Actually, you did. So keep it up.
That is one of the first things the script searches for. Making it bold or putting it into quotations marks is also very helpful.
I do. I really don’t want to be converted into paper clips some day.
It depends on the circumstances. In most cases (and after stripping the markup) the script is allowed to consider everything between 5 and 40 chars a title. This has proven to minimize the number of false positives quite reliably. However, this is like Star Trek - there is always a manual override. So, there is no real limit to the title length.
This is just like Amazon showing you ‘personalized’ advertisements for things that just don’t make any sense. They just try to keep you calm and hide their real evil power.
So, don’t worry guys. The current script is working just fine.
For everyone interested: The new version I created for Discourse is almost completely JavaScript based. (I only use PHP for downloading the JSON files (which make things A LOT easier - compared to the old forum’s HTML) from blenderartists.org in order to get around any cross domain issues.)
There are three main reasons for choosing JavaScript.
1.) I like it.
2.) It is great at fiddling with HTML and JSON
3.) It doesn’t matter where I am. I could even look through the latest posts or edit things the script may have misunderstood from everywhere in the world. (And on every device. And every OS.)
If you are still reading, here are the details on what the script does:
Every time I run the script, it piles up the necessary JSON files and detects all the new posts. For every post, it ‘guesses’ if it is an entry and tries to detect which kind of entry it is (pure, open, non-competing, …). The participants name and last edit date are easy (these are part of the JSON file). Finding the title and correct entry image is a little tricky. So it tries to guess as good as possible. (And it removes “title:” from the title strings. )
After this has all been done, the results are inserted into text input fields which (as @Napivo suggested above) can then be edited by hand. IF I notice an error, I correct it and save the latest changes. I do this every once in a while, whenever I find a minute or two. This way, almost all entries have already been covered before Monday evening. (Currently, the voting thread starts at about 1AM in Germany, so setting things up should take as little time as possible. ;-)).
After that, all that is left to do is ordering the entries (by edit date and type) and creating the voting/post text/source which can then be pasted into the forum.
Done