Bloody cunts at Twitter could provide the fucking grammar for their filtering rules...

Now I have to write the grammar for the lexer and the parser from scratch (in fucking JavaScript to boot 😡).

Mind you, I know my lex and bison, but I haven't done this shit in fucking ages, and the combo of JavaScript debauchery and being rusty, is making me want to send angry tweets to Elon musk, see if they can provide decent tools for their shit API.

It would be more bearable if they at least provided clear instructions.

Sure, they give info on operator precedence, but nothing on escaping mechanisms, nor validation.

Sure, quotes allow you to match for exact multiword matches. Do you allow escaping quotes? API doesn't complain either way, but you can't be sure how that's being interpreted...

Can you also escape other metachars? What is the arity of each operator? Do you respect Shannon's theorem? (!a or !b == !(a and b))?

They give some partial answers to all of those but my findings demonstrate even their partial answers aren't always true.

@kaki

Thanks for the suggestion, but this ain't about machine learning.

The grammar and syntax of this language are definitely set in stone. I'm just trying to reverse engineer it.

It's not a matter of how accurate I can get. There is *one* right answer, and the rest are wrong. This is not something ML is needed for.

@CoreFusionX that’s a ballache of Gaussian proportions