feat: Saving image previews

[Lyrete]

Describe your changes

• Implemented a function to scrape the og_image from the resource, convert it to a .webp and write it to a file in static/previews.
• Update ListItem.svelte to use the base url in front of the og_preview so they point correctly.
• Some other minor nitpick fixes

New dependencies

• Added httpx as a dependency, will need it for parallelization and good usage already here since it's a good idea to mask the user-agent to not get blocked (as often) when scraping the images.
• Also upgraded pip-tools version to 6.13.0 as it seems 6.12 i broken with recent versions of pip: TypeError: WheelCache.__init__() takes 2 positional arguments but 3 were given jazzband/pip-tools#1854

Related issue number/link

Fixes #219
Will probably help with #201 as well

[VeckoTheGecko]

This looks awesome!

Re. your #219 (comment)

What was the idea with the naming schema? For an initial thing I implemented a 8 char hex digest off the resource name...

Using the URL instead of resource name would ensure no name collisions. You can use a hex digest of that, or make the URL filename safe (the former probably a better option to avoid filename length issues).

Otherwise this is fairly done so it's a question of whether you'd prefer two PRs if I do #218 too.

I think two separate PRs would be great and be a bit easier to review (especially since I'm less familiar with async stuff so want better visibility on that). I'm happy to review this ASAP so that your workflow isn't interrupted, feel free to mark as "ready for review" when you're ready :) .

[VeckoTheGecko]

ik its a draft, but here are some preliminary thoughts :) .

[Lyrete]

Using the URL instead of resource name would ensure no name collisions. You can use a hex digest of that, or make the URL filename safe (the former probably a better option to avoid filename length issues).

Was actually already doing that while you posted your comment :D

I'll do a bit of clean up and then should be good to go :) There's a good chance I have to touch this functionality anyway while doing the parallelization, so better not to go too deep with the optimizing.

[VeckoTheGecko]

so better not to go too deep with the optimizing.

That's fair. I think rather than rescaling then, let's just make sure that we properly do the conversion to webp and jpeg formats. This is quick convo I had with ChatGPT that may help with development. Looks like a quick call to Pillow.

Rescaling can be left to another PR

[Lyrete]

Implemented the webp thing so should be fairly good to go now. Fairly big bump (30-40s) to the build time but I guess that should get optimized in the next one.

Oh wait the Pillow dep isn't in there yet, oops.

[Lyrete]

There's some github actions issue btw, where the build workflow happens if a precommit was made on github on a PR but not if it wasn't. I don't know which is your intention.

[VeckoTheGecko]

I'll implement these, just putting here for transparency

[VeckoTheGecko]

# TODO: Maybe add additional behaviour to check if the image is large enough

I've removed this TODO comment. I think its a very low priority issue, and would be the 'fault' of the website we're scraping from. If we add image rescaling/optimisation, this can be rolled into that I guess.

[VeckoTheGecko]

There's some github actions issue btw

I think thats due to the permissions and me having to authorize workflows. I don't think its an issue

[VeckoTheGecko]

I think 22798d3 was breaking 😅 At least in the VScode Markdown preview. Reverting for wider compatability

[VeckoTheGecko]

LGTM! Thanks for this PR @Lyrete 🥳🥳🥳🥳

[VeckoTheGecko]

@all-contributors please add @Lyrete for code

[allcontributors]

@VeckoTheGecko

I've put up a pull request to add @Lyrete! 🎉

[Lyrete]

I think 22798d3 was breaking 😅 At least in the VScode Markdown preview. Reverting for wider compatability

Interestingz for me the emoji was breaking 🤔 the link in the radme doesn't work with it.

[VeckoTheGecko]

Ah, didn't think to check the README on GitHub 😅. I'll revert the reversion 😁