2k tool meant to automate editing on Stack Overflow, because there's too much crap to do it all manually.
Disclaimer: There's going to be false positives. All use of this tool is at your own risk.
- Python 3 (because there's few languages with good API interfaces, and C++ isn't one of them)
- Various dependencies:
python3 -m pip install -r requirements.txt - 2k+ reputation on Stack Overflow (this tool actively blocks non-moderators with <2k reputation to avoid review queue flooding)
- A terminal that supports colorama (potentially optional, but definitely a requirement for QOL)
Because Dragon requires write access, it has to use the client-side OAuth flow. It's set up to redirect to a simple page (source code) that echos and copies the token. The page itself doesn't save your token (or, as the page source also says, I don't save it. I have no idea if GitHub saves URLs for security purposes) - copy it, or it'll be lost
You can save this token into .oauth.txt manually, or paste it into the program when prompted. If you use Stack Overflow in a container tab in Firefox, the URL is also explicitly printed so it's copy-pastable. (reopening in the appropriate tab doesn't work - blame Stack Overflow for not following the redirect URL if you're already logged in)
The reason Dragon only supports Stack Overflow is the patterns.
Blanket-removing "thanks" and "please" will not work on a language site, for an instance. Same with a few other grammar rules, because any misspelled words may actually be intentionally written that way, and be critical for the question.
Other sites in general also have other patterns - consequently, what Dragon has to be efficient on SO may not be enough to be efficient on [insert your favorite site].
As a result, to adapt Dragon to another site, all filters need to be re-written. This also includes for localized versions of SO, as several of the filters are English-centric, and may introduce problems with other languages.
- Add something (state machine?) that attempts to determine certain aspects about a post. This may end up being heavy, at least if we expand outside the realm of regex. May require machine learning and end up being unfeasible.
- Reupload third party images to stack.imgur.com
- Find more uncaught stuff to put in uncaught.md
- Clear out uncaught.md