hdiff

Hash-based Diffing for AST's

Monthly Downloads: 2

Programming language: Haskell

License: MIT License

Tags: Other    

Latest version: v0.0.5

README

hdiff: Hash-based Diffing for AST's

This README provides some general info on the library. For detailed information on the algorithm, check our ICFP Paper. If you want even more details, please check my PhD thesis

The hdiff tool computes provides your usual differencing functions:

diff  :: a -> a -> Patch a

apply :: Patch a -> a -> Maybe a

The difference with respect to the other tree differencing tools out there is that HDiff.Patch a is a pattern-expression based patch. A picture is worth 1024 words: Take the following files:

   O.while                 A.while

a := n;                | a := 1;
while n > 0 do {       | while n > 0 do {
  n := n - 1;          |   a := a * n;
  a := a * n;          |   n := n - 1;
}                      | }
res := n;              | res := n;

Running hdiff examples/While/Factorial/{O,A}.while to compute their differences will the following patch:

[0] := n;       -|+  [0] := 1;
while [1] do {  -|+  while [1] do {
  [2];          -|+    [3];
  [3];          -|+    [2];
}               -|+  }
[4];            -|+  [4];

Everything to the left of -|+ is called the pattern, or deletion context. Everything to the right ot -|+ is called the expression, or insertion context. The square brackets denote metavariables, which is how hdiff copies information over. If we overlay the deletion context above over O.while, we will get an assignment of variables that we can use in the insertion context to produce A.while.

In reality, the output is a bit more complicated since it includes information about where each individual change happened. In this case, there are five changes:

1. Copy of identifier a
2. Change n into 1
3. Copy condition of while loop
4. Permute statements inside while loop
5. Copy rest of the prorgam

The actual output is more like the following:

{-# [Cpy #0] #-} := {-# {- n -}{+ 1 +} #-}
while {-# [Cpy #1] #-} do {
  {-# [Prm #2 <=> #3]
      [Prm #3 <=> #2]  #-}
}
{-# [Cpy #4] #-}

Everything inside a {-# ... #-} is a change that happens in its own individual scope. Some changes are copies, some changes involve permutations and some changes delete match on something and insert something else.

But in reality, you will see the following ugly output because we have not implemented pretty-printing yet.

(Seq
 (:
  (Assign {-# [Cpy #7 ] #-} {-# {{- (Var "n") -} {+ (IntConst 1) +}} #-})
  (:
   (While
    {-# [Cpy #4 ] #-}
    (Seq {-# (: [Prm #2 <=> #3 ] (: [Prm #3 <=> #2 ] [])) #-}))
   {-# [Cpy #6 ] #-})))

Just like line-based diff, hdiff can be configured in many ways. Please, refer to hdiff --help for all the modes and options.

To see the result of a merge run with --verbose or -v; to see the patch produced by a merge run with --debug.