From 6550828ccef29074eacfe5fe6ff29618661da1d5 Mon Sep 17 00:00:00 2001
From: Boon Low <boon.low@mac.com>
Date: Wed, 21 Nov 2018 17:21:13 +0000
Subject: [PATCH] provide a `sort` boolean option to enable/disable ranking

---
 README.md                    | 19 ++++++++++++++++++-
 lib/information_retrieval.ex | 22 ++++++++++++++--------
 test/search_test.exs         |  8 ++++++++
 3 files changed, 40 insertions(+), 9 deletions(-)

diff --git a/README.md b/README.md
index 63fa440..715f901 100644
--- a/README.md
+++ b/README.md
@@ -50,7 +50,7 @@ a corpus and index of 1000 (max) docs from the
 CSV dataset if pre-created index / corpus are not supplied.
 The query will be issued on this small index.
 
-Ranked docs (ids) with scores are currently being returned.
+Doc ids and scores are currently being returned as results.
 
 ```elixir
   # quick search test with up to 1000 max docs
@@ -77,6 +77,23 @@ Ranked docs (ids) with scores are currently being returned.
   iex> IR.q "van eyck", index: index, corpus: corpus, op: :and
 
 
+```
+
+Ranking and sorting of results can be toggled with the `:sort` option.
+
+```elixir
+  iex> IR.q "christopher columbus carlo eyck galileo galilei", sort: false
+  Indexing..
+  Found 5 results.
+  [{1, 1.25276}, {4, 0.55962}, {5, 0.55962}, {6, 0.55962}, {7, 5.01105}]
+
+  # ranking with relevancy
+  iex(6)> IR.q "christopher columbus carlo eyck galileo galilei", sort: true
+  Indexing..
+  Found 5 results.
+  [{7, 5.01105}, {1, 1.25276}, {4, 0.55962}, {5, 0.55962}, {6, 0.55962}]
+
+
 ```
 
 ## CSV data
diff --git a/lib/information_retrieval.ex b/lib/information_retrieval.ex
index f3adae5..153b7f3 100644
--- a/lib/information_retrieval.ex
+++ b/lib/information_retrieval.ex
@@ -167,6 +167,7 @@ defmodule IR do
   - `:op` - default `:or`, match ALL (`:and`) or ANY (`:or`) terms in the query
   - `:corpus`, parsed data required for results display and ranking purposes
   - `:index`, pre-created search data for querying and ranking purposes
+  - `:sort` default `true`, rank results by relevancy
 
   ### Example
 
@@ -199,7 +200,7 @@ defmodule IR do
   ```
   """
   @spec q(binary, keyword) :: list[binary]
-  def q(query, opts \\ [index: nil, corpus: nil, op: :or])
+  def q(query, opts \\ [index: nil, corpus: nil, op: :or, sort: true])
   def q(query, opts) do
     op = if opts[:op], do: opts[:op], else: :or
 
@@ -207,14 +208,14 @@ defmodule IR do
     # if no index / corpus are provided
     if is_nil(opts[:index]) or is_nil(opts[:corpus]) do
       {:ok, index, corpus} = indexing(1000, corpus: true)
-      q(query, index, corpus, op)
+      q(query, index, corpus, op, opts[:sort])
     else
-      q(query, opts[:index], opts[:corpus], op)
+      q(query, opts[:index], opts[:corpus], op, opts[:sort])
     end
   end
 
   @doc false
-  def q(query, index, corpus, op) do
+  def q(query, index, corpus, op, sort) do
     terms = query |> analyse
     posting_sets = terms |> Enum.map(&(index[&1]))
 
@@ -244,11 +245,16 @@ defmodule IR do
     #
     # next: could derive scores by computing cosine similarity 
     # between doc vectors and search term vector
-    ranked_ids_with_scores = term_doc_matrix 
-    |> Enum.map(fn {doc_id, vector} -> {doc_id, Enum.sum(vector) |> Float.round(5)} end) 
-    |> Enum.sort_by(&(elem(&1,1)),  &>=/2)
+    unranked_ids_with_scores = term_doc_matrix
+    |> Enum.map(fn {doc_id, vector} -> {doc_id, Enum.sum(vector) |> Float.round(5)} end)
+
+    case sort do
+      true ->
+        unranked_ids_with_scores |> Enum.sort_by(&(elem(&1,1)),  &>=/2)
+      false -> unranked_ids_with_scores
+      _ -> unranked_ids_with_scores 
+    end
 
-    ranked_ids_with_scores
   end
 
   # single keyword postings
diff --git a/test/search_test.exs b/test/search_test.exs
index 0f1f712..79b6498 100644
--- a/test/search_test.exs
+++ b/test/search_test.exs
@@ -46,4 +46,12 @@ defmodule IRSearchTest do
     assert doc_ids == []
   end
 
+  test "ranking keywords search results", %{index: index, corpus: corpus} do
+    doc_ids = IR.q "christopher columbus carlo eyck galileo galilei", index: index, corpus: corpus, sort: false
+    assert doc_ids |> Enum.map(&(elem(&1,0))) == [1, 4, 5, 6, 7]
+
+    doc_ids = IR.q "christopher columbus carlo eyck galileo galilei", index: index, corpus: corpus, sort: true
+    assert doc_ids |> Enum.map(&(elem(&1,0))) == [7, 1, 4, 5, 6]
+  end
+
 end