Introduction

public int add
  ArrayList<St
  int nWidth =
  while (st <
    int height
    if (st * r
      img.chec
    } else if 
...

Contents

Disclaimer: Not much has been done.

Why is Software So Important?

Software controls...

How many had software glitch in the past month?

Software = Too much $$$

• Software costs >50% of the car R&D today.

Software = Life or Death

• Toyota "Unintended Acceleration" circa 2010.
  • Class action: $1.2B.
  • Faulty software on ETCS?
  • 2,272 global variables.
  • 80,000 violations of MISRA C.
  • Over 67 functions "spaghetti".
• Huge U.S. Blackout in 2003.
• WannaCry attacked a UK hospital.
• Therac-25 (killed 3 people).
  

  pixnio.com

Software = So Subtle

Apple's "GotoFail" (2014)

static OSStatus
SSLVerifySignedServerKeyExchange(...)
{
    OSStatus        err;
    ...
    if ((err = SSLHashSHA1.update(&hashCtx, &serverRandom)) != 0)
        goto fail;
    if ((err = SSLHashSHA1.update(&hashCtx, &signedParams)) != 0)
        goto fail;
        goto fail;
    if ((err = SSLHashSHA1.final(&hashCtx, &hashOut)) != 0)
        goto fail;
    ...
 
fail:
    SSLFreeBuffer(&signedHashes);
    SSLFreeBuffer(&hashCtx);
    return err;
}

Problems

• Software maintenance : 60 ~ 80% of cost.
  • "Suppose that you're asked to
    rewrite Halmet in which
    the main character is female."
• Code understanding : 50% ~ of time.
  • Need all the details!
  • Error prone.

Attacking the Monster

• No "Silver Bullet".
• Multi-faceted solutions.
  • Make better tools.
  • Establish better practices.
  • ...
  • Help reducing cognitive burden.

Research Question

a = 100;
b = 2;
c = a + b;
d = a * c;
...

Example 1

if (c == 1) {
    b = b - 100;
    a1 = a1 + 1;
} else if (c == 2) {
    b = b - 300;
    a2 = a2 + 1;
} else
...

• Q. What's the role of variable b?

Example 1

if (choice == 1) {
    money = money - 100;
    item1 = item1 + 1;
} else if (choice == 2) {
    money = money - 300;
    item2 = item2 + 1;
} else
...

• money : Quantity you can exchange with any item.

Example 2

if (x) {
    y = y - 1;
}
if (y == 0) {
    g();
}
...

• Q. What's the role of variable y?

Example 2

if (hit) {
    life = life - 1;
}
if (life == 0) {
    gameover();
}
...

Motral Kombat Wiki

• life : Quantiy that decreases when you're hit.
  When it reaches zero, Game Over.

Basic Idea

• Convert a source code to
  a "Dataflow graph".

  pow(x, y) {
      z = 1;
      while (0 < y) {
          z *= x;
          y -= 1;
      }
      return z;
  }
  

Pattern Matching

• Learn mappings from a large corpus.

  if (c == 1) {
      b = b - 10
      a1 = a1 + 
  } else if (c =
      ...
  

  Apply patterns to the target code.

Current Progress

1. Download top 1,000 GitHub repos.
   (11GB zipped text)
2. Parse .java codes. [Eclipse JDT]
   (480,627 files. 74,088,883 loc.)
3. Store dataflow graphs.
   (4,199,301 functions, 42,133,919 nodes)
4. Implement pattern matching.

Conclusions

• Help programmers to understand code.
• Obtain patterns from existing code.
• Assign meaning to program variables.

References

1. This Car Runs on Code, IEEE Specrum, 2009
2. Koopman, "A Case Study of Toyota Unintended Acceleration and Software Safety", 2014
3. Glass, Facts and Fallacies of Software Engineering, Addison-Wesley, 2002
4. Eclipse Java development tools (JDT)