Add a disjoint sets (union-find) utility #6797
Conversation
This will be used in an upcoming type optimization pass and may be generally useful.
src/support/disjoint_sets.h
Outdated
|
|
||
| namespace wasm { | ||
|
|
||
| // A disjoint set forest (a.k.a. union-find) implementation. |
There was a problem hiding this comment.
| // A disjoint set forest (a.k.a. union-find) implementation. | |
| // A disjoint set forest (a.k.a. union-find) implementation. See | |
| // https://en.wikipedia.org/wiki/Disjoint-set_data_structure |
| info[root2].parent = root1; | ||
| // If the ranks were equal, the new root has a larger rank. | ||
| if (info[root1].rank == info[root2].rank) { | ||
| ++info[root1].rank; |
There was a problem hiding this comment.
| ++info[root1].rank; | |
| info[root1].rank++; |
Unless there is some reason for it? To me the default ++ reads more clearly in general.
| constexpr size_t count = 16; | ||
| DisjointSets sets; | ||
| size_t elems[count]; | ||
| for (size_t i = 0; i < count; ++i) { |
There was a problem hiding this comment.
Also, can we standardize on i++ for such increments?
There was a problem hiding this comment.
I would prefer to standardize on ++i. For integers it obviously doesn't matter, but for nontrivial iterators it can make a big difference because postincrement requires making a copy of the iterator while preincrement does not.
There was a problem hiding this comment.
In general that is true, but in a for loop like this, the output of ++i / i++ is not being read? The next time i is read is in the condition check, which reads the updated i anyhow.
There was a problem hiding this comment.
Anyhow, I don't mean to block this PR on it, but I think it's separately worth making a choice here.
This will be used in an upcoming type optimization pass and may be
generally useful.