Release v0.1.4: Bump dependencies and refactor for code quality

Updates the application to version 0.1.4, aligning dependencies with the
latest North Shore AI ecosystem releases and addressing static analysis
findings.

Dependency Updates:
- Update crucible_datasets to ~> 0.5.3 (pulling in crucible_ir 0.2.0).

Code Quality & Refactoring:
- Refactor EvalEx.Comparison to decompose comparison logic and extract
  helper functions for metric normalization and formatting.
- Update EvalEx.Metrics to prefer Enum.empty?/1 over length/1 checks
  for improved efficiency and style compliance.
- Fix compile warning for undefined EvalEx.Datasets.load/1 usage via
  runtime dispatch.
- Clean up module aliases in EvalEx.Task.Registry and test suites.

This maintenance release focuses on stability, code health, and ensuring
compatibility with updated upstream dataset definitions.

Author: nshkrdotcom

CHANGELOG.md

@@ -5,6 +5,17 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [0.1.4] - 2025-12-25
+
+### Fixed
+
+- Fix compile warning for undefined `EvalEx.Datasets.load/1` by using `apply/3` for runtime dispatch
+- Credo fixes for code quality
+
+### Changed
+
+- Bump `crucible_datasets` dependency to `~> 0.5.3`
+
 ## [0.1.3] - 2025-12-25
 
 ### Changed

README.md

@@ -129,7 +129,7 @@ Add `eval_ex` to your list of dependencies in `mix.exs`:
 ```elixir
 def deps do
   [
-    {:eval_ex, "~> 0.1.3"}
+    {:eval_ex, "~> 0.1.4"}
   ]
 end
 ```

lib/eval_ex/comparison.ex

@@ -34,7 +34,8 @@ defmodule EvalEx.Comparison do
 
   """
   @spec compare(list(Result.t())) :: t()
-  def compare(results) when is_list(results) and length(results) > 0 do
+  def compare(results) when is_list(results) do
+    if Enum.empty?(results), do: raise(ArgumentError, "results cannot be empty")
     metric_comparisons = compare_metrics(results)
     rankings = rank_results(results, metric_comparisons)
     best = determine_best(rankings)
@@ -133,13 +134,7 @@ defmodule EvalEx.Comparison do
             0.0
 
           %{values: values} ->
-            max_value = values |> Enum.map(&elem(&1, 1)) |> Enum.max()
-
-            if max_value > 0 do
-              stats.mean / max_value
-            else
-              0.0
-            end
+            normalize_metric_score(stats.mean, values)
         end
       end)
 
@@ -153,6 +148,38 @@ defmodule EvalEx.Comparison do
   defp determine_best([]), do: nil
   defp determine_best([{best, _score} | _rest]), do: best
 
+  defp normalize_metric_score(mean, values) do
+    max_value = values |> Enum.map(&elem(&1, 1)) |> Enum.max()
+    if max_value > 0, do: mean / max_value, else: 0.0
+  end
+
+  defp build_pairwise_test(r1, r2, metric) do
+    v1 = get_metric_values(r1, metric)
+    v2 = get_metric_values(r2, metric)
+
+    if not Enum.empty?(v1) and not Enum.empty?(v2) do
+      t_stat = calculate_t_statistic(v1, v2)
+
+      %{
+        pair: "#{r1.name} vs #{r2.name}",
+        t_statistic: t_stat,
+        significant: abs(t_stat) > 1.96
+      }
+    else
+      nil
+    end
+  end
+
+  defp format_metric_test({metric, test_results}) do
+    results_str =
+      Enum.map_join(test_results, ", ", fn test ->
+        sig = if test.significant, do: "*", else: ""
+        "#{test.pair}: t=#{Float.round(test.t_statistic, 2)}#{sig}"
+      end)
+
+    "  #{metric}: #{results_str}"
+  end
+
   defp run_statistical_tests(results) when length(results) < 2 do
     %{note: "Need at least 2 results for statistical testing"}
   end
@@ -176,20 +203,7 @@ defmodule EvalEx.Comparison do
     results
     |> Enum.chunk_every(2, 1, :discard)
     |> Enum.map(fn [r1, r2] ->
-      v1 = get_metric_values(r1, metric)
-      v2 = get_metric_values(r2, metric)
-
-      if length(v1) > 0 and length(v2) > 0 do
-        t_stat = calculate_t_statistic(v1, v2)
-
-        %{
-          pair: "#{r1.name} vs #{r2.name}",
-          t_statistic: t_stat,
-          significant: abs(t_stat) > 1.96
-        }
-      else
-        nil
-      end
+      build_pairwise_test(r1, r2, metric)
     end)
     |> Enum.reject(&is_nil/1)
   end
@@ -255,15 +269,7 @@ defmodule EvalEx.Comparison do
     if Map.has_key?(tests, :note) do
       "  #{tests.note}"
     else
-      Enum.map_join(tests, "\n", fn {metric, test_results} ->
-        results_str =
-          Enum.map_join(test_results, ", ", fn test ->
-            sig = if test.significant, do: "*", else: ""
-            "#{test.pair}: t=#{Float.round(test.t_statistic, 2)}#{sig}"
-          end)
-
-        "  #{metric}: #{results_str}"
-      end)
+      Enum.map_join(tests, "\n", &format_metric_test/1)
     end
   end
 

