tans-parser 0.1.4 → 0.1.5

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 907b90ef203876bd99cc0dbca8eb6b184ed6472b580c6d2918e31469d0f7cc12
-  data.tar.gz: b828065265563752bb2acf5ef380c2ba805f4b4e785fc602337c2b5048267824
+  metadata.gz: 0ea2a71fb0db321851e51510e4250d9c2afcd2830495771ebb3d0adf36e707b0
+  data.tar.gz: 7198b01f46232a42fc80b8d09668b07dc20834c47d16e0a5502d1bb18b7a5d8f
 SHA512:
-  metadata.gz: ae5ea0f42c3663d0edfc35e86bb11064396099f73beebcdcf2fe5d43cfc2b65d810ab2dc055cf5d186a3571aa11703abc006962407a30b979c1a7d4d6e46f3f7
-  data.tar.gz: b3379404c4cee09f3e49ceeaaa0e3ec90322a665fbb671a7d5e19506759b72e338fce4e782cc2017779d7812eadfddc098df82ba6bd0cb4ecec50a283d4ce044
+  metadata.gz: 165b2cd6b7cc5c0edb4bf4dd510fca304bd48fdcd0bbc7757d9edaf9ef4ef9fd76e3bfc39708301eb16ad263df2a3ec7f2b450ccca9a9b9ed9c24aca64f68f0b
+  data.tar.gz: 8eaede4178263a217f715d9ee79e263aff0169801bb7f9eecc6b82a81608e32f8c442788e17afff67e3a6bd68165de6612a5ca0a80d661b3711fc092b4871561

data/CHANGELOG.md

@@ -1,5 +1,34 @@
 # CHANGELOG
 
+## 0.1.5
+
+- **Confidence scoring** — each detected element now carries a `confidence` value (0.0–1.0):
+  - `Element#confident?` — returns `true` when confidence ≥ 0.5 (or nil, for backward compatibility)
+  - Scoring heuristics per role:
+    - **Buttons**: 0.9 (square), 0.85 (round), 0.75 (angle); −0.2 penalty for single-character text
+    - **Checkboxes**: 0.9 (checked), 0.85 (unchecked)
+    - **Dialogs**: 0.9 base; +0.05 bonus for titled borders (text in top border)
+    - **Statusbars**: 0.9 (inverse colors), 0.85 (separator-preceded footer), 0.5 (≥30-char fallback)
+    - **Progress bars**: 0.9 (incomplete), 0.95 (100% complete)
+    - **Inputs**: 0.9
+    - **Labels**: 0.8 (single-word), 0.85 (multi-word)
+    - **Menus**: 0.9 (3+ items), 0.85 (2 items), 0.8 (dropdown `> Item`)
+    - **Tabs**: 0.85 (3+ tabs), 0.7 (2 tabs); +0.05 for focused tab
+    - **Annotations**: 1.0 (manually annotated); can be overridden via `confidence:` keyword
+  - `confidence` included in `Element#to_h` when set; excluded when nil (backward compatible)
+- **Reduced false positives** — tighter heuristics to avoid misdetection:
+  - **Buttons**: skip numeric-only brackets (e.g. `[12]`, `[3]`)
+  - **Labels**: skip URL schemes (`https://example.com`) and time patterns (`Meeting at 3:00`)
+  - **Progress bars**: minimum width of 6 characters (`[##]` is no longer detected)
+- **Negative tests** — 15 new tests covering edge cases (numeric brackets, URLs, time patterns, short progress bars, tabs across rows, incomplete boxes, prompt-like menus, etc.)
+- **Confidence tests** — 15 new tests verifying scoring for every element type and edge case (titled dialogs, focused tabs, complete progress bars, annotation override, etc.)
+- **Benchmarks** — `benchmark-ips` suite for parser, diff, and selector:
+  - `benchmarks/parser_benchmark.rb` — plain, ANSI, cursor, complex, and dialog-like workloads
+  - `benchmarks/diff_benchmark.rb` — full, chars_only, and ignore_rows modes
+  - `benchmarks/selector_benchmark.rb` — full scan with mixed UI elements
+  - `benchmark-ips` added as development dependency
+- 30 new tests, 374 total, 100% line and branch coverage maintained
+
 ## 0.1.4
 
