When user enters something invalid (like number 123), it wuold still hit
FB webpage and return HTML. I do simple detection by looking at title
when it attempts to parse the DOM directly. When it contains "Content
Not Found", it is skipped.
Non-existent web pages cannot be parsed by using LDJSON parser and
looking at null data.
Server now downloads the HTML file via new endpoint but the parsing logic now happens in
the browser. The reason for this is to have a way same code for both
environments.
If the JavaScript is disabled, it's still possible to call the previous
endpoint and download the file from the server.
Ondřej Synáček