Giters

avinassh / rockstar

Makes you a Rockstar C++ Programmer in 2 minutes

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Make real code commits instead of random string

avinassh opened this issue · comments

commented

Currently the code commits random strings in code file and in commit message (uuid.uuid1). Instead it should write some real code and commit it with an intelligent commit message.

here's how I am thinking to implement: find some popular libraries (with MIT License), maintain a directory of those snippets and copy the code from it with commit message.

Idea originally suggested by Laurentvw on HN.

commented

I have no idea how RNN works (not cs rockstar yet :/)

Let me know if someone wants to take this task.

commented

@ybur-yug will be implementing this.

commented

I think a neural network is a bit of overengineering something thats such a simple toy application @cscorley. I'll be working on a keep-things-simple solution today or tomorrow (depends when I have a the time).

commented

What if you just grab code snippets from stackoverflow?

commented

@avinassh Whai if we could grab code snippets from different websites on run time and then commit them. That would be much cooler like a rockstar no?

commented

@rahulxxarora there will be licensing issues. but it would be really really cool actually.

@jaredly cool idea! 😎 However we need to make sure we are following the CC License correctly 0, 1

commented

Looks like the code snippets are fine to use under "fair use": http://meta.stackexchange.com/questions/25956/what-is-up-with-the-source-code-license-on-stack-overflow/25957#25957

That said, a snippet of code falls under excerpt category and thus should be free to use under fair use.

  • Jeff Atwood
commented

Guys any update?

commented

RNN, AI, snippets, that's too much work for a rockstar.

commented

@avinassh @cscorley @ybur-yug I am aware of how DL models work, especially RNN. If you are working on this task, let me know. I can help.

BTW, this repo is some awesome shit I found on git. Great Job!!

commented

What we are looking at it something called code generation using DL. Have a look here - https://ml4code.github.io/base-taxonomy/generative.html

I have previously worked on RNNs for language encoding and QA systems. Let me know the details and we will plan accordingly.

commented

Also, if you don't want to use DL, we can generate random senseless code with a simple copy-pasting logic with language specific simple parsing. Any votes on this?