How would I parse a flat Ruby hash in to a nested hash?

Question

I have the following flat Ruby hash:

{ 
  "builder_rule_0_filter"=>"artist_name", 
  "builder_rule_0_operator"=>"contains", 
  "builder_rule_0_value_0"=>"New Found Glory", 

  "builder_rule_1_filter"=>"bpm", 
  "builder_rule_1_operator"=>"less", 
  "builder_rule_1_value_0"=>"150", 

  "builder_rule_2_filter"=>"days_ago", 
  "builder_rule_2_operator"=>"less", 
  "builder_rule_2_value_0"=>"40", 

  "builder_rule_3_filter"=>"release_date_start", 
  "builder_rule_3_operator"=>"between", 
  "builder_rule_3_value_0"=>"2019-01-01", 
  "builder_rule_3_value_1"=>"2019-12-31", 
}

I'd like to convert it to a nested hash that's a bit more organized/useable:

{
  "0"=>
    {
      "filter"=>"artist_name",
      "operator"=>"contains",
      "values"=>
      {
        "0"=>"New Found Glory"
      }
    },
  "1"=>
    {
      "filter"=>"bpm",
      "operator"=>"less",
      "values"=>
      {
        "0"=>"150"
      }
    }
  "2"=>
    {
      "filter"=>"days_ago",
      "operator"=>"less",
      "values"=>
      {
        "0"=>"40"
      }
    }
  "3"=>
    {
      "filter"=>"release_date_start",
      "operator"=>"between",
      "values"=>
      {
        "0"=>"2019-01-01"
        "1"=>"2019-12-31"
      }
    }
}

So, how can convert a flat hash (which I'm getting from a form) and convert it to be nested based on on those key names?

Answer 1

First you can split up the hash into multiple hashes according to their prefix number (feel free to run this to see the return val)

groups = input.
  group_by { |k,v| k.match(/builder_rule_(\d+)/)[1] }.
  transform_values(&:to_h)

at this point, creating the inner objects is easier if you just use some hard-coding to build the key-valls:

result = groups.each_with_object({}) do |(prefix, hash), memo|
  memo[prefix] = {
    "filter" => hash["builder_rule_#{prefix}_filter"],
    "operator" => hash["builder_rule_#{prefix}_operator"],
    "values" => hash.select do |key, val|
      key =~ /builder_rule_#{prefix}_value/
    end.sort_by { |key, val| key }.map { |(key, val)| val }
  }
end

It might possibly be confusing what .sort_by { |key, val| key }.map { |(key, val)| val } means. I can spell it out:

• hash.select { |key, val| key =~ /builder_rule_#{prefix}_value/ } gets the key-vals which are going to be used for the "values" array. It returns a hash.
• .sort_by { |key, val| key } turns the hash into an array of [key, val] tuples , sorted by the key. This is so that the values appear in the correct order.
• .map { |(key, val)| val } turns the nested array into a single-level array, discarding the keys

You could also use sort_by(&:first).map(&:second), though you need active support to use Array#second

Answer 2

The accepted answer is good, but does a bit more work than is strictly necessary. I would do all of the work with each_with_object:

input.each_with_object({}}) do |(key, val), hsh|
  _, num, name = *key.match(/^builder_rule_(\d+)_(.*)/)

  hsh[num] = {} unless hsh.key?(num)

  if name.start_with?(/value_\d/)
    hsh[num]["values"] = [] unless hsh[num].key?("values")
    hsh[num]["values"] << val
  else
    hsh[num][name] = val
  end
end
# => {
#      "0" => {
#        "filter" => "artist_name",
#        "operator" => "contains",
#        "values" => ["New Found Glory"]
#      },
#      "1" => {
#        "filter" => "bpm",
#        "operator" => "less",
#        "values" => ["150"]
#      },
#      "2" => {
#        "filter" => "days_ago",
#        "operator" => "less",
#        "values" => ["40"]
#      },
#      "3" => {
#        "filter" => "release_date_start",
#        "operator" => "between",
#        "values" => ["2019-01-01", "2019-12-31"]
#      }
#    }

See it in action on repl.it: https://repl.it/@jrunning/GiddyEnragedLinks

Answer 3

Assuming the depth is fixed at 3 and the rules are consistent (as shown in the example), you can make this a single pass transformation as follows:

h.each_with_object(Hash.new {|h,k| h[k]= {}}) do |(k,v),obj|
  new_k, k2, k3 = k[/(?<=builder_rule_).*/].split('_')
  k2 == 'value' ? (obj[new_k]["values"] ||= {}).merge!({k3 => v}) :  obj[new_k][k2] = v
end

Here we are constructing a new Hash where first level keys default to an empty Hash (h.each_with_object(Hash.new {|h,k| h[k]= {}}) do |(k,v),obj|).

Then we split the key based on the underscore after 'builder_rule_' new_k, k2, k3 = k[/(?<=builder_rule_).*/].split('_')

Then if the second level key (k2) equals 'value' we need to make it values and merge the third level key (k3) and the value into a new hash otherwise we assign k2 to the value. k2 == 'value' ? (obj[new_k]["values"] ||= {}).merge!({k3 => v}) : obj[new_k][k2] = v

This will actually return the requested requirement where values is a Hash with key indexes

{
  "0"=>{
      "filter"=>"artist_name",
      "operator"=>"contains",
      "values"=>
      {
        "0"=>"New Found Glory"
      }
    },
  "1"=>{
      "filter"=>"bpm",
      "operator"=>"less",
      "values"=>
      {
        "0"=>"150"
      }
    },
  "2"=>{
      "filter"=>"days_ago",
      "operator"=>"less",
      "values"=>
      {
        "0"=>"40"
      }
    },
  "3"=>{
      "filter"=>"release_date_start",
      "operator"=>"between",
      "values"=>
      {
        "0"=>"2019-01-01",
        "1"=>"2019-12-31"
      }
    }
}