain milestones (life-cycle events) of its proccesses. Adding a proper pre-commit script let’s me inspect or change the contents of any file that is to be committed before committing it. The script can be in any language, shell, Ruby or my weapon of choice, Python. Great!

And it miraculously works!

I’ve never automated anything in a build process before, but it had this insider feeling, so I had to do it. And in fact it was faster than anticipated to write a Python script, that:

  1. Checks the list of files to be committed, and picks the .md files from the _posts directory (if there’s any).
  2. Extracts the header-id variable from the front-matter for each post.
  3. Retrieves the url of the properly sized photo and the photographer’s name and username from the Unsplash API.
  4. Replaces header-id with these new variables and overwrites the post.
  5. Stages the modified post(s) in git, before committing anything.

It’s definitely not the nicest Python script ever written in the long history of the seven kingdoms of Westeros, but it works.

So now I only have to chose my photo, copy its Unsplash id to the top of my post, and when I first commit the new post, the photo is inserted and proper credits are given to the author in an ‘Unsplash-compatible’ way (I had to also slightly tweak the page templates for that).

I have some other revolutionary ideas with the header images and other stuff on the blog, but for now I just want to enjoy the vast amount of precious time I saved myself with this huge hack.

So here’s this ugly little script:

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

import subprocess
import re
import requests

def get_img_data(id):
    '''Get photo details from Unsplash API by photo id'''
    url = '{}'.format(id)
    params = {
        'w': 1900,
        'client_id': 'XXXXXXX'
    r = requests.get(url, params=params)

    json_result = r.json()
    img_url = json_result['urls']['custom']
    author_name = json_result['user']['name']
    author_profile_url = json_result['user']['username']

    return (img_url, author_name, author_profile_url)

def update_front_matter(path):
    '''Read in .md file, loop through the first line and replace header-id with image details'''
    lines = []
    updated = False
    # Open post for reading
    with open(path, 'r') as post:
        dash_counter = 0
        lines = post.readlines()
        for idx, line in enumerate(lines):
            tokens = line.strip().replace(' ', '').split(':')
            if tokens[0] == '---':
                # Front-matter starts and ends with '---'
                dash_counter += 1
            elif tokens[0] == 'header-id':
                id = tokens[1]
                details = get_img_data(id)

                # New lines to add to front-matter
                url_line = 'header-url: {}\n'.format(details[0])
                photo_author = 'photo-author: {}\n'.format(details[1])
                photo_profile = 'photo-author-profile: {}\n'.format(details[2])

                # Remove header-img line and add new lines
                lines.insert(idx, url_line)
                lines.insert(idx, photo_author)
                lines.insert(idx, photo_profile)
                updated = True

            if dash_counter == 2:
                # Stop if second '---' is found
    if updated:
        # Open post for writing
        # (There should be a simpler way to avoid opening it twice)
        with open(path, 'w') as post:
    return updated

if __name__ == '__main__':
    # Get the list of all (staged, unstaged, untracked) files to be committed
    result ='git status --porcelain', shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
    if result.returncode == 0:
        filelist = result.stdout.decode('UTF-8').strip().split('\n')
        for entry in filelist:
            file_path = entry.split(' ')[1]
            if re.match('_posts/.*\.md', file_path):
                # Only continue with *.md files
                update = update_front_matter(file_path)
                if update:
                    print('--> {} front-matter updated'.format(file_path))
                    # Add newly updated post to staging area
          'git add {}'.format(file_path), shell=True)
    elif result.returncode == 1:
        print('Error when calling pre-commit hook:')

Header photo by Adi Goldstein/ Unsplash