lib/eval_ex/crucible.ex

@@ -55,8 +55,7 @@ defmodule EvalEx.Crucible do
   end
 
   defp submit_to_crucible(experiment_data) do
-    # TODO: Implement actual Crucible integration
-    # This would call Crucible.Experiment.create/1 or similar
+    # NOTE: Placeholder - production would call Crucible.Experiment.create/1 or similar
     # For now, we just log
     log_experiment(experiment_data)
     {:ok, :submitted}

lib/eval_ex/metrics.ex

@@ -94,7 +94,7 @@ defmodule EvalEx.Metrics do
     n = min(max_n, min(length(pred_tokens), length(truth_tokens)))
 
     if n == 0 do
-      if length(pred_tokens) == 0 and length(truth_tokens) == 0 do
+      if Enum.empty?(pred_tokens) and Enum.empty?(truth_tokens) do
         1.0
       else
         0.0
@@ -108,7 +108,7 @@ defmodule EvalEx.Metrics do
           common = MapSet.intersection(MapSet.new(pred_ngrams), MapSet.new(truth_ngrams))
           common_count = MapSet.size(common)
 
-          if length(pred_ngrams) == 0 do
+          if Enum.empty?(pred_ngrams) do
             0.0
           else
             common_count / length(pred_ngrams)
@@ -138,11 +138,14 @@ defmodule EvalEx.Metrics do
 
     lcs_length = lcs(pred_tokens, truth_tokens)
 
-    if length(pred_tokens) == 0 and length(truth_tokens) == 0 do
+    pred_len = length(pred_tokens)
+    truth_len = length(truth_tokens)
+
+    if pred_len == 0 and truth_len == 0 do
       1.0
     else
-      precision = if length(pred_tokens) > 0, do: lcs_length / length(pred_tokens), else: 0.0
-      recall = if length(truth_tokens) > 0, do: lcs_length / length(truth_tokens), else: 0.0
+      precision = if pred_len > 0, do: lcs_length / pred_len, else: 0.0
+      recall = if truth_len > 0, do: lcs_length / truth_len, else: 0.0
 
       if precision + recall > 0 do
         2 * (precision * recall) / (precision + recall)
@@ -160,7 +163,7 @@ defmodule EvalEx.Metrics do
   """
   @spec entailment(term(), term()) :: float()
   def entailment(prediction, ground_truth) do
-    # TODO: Integrate with actual NLI model (DeBERTa-v3)
+    # NOTE: Placeholder - production would integrate with NLI model (DeBERTa-v3)
     # For now, use token overlap as proxy
     f1(prediction, ground_truth)
   end
@@ -185,7 +188,7 @@ defmodule EvalEx.Metrics do
     # For structured predictions, check citation fields
     citations = Map.get(prediction, :citations, Map.get(prediction, "citations", []))
 
-    if is_list(citations) and length(citations) > 0 do
+    if is_list(citations) and not Enum.empty?(citations) do
       # Validate citations exist in ground truth evidence
       truth_evidence = Map.get(ground_truth, :evidence, Map.get(ground_truth, "evidence", []))
       validate_citations(citations, truth_evidence)
@@ -262,8 +265,10 @@ defmodule EvalEx.Metrics do
       matches = count_matches(pred_tokens, truth_tokens)
 
