ankigammon 1.0.0__py3-none-any.whl

ankigammon/parsers/gnubg_parser.py

@@ -0,0 +1,454 @@
+"""
+GNU Backgammon text output parser.
+
+Parses analysis output from gnubg-cli.exe into Decision objects.
+"""
+
+import re
+from typing import List, Optional
+
+from ankigammon.models import Decision, DecisionType, Move, Player, Position
+from ankigammon.utils.xgid import parse_xgid
+
+
+class GNUBGParser:
+    """Parse GNU Backgammon analysis output."""
+
+    @staticmethod
+    def parse_analysis(
+        gnubg_output: str,
+        xgid: str,
+        decision_type: DecisionType
+    ) -> Decision:
+        """
+        Parse gnubg output into Decision object.
+
+        Args:
+            gnubg_output: Raw text output from gnubg-cli.exe
+            xgid: Original XGID for position reconstruction
+            decision_type: CHECKER_PLAY or CUBE_ACTION
+
+        Returns:
+            Decision object with populated candidate_moves
+
+        Raises:
+            ValueError: If parsing fails
+        """
+        # Parse XGID to get position and metadata
+        position, metadata = parse_xgid(xgid)
+
+        # Parse moves based on decision type
+        if decision_type == DecisionType.CHECKER_PLAY:
+            moves = GNUBGParser._parse_checker_play(gnubg_output)
+        else:
+            moves = GNUBGParser._parse_cube_decision(gnubg_output)
+
+        if not moves:
+            raise ValueError(f"No moves found in gnubg output for {decision_type.value}")
+
+        # Extract winning chances from metadata or output
+        winning_chances = GNUBGParser._parse_winning_chances(gnubg_output)
+
+        # Build Decision object
+        decision = Decision(
+            position=position,
+            on_roll=metadata.get('on_roll', Player.O),
+            decision_type=decision_type,
+            candidate_moves=moves,
+            dice=metadata.get('dice'),
+            xgid=xgid,
+            score_x=metadata.get('score_x', 0),
+            score_o=metadata.get('score_o', 0),
+            match_length=metadata.get('match_length', 0),
+            cube_value=metadata.get('cube_value', 1),
+            cube_owner=metadata.get('cube_owner'),
+        )
+
+        # Add winning chances to decision if found
+        if winning_chances:
+            decision.player_win_pct = winning_chances.get('player_win_pct')
+            decision.player_gammon_pct = winning_chances.get('player_gammon_pct')
+            decision.player_backgammon_pct = winning_chances.get('player_backgammon_pct')
+            decision.opponent_win_pct = winning_chances.get('opponent_win_pct')
+            decision.opponent_gammon_pct = winning_chances.get('opponent_gammon_pct')
+            decision.opponent_backgammon_pct = winning_chances.get('opponent_backgammon_pct')
+
+        return decision
+
+    @staticmethod
+    def _parse_checker_play(text: str) -> List[Move]:
+        """
+        Parse checker play analysis from gnubg output.
