Benchmark Dictionary implementation in Swarm

[xsebek]

• add dictionary benchmarks in Swarm language
• add more tests
• fix discovered bugs

You can test it with:

run "example/recursive-containers.sw";
benchmark_tree

Screenshot

• Closes Dictionaries #2159

[xsebek]

@byorgey without import you can copy the relevant part of the header and use it like this:

tydef Maybe a = Unit + a end

tydef IDict d k v =
  [ empty: d
  , insert: k -> v -> d -> d
  , delete: k -> d -> d
  , get: k -> d -> Maybe v
  , contains: k -> d -> Bool
  , pretty: d -> Text
  ]
end
def undefined_dict : any -> IDict any k v =\empty.
    [empty=empty, insert=\x.undefined, delete=\x.undefined, get=\x.undefined, contains=\x.undefined, pretty=\x.undefined]
end

tydef RBTree k = rec b. Maybe [red: Bool, k: k, l: b, r: b] end
tydef Dict k v = RBTree (k * v) end
def tree_dict : IDict (Dict k v) k v = undefined_dict (inl ()) end

run "recursive-containers.sw";
log $ tree_dict.pretty $ tree_dict.insert 1 "b" $ tree_dict.insert 0 "a" $ tree_dict.insert 4 "e" tree_dict.empty

I hoped the records would make this shorter, but they are values, so I can not just declare them undefined.

EDIT: I wonder if IDict (Dict k v) k v is not too obscure - non Java/C# people might not understand the I means interface convention 🤔

[byorgey]

What's the purpose of defining flat_set et al as records with lots of undefined at first, and then later redefining them?

[xsebek]

It's an overview, as it can be hard to spot between the Red-Black Tree implementation and tests.

It also works as a "header" of sorts, if you wanted to copy paste part of it in a scenario definition/other Swarm script and then run this file.

[byorgey]

Scoped type variables! 😍