       # Calculate precision and recall
-      precision = if length(pred_tokens) > 0, do: matches / length(pred_tokens), else: 0.0
-      recall = if length(truth_tokens) > 0, do: matches / length(truth_tokens), else: 0.0
+      pred_len = length(pred_tokens)
+      truth_len = length(truth_tokens)
+      precision = if pred_len > 0, do: matches / pred_len, else: 0.0
+      recall = if truth_len > 0, do: matches / truth_len, else: 0.0
 
       # Calculate F-mean with higher weight on recall
       if precision + recall > 0 do
@@ -312,7 +317,7 @@ defmodule EvalEx.Metrics do
   """
   @spec bert_score(term(), term()) :: map()
   def bert_score(prediction, ground_truth) do
-    # TODO: Integrate with actual BERT/transformer model
+    # NOTE: Placeholder - production would integrate with BERT/transformer model
     # For now, return token-based similarity as placeholder
     similarity = f1(prediction, ground_truth)
 
@@ -438,10 +443,10 @@ defmodule EvalEx.Metrics do
         Enum.any?(evidence, fn ev -> matches_citation?(ev, citation_id) end)
       end)
 
-    if length(citations) > 0 do
-      valid_count / length(citations)
-    else
+    if Enum.empty?(citations) do
       0.0
+    else
+      valid_count / length(citations)
     end
   end
 

lib/eval_ex/runner.ex

@@ -41,9 +41,11 @@ defmodule EvalEx.Runner do
   defp get_ground_truth(evaluation, opts) do
     case Keyword.get(opts, :ground_truth) do
       nil ->
-        # Try to load from dataset
-        if Code.ensure_loaded?(EvalEx.Datasets) do
-          apply(EvalEx.Datasets, :load, [evaluation.dataset()])
+        # Build module name dynamically to avoid compile-time warning on optional module
+        datasets_module = Module.concat([EvalEx, Datasets])
+
+        if Code.ensure_loaded?(datasets_module) do
+          datasets_module.load(evaluation.dataset())
         else
           {:error, :no_ground_truth}
         end

lib/eval_ex/task/registry.ex

@@ -25,6 +25,8 @@ defmodule EvalEx.Task.Registry do
 
   use GenServer
 
+  alias EvalEx.Task.Definition
+
   @doc """
   Starts the registry GenServer.
 
@@ -111,8 +113,8 @@ defmodule EvalEx.Task.Registry do
     registry = Keyword.get(opts, :registry, __MODULE__)
 
     case GenServer.call(registry, {:get, task_id}) do
-      {:ok, %EvalEx.Task.Definition{} = defn} ->
-        {:ok, EvalEx.Task.Definition.invoke(defn, args)}
+      {:ok, %Definition{} = defn} ->
+        {:ok, Definition.invoke(defn, args)}
 
       {:ok, module} when is_atom(module) ->
         {:ok, EvalEx.Task.from_module(module)}
@@ -126,7 +128,7 @@ defmodule EvalEx.Task.Registry do
   def init(_), do: {:ok, %{}}
 
   @impl true
-  def handle_call({:register, %EvalEx.Task.Definition{} = definition}, _from, state) do
+  def handle_call({:register, %Definition{} = definition}, _from, state) do
     {:reply, :ok, Map.put(state, definition.name, definition)}
   end
 

mix.exs

@@ -1,7 +1,7 @@
 defmodule EvalEx.MixProject do
   use Mix.Project
 
-  @version "0.1.3"
+  @version "0.1.4"
   @source_url "https://github.com/North-Shore-AI/eval_ex"
 
   def version, do: @version
@@ -60,7 +60,7 @@ defmodule EvalEx.MixProject do
   defp deps do
     [
       {:jason, "~> 1.4"},
-      {:crucible_datasets, "~> 0.5.2"},
+      {:crucible_datasets, "~> 0.5.3"},
       {:statistics, "~> 0.6"},
       {:ex_doc, "~> 0.31", only: :dev, runtime: false},
       {:credo, "~> 1.7", only: [:dev, :test], runtime: false},