+
+        Expected format:
+            1. Cubeful 4-ply    21/16 21/15                  Eq.:  -0.411
+               0.266 0.021 0.001 - 0.734 0.048 0.001
+                4-ply cubeful prune [4ply]
+            2. Cubeful 4-ply    9/4 9/3                      Eq.:  -0.437 ( -0.025)
+               0.249 0.004 0.000 - 0.751 0.021 0.000
+                4-ply cubeful prune [4ply]
+
+        Args:
+            text: gnubg output text
+
+        Returns:
+            List of Move objects sorted by rank
+        """
+        moves = []
+        lines = text.split('\n')
+
+        # Pattern for gnubg move lines
+        # Matches: "    1. Cubeful 4-ply    21/16 21/15                  Eq.:  -0.411"
+        #          "    2. Cubeful 4-ply    9/4 9/3                      Eq.:  -0.437 ( -0.025)"
+        move_pattern = re.compile(
+            r'^\s*(\d+)\.\s+(?:Cubeful\s+\d+-ply\s+)?(.*?)\s+Eq\.?:\s*([+-]?\d+\.\d+)(?:\s*\(\s*([+-]?\d+\.\d+)\))?',
+            re.IGNORECASE
+        )
+
+        # Pattern for probability line
+        # Matches: "       0.266 0.021 0.001 - 0.734 0.048 0.001"
+        prob_pattern = re.compile(
+            r'^\s*(\d\.\d+)\s+(\d\.\d+)\s+(\d\.\d+)\s*-\s*(\d\.\d+)\s+(\d\.\d+)\s+(\d\.\d+)'
+        )
+
+        for i, line in enumerate(lines):
+            match = move_pattern.match(line)
+            if match:
+                rank = int(match.group(1))
+                notation = match.group(2).strip()
+                equity = float(match.group(3))
+                error_str = match.group(4)
+
+                # Parse error (if shown)
+                error = float(error_str) if error_str else 0.0
+                abs_error = abs(error)
+
+                # Look for probability line on next line
+                player_win = None
+                player_gammon = None
+                player_backgammon = None
+                opponent_win = None
+                opponent_gammon = None
+                opponent_backgammon = None
+
+                if i + 1 < len(lines):
+                    prob_match = prob_pattern.match(lines[i + 1])
+                    if prob_match:
+                        # Convert from decimal to percentage
+                        player_win = float(prob_match.group(1)) * 100
+                        player_gammon = float(prob_match.group(2)) * 100
+                        player_backgammon = float(prob_match.group(3)) * 100
+                        opponent_win = float(prob_match.group(4)) * 100
+                        opponent_gammon = float(prob_match.group(5)) * 100
+                        opponent_backgammon = float(prob_match.group(6)) * 100
+
+                moves.append(Move(
+                    notation=notation,
+                    equity=equity,
+                    rank=rank,
+                    error=abs_error,
+                    xg_error=error,
+                    xg_notation=notation,
+                    xg_rank=rank,
+                    from_xg_analysis=True,
+                    player_win_pct=player_win,
+                    player_gammon_pct=player_gammon,
+                    player_backgammon_pct=player_backgammon,
+                    opponent_win_pct=opponent_win,
+                    opponent_gammon_pct=opponent_gammon,
+                    opponent_backgammon_pct=opponent_backgammon
+                ))
+
+        # If no moves found, try alternative pattern
+        if not moves:
+            # Try simpler pattern without rank numbers
+            alt_pattern = re.compile(
+                r'^\s*([0-9/\s*bar]+?)\s+Eq:\s*([+-]?\d+\.\d+)',
+                re.MULTILINE
+            )
+            for i, match in enumerate(alt_pattern.finditer(text), 1):
+                notation = match.group(1).strip()
+                equity = float(match.group(2))
+
+                moves.append(Move(
+                    notation=notation,
+                    equity=equity,
+                    rank=i,
+                    error=0.0,
+                    from_xg_analysis=True
+                ))
+
+        # Sort by equity (highest first) and recalculate errors
+        if moves:
+            moves.sort(key=lambda m: m.equity, reverse=True)
+            best_equity = moves[0].equity
+
+            for i, move in enumerate(moves, 1):
+                move.rank = i
+                move.error = abs(best_equity - move.equity)
+
+        return moves
+
+    @staticmethod
+    def _parse_cube_decision(text: str) -> List[Move]:
+        """
+        Parse cube decision analysis from gnubg output.
