| T O P I C R E V I E W |
| Uniwares |
Posted - Aug 18 2020 : 10:10:10 AM
Just found what I wish VA would do: https://marketplace.visualstudio.com/items?itemName=jinyafeng.DocumentationAssistant
Creates pretty nifty documentation with auto derived content from member name and type. |
| 7 L A T E S T R E P L I E S (Newest First) |
| Uniwares |
Posted - Aug 27 2020 : 08:45:56 AM
Can always count on me!  |
| feline |
Posted - Aug 27 2020 : 07:56:59 AM
Thank you for volunteering to support this new and exciting feature |
| Uniwares |
Posted - Aug 26 2020 : 3:58:14 PM
When? VA2390?  |
| accord |
Posted - Aug 26 2020 : 3:54:55 PM
I have mentioned that GPT-3 might be able to help with this and someone who was lucky to get an early access is already playing around with this with surprisingly good results:
https://twitter.com/amasad/status/1285789362647478272 |
| feline |
Posted - Aug 19 2020 : 10:43:53 AM
Urg, considering this from the support point of view, when you start getting variables created on the fly with macros, lots of one character variable names, etc, I wonder how easy it would be to break this. That's before you even start with the same name used for different things in different scopes!
Reminds me of the wonderful idea that good code does not require comments, since it is clear, and comments are to explain things that are not obvious. Still not sold on this idea, but I like the theory  |
| Uniwares |
Posted - Aug 18 2020 : 4:50:07 PM
Should not be too hard to analyze a method and create the documentation from this info. Would be typical VA functionality.
Guess its out with the next release, right? |
| accord |
Posted - Aug 18 2020 : 3:56:20 PM
While it looks fancy, it doesn't really add any additional information which isn't immediately obvious from the method signature, in my opinion.
Now, risking to be a bit futuristic, this is what I would imagine to be generated by a neural network that was trained on a huge amount of method body and their comment pairs. GPT-3 and other approaches show promise with similar tasks, but this is just an idea, not a promise that it will be implemented Since it's not only highly experimental at this point but also requires unreal amount of computational capacity for the training. |