 - **Unicode width support** — correct display width for CJK, emoji, and combining characters:

data/README.md

@@ -152,7 +152,7 @@ selector.statusbars  # => includes annotated statusbar
 
 # Annotations accept extra attributes
 state.annotate_role(:button, row: 0, col: 0, width: 6, height: 1,
-                    text: "Submit", fg: "green", disabled: false)
+                    text: "Submit", fg: "green", disabled: false, confidence: 0.8)
 ```
 
 ### State comparison (diff)
@@ -197,13 +197,15 @@ el.height    # => 1
 el.checked   # => true/false/nil
 el.focused   # => true/false/nil
 el.disabled  # => true/false/nil
+el.confidence # => 0.9 (Float 0.0-1.0) or nil when not set
 el.fg        # => "default"
 el.bg        # => "default"
-el.to_h      # => {role: :button, text: "OK", row: 1, col: 2, ...}
+el.to_h      # => {role: :button, text: "OK", row: 1, col: 2, confidence: 0.9, ...}
 
 # Predicates
 el.checked?   # => false (always boolean)
 el.disabled?  # => false (always boolean)
+el.confident? # => true when confidence >= 0.5 (or nil)
 
 # Geometry
 el.bounds     # => {row: 1, col: 2, width: 4, height: 1}
@@ -214,6 +216,49 @@ el.type("hello")    # => {action: :type, target: el, row: 1, col: 4, text: "hell
 el.press_key(:tab)  # => {action: :press_key, target: el, key: :tab}
 ```
 
+### Confidence scoring
+
+Each detected element carries a `confidence` value (0.0–1.0) reflecting how sure the heuristics are:
+
+```ruby
+btn = selector.button
+btn.confidence  # => 0.9 (square-bracket buttons are high confidence)
+btn.confident?  # => true
+
+# Low-confidence detections can be filtered out
+reliable = selector.buttons.select(&:confident?)  # confidence >= 0.5
+```
+
+Confidence values per role and context:
+
+| Role | Scenario | Confidence |
+|------|----------|------------|
+| `:button` | `[ OK ]` square brackets | 0.9 |
+| `:button` | `(Cancel)` round brackets | 0.85 |
+| `:button` | `<Submit>` angle brackets | 0.75 |
+| `:button` | Single-character text | −0.2 penalty |
+| `:checkbox` | `[x]` checked | 0.9 |
+| `:checkbox` | `[ ]` unchecked | 0.85 |
+| `:input` | `[________]` underscore brackets | 0.9 |
+| `:label` | `Project Name:` (multi-word) | 0.85 |
+| `:label` | `Username:` (single-word) | 0.8 |
+| `:menu` | 3+ items on menu bar | 0.9 |
+| `:menu` | 2 items on menu bar | 0.85 |
+| `:menu` | `> Item` dropdown | 0.8 |
+| `:tab` | 3+ tabs | 0.85 |
+| `:tab` | 2 tabs | 0.7 |
+| `:tab` | Focused tab (underline/bg) | +0.05 bonus |
+| `:dialog` | Complete box with all 4 corners | 0.9 |
+| `:dialog` | Titled border (text in top border) | 0.95 |
+| `:statusbar` | Inverse colors + ≥3 colored cells | 0.9 |
+| `:statusbar` | Separator-preceded footer | 0.85 |
+| `:statusbar` | Fallback (≥30 chars, no bg info) | 0.5 |
+| `:progress` | `[#####     ]` incomplete | 0.9 |
+| `:progress` | `[##########]` 100% complete | 0.95 |
+| Annotation | Manually annotated via `annotate_role` | 1.0 |
+
+`confidence` is excluded from `to_h` when nil (backward compatible).
+
 ### Recognized element patterns
 
 | Role | Pattern | Example |
@@ -241,8 +286,9 @@ Each cell is a Hash with these keys:
 | `italic` | Boolean | Italic style |
 | `underline` | Boolean | Underline style |
 | `blink` | Boolean | Blink style |
+| `width` | Integer | Display width (1 for normal, 2 for CJK/emoji, 0 for continuation) |
 