+
+        Expected format:
+            Cubeful equities:
+            1. No double           +0.172
+            2. Double, take        -0.361  (-0.533)
+            3. Double, pass        +1.000  (+0.828)
+
+            Proper cube action: No double
+
+        Generates all 5 cube options (like XG parser):
+        - No double/Take
+        - Double/Take
+        - Double/Pass
+        - Too good/Take (synthetic)
+        - Too good/Pass (synthetic)
+
+        Args:
+            text: gnubg output text
+
+        Returns:
+            List of Move objects with all 5 cube options
+        """
+        moves = []
+
+        # Look for "Cubeful equities:" section
+        if 'Cubeful equities' not in text and 'cubeful equities' not in text:
+            return moves
+
+        # Parse the 3 equity values from gnubg
+        # Pattern to match cube decision lines:
+        # "1. No double           +0.172"
+        # "2. Double, take        -0.361  (-0.533)"
+        # "3. Double, pass        +1.000  (+0.828)"
+        pattern = re.compile(
+            r'^\s*\d+\.\s*(No (?:re)?double|(?:Re)?[Dd]ouble,?\s*(?:take|pass|drop))\s*([+-]?\d+\.\d+)(?:\s*\(([+-]\d+\.\d+)\))?',
+            re.MULTILINE | re.IGNORECASE
+        )
+
+        # Store parsed equities in order they appear
+        gnubg_moves_data = []  # List of (normalized_notation, equity, gnubg_error)
+        for match in pattern.finditer(text):
+            notation = match.group(1).strip()
+            equity = float(match.group(2))
+            error_str = match.group(3)
+
+            # Parse gnubg's error (in parentheses)
+            gnubg_error = float(error_str) if error_str else 0.0
+
+            # Normalize notation: "Double, take" -> "Double/Take"
+            normalized = notation.replace(', ', '/').replace(',', '/')
+            normalized = GNUBGParser._normalize_cube_notation(normalized)
+
+            gnubg_moves_data.append((normalized, equity, gnubg_error))
+
+        if not gnubg_moves_data:
+            return moves
+
+        # Build equity map for easy lookup
+        equity_map = {data[0]: data[1] for data in gnubg_moves_data}
+
+        # Parse "Proper cube action:" to determine best move
+        best_action_match = re.search(
+            r'Proper cube action:\s*(.+?)(?:\n|$)',
+            text,
+            re.IGNORECASE
+        )
+
+        best_action_text = None
+        if best_action_match:
+            best_action_text = best_action_match.group(1).strip()
+
+        # Determine if using "double" or "redouble" terminology
+        use_redouble = any('redouble' in data[0].lower() for data in gnubg_moves_data)
+        double_term = "Redouble" if use_redouble else "Double"
+
+        # Generate all 5 cube options with appropriate terminology
+        all_options = [
+            f"No {double_term}/Take",
+            f"{double_term}/Take",
+            f"{double_term}/Pass",
+            f"Too good/Take",
+            f"Too good/Pass"
+        ]
+
+        # Assign equities
+        no_double_eq = equity_map.get("No Double", None)
+        double_take_eq = equity_map.get("Double/Take", None)
+        double_pass_eq = equity_map.get("Double/Pass", None)
+
+        option_equities = {}
+        if no_double_eq is not None:
+            option_equities[f"No {double_term}/Take"] = no_double_eq
+        if double_take_eq is not None:
+            option_equities[f"{double_term}/Take"] = double_take_eq
+        if double_pass_eq is not None:
+            option_equities[f"{double_term}/Pass"] = double_pass_eq
+
+        # For "Too good" options, use same equity as Double/Pass
+        if double_pass_eq is not None:
+            option_equities["Too good/Take"] = double_pass_eq
+            option_equities["Too good/Pass"] = double_pass_eq
+
+        # Determine best notation from "Proper cube action:" text
+        best_notation = GNUBGParser._parse_best_cube_action(best_action_text, double_term)
+
+        # Create Move objects for all 5 options
+        for option in all_options:
+            equity = option_equities.get(option, 0.0)
+            is_from_gnubg = not option.startswith("Too good")
+
+            moves.append(Move(
+                notation=option,
+                equity=equity,
+                error=0.0,  # Will calculate below
+                rank=0,  # Will assign below
+                xg_error=None,
+                xg_notation=option if is_from_gnubg else None,
+                xg_rank=None,
+                from_xg_analysis=is_from_gnubg
+            ))
+
+        # Sort by equity (highest first) to determine ranking
+        moves.sort(key=lambda m: m.equity, reverse=True)
+
+        # Assign ranks
+        if best_notation:
+            rank_counter = 1
+            for move in moves:
+                if move.notation == best_notation:
+                    move.rank = 1
+                else:
+                    if rank_counter == 1:
+                        rank_counter = 2
+                    move.rank = rank_counter
+                    rank_counter += 1
+        else:
+            # Best wasn't identified, rank purely by equity
+            for i, move in enumerate(moves, 1):
+                move.rank = i
+
+        # Calculate errors relative to best move
+        if moves:
+            best_move = next((m for m in moves if m.rank == 1), moves[0])
+            for move in moves:
+                move.error = abs(best_move.equity - move.equity)
+
+        return moves
+
+    @staticmethod
+    def _normalize_cube_notation(notation: str) -> str:
+        """
+        Normalize cube notation to standard format.
