Adds opengraph example

[dom96]

Uses HTMLRewriter to inject meta tags

Deployed to: https://python-opengraph.runtime-playground.workers.dev/

CC @ryanking13 (I can't add you as a reviewer yet)

[dom96]

Without to_js here we get an unclear error:

✘ [ERROR] Uncaught Error: This borrowed attribute proxy was automatically destroyed in the process of destroying the proxy it was borrowed from. Try using the 'copy' method.

  For more information about the cause of this error, use `pyodide.setDebug(true)` Error
      at _getAttrs (pyodide-internal:generated/emscriptenSetup:22274:19)
      at _adjustArgs (pyodide-internal:generated/emscriptenSetup:22292:77)
      at apply (pyodide-internal:generated/emscriptenSetup:23240:37)
      at apply (pyodide-internal:generated/emscriptenSetup:23083:20)


✘ [ERROR] Uncaught Error: This borrowed attribute proxy was automatically destroyed in the process of destroying the proxy it was borrowed from. Try using the 'copy' method.

  For more information about the cause of this error, use `pyodide.setDebug(true)`


✘ [ERROR] Uncaught Error: This borrowed attribute proxy was automatically destroyed in the process of destroying the proxy it was borrowed from. Try using the 'copy' method.

  For more information about the cause of this error, use `pyodide.setDebug(true)`

After reading pyodide/pyodide#1855 I was able to deduce that to_js is at least the first thing I need to do in order to then be able to copy as well. Looks like just doing to_js works here, but maybe there is a hidden bug here that I'm not aware of.

Anyway, we should put HTMLRewriter into our SDK to make this easier for devs.

[ryanking13]

create_proxy would also work, but I found another example using to_js and it is probably good to be consistent.

[hoodmane]

I'd say it should be create_proxy here.

[hoodmane]

To get the resource management right, it should probably be:

        meta_remover = create_proxy(ExistingMetaRemover())
        meta_injector = create_proxy(MetaTagInjector(og_data))

        try:
            # Create an HTMLRewriter instance
            rewriter = HTMLRewriter.new()
            rewriter.on('meta[property^="og:"]', meta_remover)
            rewriter.on("head", to_js(meta_injector))
            return rewriter.transform(response.js_object)
        finally:
            meta_remover.destroy()
            meta_injector.destroy()

We should probably add a wrapper for HTMLRewriter to the sdk that can manage the handler lifetimes.

[dom96]

Okay, done. I needed to add a few copy in there too. Hopefully that makes sense.

[ryanking13]

CC @ryanking13 (I can't add you as a reviewer yet)

Yeah, do you know how I can be added to the GitHub org?

[ryanking13]

create_proxy would also work, but I found another example using to_js and it is probably good to be consistent.

[hoodmane]

I think we don't need these copies.

[dom96]

it fails without them

[hoodmane]

Does the rewriter do any work after rewriter.transform returns?

[dom96]

I wouldn't think that it does, but I guess the error implies this

I removed the copy() and the destroy() for now