-Default cell: `{char: " ", fg: "default", bg: "default", bold: false, italic: false, underline: false, blink: false}`
+Default cell: `{char: " ", fg: "default", bg: "default", bold: false, italic: false, underline: false, blink: false, width: 1}`
 
 ## Supported ANSI sequences
 
@@ -255,7 +301,8 @@ Default cell: `{char: " ", fg: "default", bg: "default", bold: false, italic: fa
 - **Cursor style** — DECSCUSR
 - **Mouse tracking** — DEC private modes 1000, 1002, 1003, 1006
 - **ISO 2022** — G0/G1 charset switching, DEC Special Graphics
-- **UTF-8** — Multi-byte characters including emoji
+- **UTF-8** — Multi-byte characters including CJK, emoji (correct display width via `unicode-display_width`)
+- **Combining characters** — Zero-width combining marks appended to previous cell
 
 ## License
 

data/lib/tans_parser/element.rb

@@ -11,6 +11,7 @@ module TansParser
     :focused,
     :fg, :bg,
     :disabled,
+    :confidence,
     keyword_init: true,
   ) do
     def checked?
@@ -37,6 +38,10 @@ module TansParser
       { action: :press_key, target: self, key: key }
     end
 
+    def confident?
+      confidence.nil? || confidence >= 0.5
+    end
+
     def to_h
       {
         role: role,
@@ -47,6 +52,7 @@ module TansParser
         focused: focused,
         fg: fg, bg: bg,
         disabled: disabled,
+        confidence: confidence,
       }.compact
     end
   end

data/lib/tans_parser/selector.rb

@@ -147,7 +147,7 @@ module TansParser
 
     # Detects annotations: manually annotated roles from State#annotate_role
     def detect_annotations
-      @state.annotations.map { |a| Element.new(a) }
+      @state.annotations.map { |a| Element.new(**a, confidence: a[:confidence] || 1.0) }
     end
 
     # Detects buttons: [ OK ], (Cancel), <Submit>
@@ -163,8 +163,18 @@ module TansParser
           next if text.empty?
           next if text.match?(/^_+$/)
           next if text.match?(/^[ xX*]$/) # skip checkbox markers
+          next if text.match?(/^\d+$/)    # skip numeric-only brackets (e.g. [12])
 
           col = ::Regexp.last_match.begin(0)
+          confidence = if ::Regexp.last_match[1]
+                         0.9
+                       elsif ::Regexp.last_match[2]
+                         0.85
+                       else
+                         0.75
+                       end
+          confidence -= 0.2 if text.length == 1 # penalize single-character buttons
+
           buttons << Element.new(
             role: :button,
             text: text,
@@ -172,6 +182,7 @@ module TansParser
             width: ::Regexp.last_match[0].length, height: 1,
             fg: row[col][:fg],
             bg: row[col][:bg],
+            confidence: confidence,
           )
         end
       end
@@ -190,12 +201,14 @@ module TansParser
         checked = match[2] != " "
         label_text = match[3].strip
         col = match.begin(3)
+        confidence = checked ? 0.9 : 0.85
         checkboxes << Element.new(
           role: :checkbox,
           text: label_text,
           row: r, col: col,
           width: label_text.length, height: 1,
           checked: checked,
+          confidence: confidence,
         )
       end
       checkboxes
@@ -218,11 +231,16 @@ module TansParser
           if bottom_r
             height = bottom_r - r + 1
             text = extract_dialog_text(r + 1, tl_idx + 1, width - 2, height - 2)
+            confidence = 0.9
+            # Bonus for titled borders (text in top border)
+            top_border = line[tl_idx..(tl_idx + width - 1)]
+            confidence = (confidence + 0.05).round(2) if top_border.match?(/[A-Za-z]/)
             dialogs << Element.new(
               role: :dialog,
               text: text,
               row: r, col: tl_idx,
               width: width, height: height,
+              confidence: confidence,
             )
           end
           tl_idx += 1
@@ -291,6 +309,7 @@ module TansParser
           row: row_idx, col: 0,
           width: row.length, height: 1,
           bg: non_default.first[:bg],
+          confidence: 0.9,
         )
         return bars
       end
@@ -303,6 +322,7 @@ module TansParser
           role: :statusbar, text: text,
           row: grid.length - 1, col: 0,
           width: last_row.length, height: 1,
+          confidence: 0.5,
         )
         return bars
       end
@@ -326,6 +346,7 @@ module TansParser
           role: :statusbar, text: text,
           row: r, col: 0,
           width: row.length, height: 1,
+          confidence: 0.85,
         )
         return bars
       end
@@ -341,16 +362,19 @@ module TansParser
         line = row.map { |c| c[:char] }.join
         match = line.match(/\[([#>=-]+)\s*\]/)
         next unless match
+        next if match[0].length < 6 # skip too-short brackets (e.g. [##])
 
