DeepDive

This library should serve the simple aim of finding keys in deeply nested and complex data structures used in other elixir projects. The use case for which it has been conceived was to explore the depths of the Absinthe resolution structs, but may be useful in any such case when the strutures at hand are complex and when such structures are built at runtime by some intricate logic. It is intended solely as a debugging utility, and not as a production-level tool.

Premise

To understand the logic of the exploration algorithms offered, let's point out that any nested structure we are talking about may be thought of as a tree. The nodes of the tree are either the keys in a map (or a struct), or the indices of a list. The leaves of such tree are the nodes that have no more descendants. Two nodes have the same level if they have the same number of parent nodes up to the root. A set of nodes are in the same branch if they either have the same set of parent nodes up to the root or one is ancestor of the other. In the former case, they are also at the same level.

Elixir data structures and trees

For what concerns this small utility, the elixir data structure correspondents of the elements of an abstract tree are maps (and structs treated as such) and lists (with keyword lists as a special case, that is treated as maps). Any nested combination of such structures may be explored with this library.

Any key-value relationship represents descending a level, and any keys in the same structure are on the same branch (so the branching begins after the key-value jump). This is important to notice because, in the case of the DeepDive.FirstFound algorithm that is described below, if a key is found on a level (i.e. in a map or a keyword list), the search is halted for all the children branches attached to other keys (and this might sound a bit unexpected).

What's inside

It currently offers two different modules that implement a different exploration strategy:

• DeepDive.FirstFound starts a depth-first exploration and, whenever a key matches, adds it to the list of matches and drops the search on that branch.
• DeepDive.FullWalk proceeds as the above, but does not stop on first match, rather on arriving at the leaves.

Both these strategies complete when they have explored the whole tree.

In action

Take the following nested map

iex> test_map = %{
  a: %{
    b: %{
      c: 1,
      d: [
        %{e: 1},
        %{e: 2},
        %{n: %{o: %{p: %{e: 3}, o: "found"}}},
        %{e: 4},
        %{e: 5},
        %{o: {1, 2}}
      ]
    },
    f: %{g: %{e: 6}},
    h: %{i: %{j: %{k: %{e: 7}, l: %{m: %{e: 8}}}, e: 8}}
  }
}

Find, on any branch, the first occurence of the key :e

iex> DeepDive.FirstFound.find_key(atom_map, :e)
[
  {[:a, :h, :i, :e], 8},
  {[:a, :f, :g, :e], 6},
  {[:a, :b, :d, 4, :e], 5},
  {[:a, :b, :d, 3, :e], 4},
  {[:a, :b, :d, 2, :n, :o, :p, :e], 3},
  {[:a, :b, :d, 1, :e], 2},
  {[:a, :b, :d, 0, :e], 1}
]

Find all the occurences, instead

iex> DeepDive.FullWalk.find_key(atom_map, :e)
[
  {[:a, :h, :i, :e], 8},
  {[:a, :h, :i, :j, :l, :m, :e], 8},
  {[:a, :h, :i, :j, :k, :e], 7},
  {[:a, :f, :g, :e], 6},
  {[:a, :b, :d, 4, :e], 5},
  {[:a, :b, :d, 3, :e], 4},
  {[:a, :b, :d, 2, :n, :o, :p, :e], 3},
  {[:a, :b, :d, 1, :e], 2},
  {[:a, :b, :d, 0, :e], 1}
]

The key might also be a Regex or a function ((term -> bool)). This in order to allow a fuzzy matching, in case the exact key is not known.

Peek into the tests to see more examples in action.

Installation

If available in Hex, the package can be installed by adding deep_dive to your list of dependencies in mix.exs:

def deps do
  [
    {:deep_dive, "~> 1.0.0", only: :dev}
  ]
end

Documentation can be found at https://hexdocs.pm/deep_dive.

Developing

Clone this repo and change what you want. If introducing new features or modifying existing ones, please act on the tests accordingly. Always run mix credo and mix dialyzer before opening a PR.