Your best Buddies

Learning about the project

How can we learn about the project/issue/library?

Coworkers knowledge
Documentation
Issue Tracking System
Version Control System
Source code

Coworkers knowledge

Who should I even ask for help?

Spoiler alert: watch out for Context Buddy!

Documentation

Issue Tracking System

Some tickets already archived
Too many different issues to find something useful
Vague issue description
No information about the solution
Sometimes not connected to the source code at all (no MR link)

Version Control System

$ git log --all --pretty=format:"%h" \
> --grep='\(^\| \)\+\(\(refactor\)\|\(rename\)\|\(move\)\|\(remove\)\)' | wc -l
2253

$ git log --all --pretty=format:"%h" --grep='^fix$' | wc -l
118

$ git log --all --pretty=format:"%h" | wc -l
53420

Spoiler alert: watch out for Context Buddy!

Source code

import cats.data.OptionT
import cats.instances.future._

def findUserById(userId: Long): OptionT[Future, User] = { /* ... */ }

def findAccountById(accountId: Long): OptionT[Future, Account] = { /* ... */ }

def getReservedFundsForAccount(account: Account): OptionT[Future, Money] = { /* ... */ }

def getReservedFundsForUser(userId: Long): OptionT[Future, Money] = for {
  user <- findUserById(userId)
  account <- findAccountById(user.accountId)
  funds <- getReservedFundsForAccount(account)
} yield funds

Source code - reality

scanf ("%d",&steps);
for (count = 0; count < steps; count++){
    static int bbk = 0, bk = 1;
    int k = bbk + bk;
    bbk = bk; bk = k;
}

Poorly structured
5k lines long files
Complicated dependencies

Spoiler alert: watch out for Graph Buddy!

It all boils down to working effectively with the source code.

IDEs evolution

Turbo Pascal - 1992

Pajączek 1997

Eclipse 2001

Intellij 2012

Intellij 2019

What is really changing?

Various integrations straight in IDE
Language Server Protocol adoption (multi-language support in an easy way)
Advanced language support
- refactoring
- smart completion
- suggestions (recently based also on AI)

Improving existing tools and bringing more context closer to the source code

Still a lot of places to improve...

contextbuddy.virtuslab.com

Meet Graph Buddy

Idea

//File A.scala
package com.virtuslab.graphbuddy
class A(a: String)

//File B.scala
package com.virtuslab.graphbuddy
class B(b: String, a: A)

AST + SemanticDB *

AST (scala trees, scalac AST)
- information about the code structure
SemanticDB:
- Unique stable symbols across the source code
- Additional semantic data (signatures, types, modifiers, return types and more)

* Or we can just ask Context Buddy

Abstract Semantic Graph

Abstract syntax in which expression of a programming language is represented by a graph
Higher level of abstraction than AST

Bird's eye view

Never get lost again

Never get lost again (VSCode integration)

Appropriate scale

Find the path

Possible use cases

Learning about the project
- Project visualisation
- Project evolution visualisation
Refactoring (Rory Graves, ScalaClean – deep code analysis)
- Dead code elimination
- Automatic module extraction
- Complex dependency elimination
Maintaining
- Anomaly detection (automatic with CI)
- Architecture advices

Context Buddy:
The semantic platform

Provides simple API on top of languages and build tools
Design for cross-language usage (no AST)
Model is flexible enough to support non-code data (logs, unit tests results and more)

Context Buddy:
The context platform

Our first goal is to suggest experts to talk to
Next step: Reviewers suggestion [Master Thesis in progress]
We want to learn from local mistakes and development
Knowledge base about the project is ultimate goal