+
+        Args:
+            notation: Raw notation (e.g., "Double, take", "No redouble")
+
+        Returns:
+            Normalized notation (e.g., "Double/Take", "No Double")
+        """
+        # Standardize case
+        parts = notation.split('/')
+        result_parts = []
+
+        for part in parts:
+            part = part.strip().lower()
+
+            # Normalize terms
+            if 'no' in part and ('double' in part or 'redouble' in part):
+                result_parts.append("No Double")
+            elif 'double' in part or 'redouble' in part:
+                result_parts.append("Double")
+            elif 'take' in part:
+                result_parts.append("Take")
+            elif 'pass' in part or 'drop' in part:
+                result_parts.append("Pass")
+            elif 'too good' in part:
+                result_parts.append("Too good")
+            else:
+                result_parts.append(part.capitalize())
+
+        return '/'.join(result_parts)
+
+    @staticmethod
+    def _parse_best_cube_action(best_text: Optional[str], double_term: str) -> Optional[str]:
+        """
+        Parse "Proper cube action:" text to determine best move notation.
+
+        Args:
+            best_text: Text from "Proper cube action:" line
+            double_term: "Double" or "Redouble"
+
+        Returns:
+            Standardized notation matching all_options format
+        """
+        if not best_text:
+            return None
+
+        text_lower = best_text.lower()
+
+        if 'too good' in text_lower:
+            if 'take' in text_lower:
+                return "Too good/Take"
+            elif 'pass' in text_lower or 'drop' in text_lower:
+                return "Too good/Pass"
+        elif 'no double' in text_lower or 'no redouble' in text_lower:
+            return f"No {double_term}/Take"
+        elif 'double' in text_lower or 'redouble' in text_lower:
+            if 'take' in text_lower:
+                return f"{double_term}/Take"
+            elif 'pass' in text_lower or 'drop' in text_lower:
+                return f"{double_term}/Pass"
+
+        return None
+
+    @staticmethod
+    def _parse_winning_chances(text: str) -> dict:
+        """
+        Extract W/G/B percentages from gnubg output.
+
+        Looks for patterns like:
+            Cubeless equity: +0.172
+            Win: 52.3%  G: 14.2%  B: 0.8%
+
+        or:
+            0.523 0.142 0.008 - 0.477 0.124 0.006
+
+        Args:
+            text: gnubg output text
+
+        Returns:
+            Dictionary with winning chance percentages (or empty dict)
+        """
+        chances = {}
+
+        # Try pattern 1: "Win: 52.3%  G: 14.2%  B: 0.8%"
+        win_pattern = re.search(
+            r'Win:\s*(\d+\.?\d*)%.*?G:\s*(\d+\.?\d*)%.*?B:\s*(\d+\.?\d*)%',
+            text,
+            re.IGNORECASE
+        )
+        if win_pattern:
+            chances['player_win_pct'] = float(win_pattern.group(1))
+            chances['player_gammon_pct'] = float(win_pattern.group(2))
+            chances['player_backgammon_pct'] = float(win_pattern.group(3))
+
+        # Try pattern 2: Decimal probabilities "0.523 0.142 0.008 - 0.477 0.124 0.006"
+        prob_pattern = re.search(
+            r'(\d\.\d+)\s+(\d\.\d+)\s+(\d\.\d+)\s*-\s*(\d\.\d+)\s+(\d\.\d+)\s+(\d\.\d+)',
+            text
+        )
+        if prob_pattern:
+            chances['player_win_pct'] = float(prob_pattern.group(1)) * 100
+            chances['player_gammon_pct'] = float(prob_pattern.group(2)) * 100
+            chances['player_backgammon_pct'] = float(prob_pattern.group(3)) * 100
+            chances['opponent_win_pct'] = float(prob_pattern.group(4)) * 100
+            chances['opponent_gammon_pct'] = float(prob_pattern.group(5)) * 100
+            chances['opponent_backgammon_pct'] = float(prob_pattern.group(6)) * 100
+
+        return chances