Merge branch '__rultor'

.github/workflows/plantuml.yml

@@ -0,0 +1,26 @@
+name: plantuml
+on:
+  push:
+    paths:
+      - '**.puml'
+    branches:
+      - master
+permissions:
+  contents: write
+jobs:
+  plantuml:
+    runs-on: ubuntu-22.04
+    steps:
+      - name: Checkout Source
+        uses: actions/checkout@v4
+      - name: Generate SVG Diagrams
+        uses: holowinski/plantuml-github-action@main
+        with:
+          args: -v -tsvg doc/*.puml
+      - name: Commit changes
+        uses: EndBug/add-and-commit@v9
+        with:
+          author_name: ${{ github.actor }}
+          author_email: ${{ github.event.pusher.email }}
+          message: 'Diagram generated'
+          add: 'doc/*'

doc/integration.puml

@@ -0,0 +1,14 @@
+@startuml
+title LinerModel 0pdd Integration
+participant "Git Repo" as repo
+participant 0pdd
+participant LinerModel as lm
+
+0pdd -> repo
+repo --> 0pdd: .0pdd.yml
+alt model: true
+  0pdd -> lm: Puzzles
+  lm --> 0pdd: Ranked puzzles
+  0pdd --> repo: Ranked puzzles
+end
+@enduml

model/README.md

@@ -1,18 +1,18 @@
 Puzzle Ranking (Linear ML Model)
 
-<<<<<<< Updated upstream
-###### Note: This is an opt-in feature
-=======
-The data for puzzles is pre-processed and available in `~/data/proper_pdd_data_regression.csv`. In the data, The first row is the column index, the first column is the repo id the puzzle belongs to and the last column is the output variable (`y`).
->>>>>>> Stashed changes
-
 ### Internals
 
-The ML model is a linear model with PSO optimizer. The optimizer is used to train the model on puzzle data, the weights are stored and used to predict future puzzles.
+The ML model is a linear model with PSO optimizer.
+The optimizer is used to train the model on puzzle data,
+the weights are stored and used to predict future puzzles.
 
-Because of the time required, training is a non-blocking process, and puzzle prioritization uses a naive ranking approach based on puzzle estimate. Subsequent events use the linear model for prioritization.
+Because of the time required, training is a non-blocking process,
+and puzzle prioritization uses a naive ranking approach based on puzzle estimate.
+Subsequent events use the linear model for prioritization.
 
-The linear model is the external API for the model. It has one method `predict(...)` which accepts an array of puzzles in xml. The output of this model is an array of positional index of the input puzzles:
+The linear model is the external API for the model.
+It has one method `predict(...)` which accepts an array of puzzles in xml.
+The output of this model is an array of positional index of the input puzzles:
 
 ```ruby
 # usage
@@ -25,3 +25,8 @@ rank = LinearModel.new(repo_name, storage).predict(puzzles)
 #
 # rank -> array of positional index of ranked puzzles
 ```
+
+### Integration
+
+This diagram shows how this model can be integrated into 0pdd workflow:
+![integration.svg](../doc/integration.svg)

model/linear.rb

@@ -27,6 +27,9 @@
 
 #
 # Linear Model
+# @todo #532:60min Add unit-tests.
+#  We should add unit-tests for this class that checks puzzle ranking.
+#  For now its untested, don't forget to remove this puzzle.
 #
 class LinearModel
   def initialize(repo, storage)
@@ -46,9 +49,23 @@ def initialize(repo, storage)
     end
   end
 
+  # ranks the puzzles using Machine-Learning
+  # @param puzzles XML puzzles
+  # @return array of positional index of the input puzzles
+  # @todo #532:60min Implement a ranked puzzles.
+  #  Let's implement a class that will use `LinearModel` to rank puzzles.
+  #  This class is need in order to do an integration between original 0pdd
+  #  and model modules. Probably it can be a decorator for `Puzzles`
+  #  that ranks XML puzzles, and then submits them into `Puzzles`.
+  #  Don't forget to remove this puzzle.
   def predict(puzzles)
     weights = @storage.load # load weights for repo from s3
-    clf = Predictor.new(layers: [{ name: 'w1', shape: [10, 1] }, { name: 'w2', shape: [1, 1] }])
+    clf = Predictor.new(
+      layers: [
+        { name: 'w1', shape: [10, 1] },
+        { name: 'w2', shape: [1, 1] }
+      ]
+    )
     if weights.nil?
       train(clf) # find weights for repo backlog of puzzles
       ranks = naive_rank(puzzles) # naive rank of puzzles in each repo

objects/puzzles.rb

@@ -25,6 +25,11 @@
 
 #
 # Puzzles in XML/S3
+# @todo #532:60min Implement a decorator for optional model configuration load.
+#  Let's implement a class that decorates `Puzzles` and
+#  based on presence of `model: true` attribute in YAML config, decides
+#  whether the puzzles should be ranked or not.
+#  Don't forget to remove this puzzle.
 #
 class Puzzles
   def initialize(repo, storage)