         filled = match[1]
         total = match[0].length - 2
         percent = (filled.length.to_f / total * 100).round
+        confidence = percent == 100 ? 0.95 : 0.9
         bars << Element.new(
           role: :progress,
           text: "#{percent}%",
           row: r, col: ::Regexp.last_match.begin(0),
           width: match[0].length, height: 1,
           checked: percent == 100,
+          confidence: confidence,
         )
       end
       bars
@@ -369,6 +393,7 @@ module TansParser
             text: "",
             row: r, col: col,
             width: m[0].length, height: 1,
+            confidence: 0.9,
           )
         end
       end
@@ -385,13 +410,17 @@ module TansParser
 
         label_text = match[1].strip.sub(/:$/, "").strip
         next if label_text.empty? || label_text.length < 2
+        next if match[1].match?(/\d:/)            # skip patterns ending with digit before colon (e.g. "Meeting at 3:")
+        next if line[match.end(1), 2] == "//"     # skip URL schemes (e.g. "https://example.com")
 
         col = match.begin(1)
+        confidence = label_text.include?(" ") ? 0.85 : 0.8 # multi-word labels are stronger signals
         labels << Element.new(
           role: :label,
           text: label_text,
           row: r, col: col,
           width: match[1].length, height: 1,
+          confidence: confidence,
         )
       end
       labels
@@ -411,11 +440,13 @@ module TansParser
           items = stripped.split(/\s{2,}/)
           if items.length >= 2 && items.all? { |i| i.match?(/^[A-Za-z]/) }
             col = line.index(stripped)
+            confidence = items.length >= 3 ? 0.9 : 0.85
             menus << Element.new(
               role: :menu,
               text: items.join(" | "),
               row: r, col: col || 0,
               width: line.length, height: 1,
+              confidence: confidence,
             )
           end
         end
@@ -428,6 +459,7 @@ module TansParser
             text: m[0].sub(/^>\s*/, "").strip,
             row: r, col: m.begin(0),
             width: m[0].length, height: 1,
+            confidence: 0.8,
           )
         end
       end
@@ -453,12 +485,15 @@ module TansParser
 
           cell = row[m.begin(0)]
           focused = cell[:underline] || cell[:bg] != "default"
+          base_confidence = matches.length >= 3 ? 0.85 : 0.7
+          confidence = focused ? [base_confidence + 0.05, 0.9].min.round(2) : base_confidence
           tabs << Element.new(
             role: :tab,
             text: tab_text,
             row: r, col: m.begin(0),
             width: m[0].length, height: 1,
             focused: focused,
+            confidence: confidence,
           )
         end
       end

data/lib/tans_parser/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module TansParser
-  VERSION = "0.1.4"
+  VERSION = "0.1.5"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: tans-parser
 version: !ruby/object:Gem::Version
-  version: 0.1.4
+  version: 0.1.5
 platform: ruby
 authors:
 - Haluk Durmus
@@ -121,6 +121,20 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: '0.22'
+- !ruby/object:Gem::Dependency
+  name: benchmark-ips
+  requirement: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '2.13'
+  type: :development
+  prerelease: false
+  version_requirements: !ruby/object:Gem::Requirement
+    requirements:
+    - - "~>"
+      - !ruby/object:Gem::Version
+        version: '2.13'
 - !ruby/object:Gem::Dependency
   name: unicode-display_width
   requirement: !ruby/object:Gem::Requirement