Getting (+-) instead of diff (tilde) when comparing arrays

[stripedpumpkin]

Hello. I find the following behavior (using Hashdiff 1.0.1) somewhat unexpected:

Hashdiff.diff(%w[a b c], %w[a b d])
=> [["-", "[2]", "c"], ["+", "[2]", "d"]]

Would not it be better to get

Hashdiff.diff(%w[a b c], %w[a b d])
=> [["~", "[2]", "c", "d"]]

?

Is there an option to get this behavior ?

Thanks for reading!

[liufengyun]

I think it's possible to achieve that. It will be much appreciated if someone can give a hand on this.

[krzysiek1507]

This does it:

Hashdiff.diff(%w[a b c], %w[a b d], use_lcs: false)

[liufengyun]

This does it:

Hashdiff.diff(%w[a b c], %w[a b d], use_lcs: false)

This is a workaround. I'm wondering if it is possible to make the algorithm smarter. Conceptually, it seems possible.

[krzysiek1507]

Does LCS algorithm have a strict specification?

[liufengyun]

Does LCS algorithm have a strict specification?

I don't think it has. The tests + implementation are the de facto spec.
In this case, the algorithm performs better, it might be acceptable to users even if it differs from previous results.

[robkiessling]

I am running into this issue and was thinking of just adding a post-process that looks for - and + diffs at the same index and manually converts them into a ~?

In other words I would just loop through a result like:

[["-", "[2]", "c"], ["+", "[2]", "d"]]

And since [2] has both a - and a + diff, I can replace those with a:

["~", "[2]", "c", "d"]

Does this seem like a reasonable workaround or am I missing some edge case?

---

Also, I should add, I want to avoid using the use_lcs: false workaround because I like how LCS handles element shifting. For example:

# without lcs: results in many changes
Hashdiff.diff(%w[a b c d e], %w[b c d e x], use_lcs: false)
=> [["~", "[0]", "a", "b"], ["~", "[1]", "b", "c"], ["~", "[2]", "c", "d"], ["~", "[3]", "d", "e"], ["~", "[4]", "e", "x"]] 

# with lcs: can tell that really just 2 elements changed
Hashdiff.diff(%w[a b c d e], %w[b c d e x], use_lcs: true)
 => [["-", "[0]", "a"], ["+", "[4]", "x"]]