Avoid retaking all screenshots on publish unless strictly necessary

Deploy will take half an hour because it's going to need to rebuild all 450 screenshots due to an (invisible) change to the entry HTML.

Originally posted by @simonw in #81 (comment)

Changes to the content of a TIL should generate a new screenshot of that TIL.

But, right now, even the tiniest tweak to the base or entry templates (including changes that wouldn't visibly affect the shots at all) triggers a 450+ screenshot generation process.

Easiest fix: instead of incorporating the hash of the HTML templates themselves, switch to a SCREENSHOT_VERSION constant. I can then manually change that any time I want to trigger a full re-render.

I could set the first version of that to the current hash, to avoid recalculating all 450 existing shots when I deploy this change.

til/generate_screenshots.py

Lines 10 to 13 in b8ab160

    
           SHOT_HASH_PATHS = [ 
        
               (root / "templates" / "pages" / "{topic}" / "{slug}.html"), 
        
               (root / "templates" / "til_base.html"), 
        
           ]

til/generate_screenshots.py

Lines 57 to 60 in b8ab160

    
           shot_html_hash = hashlib.md5() 
        
           for filepath in SHOT_HASH_PATHS: 
        
               shot_html_hash.update(filepath.read_text().encode("utf-8")) 
        
           shot_html_hash = shot_html_hash.hexdigest()

Since those hashes are added in sequence, the constant would need to be a list in order to capture the current state.

If I got this right, the deploy should not recreate any screenshots.

Yes that worked:

	SHOT_HASH_PATHS = [
	(root / "templates" / "pages" / "{topic}" / "{slug}.html"),
	(root / "templates" / "til_base.html"),
	]

	shot_html_hash = hashlib.md5()
	for filepath in SHOT_HASH_PATHS:
	shot_html_hash.update(filepath.read_text().encode("utf-8"))
	shot_html_hash = shot_html_hash.hexdigest()