mix.lock

@@ -1,8 +1,8 @@
 %{
   "bunt": {:hex, :bunt, "1.0.0", "081c2c665f086849e6d57900292b3a161727ab40431219529f13c4ddcf3e7a44", [:mix], [], "hexpm", "dc5f86aa08a5f6fa6b8096f0735c4e76d54ae5c9fa2c143e5a1fc7c1cd9bb6b5"},
   "credo": {:hex, :credo, "1.7.14", "c7e75216cea8d978ba8c60ed9dede4cc79a1c99a266c34b3600dd2c33b96bc92", [:mix], [{:bunt, "~> 0.2.1 or ~> 1.0", [hex: :bunt, repo: "hexpm", optional: false]}, {:file_system, "~> 0.2 or ~> 1.0", [hex: :file_system, repo: "hexpm", optional: false]}, {:jason, "~> 1.0", [hex: :jason, repo: "hexpm", optional: false]}], "hexpm", "12a97d6bb98c277e4fb1dff45aaf5c137287416009d214fb46e68147bd9e0203"},
-  "crucible_datasets": {:hex, :crucible_datasets, "0.5.2", "4010e9dda67760a301c215895019874fa3ca9572031996c778995d7edb24801a", [:mix], [{:crucible_ir, "~> 0.1.1", [hex: :crucible_ir, repo: "hexpm", optional: false]}, {:jason, "~> 1.4", [hex: :jason, repo: "hexpm", optional: false]}, {:telemetry, "~> 1.3", [hex: :telemetry, repo: "hexpm", optional: false]}], "hexpm", "37f500e303db812829378240feeb2f8ac0e2e8b7e5cd86c2053d7bd82a1cbeeb"},
-  "crucible_ir": {:hex, :crucible_ir, "0.1.1", "040e09f3048955b3c1b9f933a1320543906bfc4467672fb4cd219d958b856c37", [:mix], [{:jason, "~> 1.4", [hex: :jason, repo: "hexpm", optional: false]}], "hexpm", "baf8235ed1b50caf6235f11a7dd4b6961d9a86300180372691998342ebc8a637"},
+  "crucible_datasets": {:hex, :crucible_datasets, "0.5.3", "abc8aef69106be062649cc24cb8d59578a4845a948465a2ff8a278be080771f0", [:mix], [{:crucible_ir, "~> 0.2.0", [hex: :crucible_ir, repo: "hexpm", optional: false]}, {:jason, "~> 1.4", [hex: :jason, repo: "hexpm", optional: false]}, {:telemetry, "~> 1.3", [hex: :telemetry, repo: "hexpm", optional: false]}], "hexpm", "0f801ca4ceb2fba5eec7c9dc59b71c8dfa4efcd0d8e92b4893c6c3f5001c71d9"},
+  "crucible_ir": {:hex, :crucible_ir, "0.2.0", "f4ef6a7a8a84fa6c1afe0c5efac267e5bcf0149d76e6fe820b2efc668b099d42", [:mix], [{:jason, "~> 1.4", [hex: :jason, repo: "hexpm", optional: false]}], "hexpm", "276adb7fc77d9cd862cd10abdad398d3bdc89384f944ad7118b7b23c985ec5ec"},
   "dialyxir": {:hex, :dialyxir, "1.4.7", "dda948fcee52962e4b6c5b4b16b2d8fa7d50d8645bbae8b8685c3f9ecb7f5f4d", [:mix], [{:erlex, ">= 0.2.8", [hex: :erlex, repo: "hexpm", optional: false]}], "hexpm", "b34527202e6eb8cee198efec110996c25c5898f43a4094df157f8d28f27d9efe"},
   "earmark_parser": {:hex, :earmark_parser, "1.4.44", "f20830dd6b5c77afe2b063777ddbbff09f9759396500cdbe7523efd58d7a339c", [:mix], [], "hexpm", "4778ac752b4701a5599215f7030989c989ffdc4f6df457c5f36938cc2d2a2750"},
   "erlex": {:hex, :erlex, "0.2.8", "cd8116f20f3c0afe376d1e8d1f0ae2452337729f68be016ea544a72f767d9c12", [:mix], [], "hexpm", "9d66ff9fedf69e49dc3fd12831e12a8a37b76f8651dd21cd45fcf5561a8a7590"},

test/eval_ex/task_test.exs

@@ -1,7 +1,7 @@
 defmodule EvalEx.TaskTest do
   use ExUnit.Case, async: true
 
-  alias EvalEx.{Task, Sample, Scorer.ExactMatch}
+  alias EvalEx.{Sample, Scorer.ExactMatch, Task}
 
   describe "new/1" do
     test "creates task with required fields" do