dfm-python 0.1.0__py3-none-any.whl

dfm_python/__init__.py

@@ -0,0 +1,28 @@
+"""Core nowcasting modules for DFM estimation and forecasting."""
+
+from .config import DFMConfig, AppConfig
+from .data_loader import (
+    load_config, load_config_from_yaml, load_config_from_csv, load_data, transform_data
+)
+from .dfm import DFMResult, dfm
+from .kalman import run_kf, skf, fis, miss_data
+from .news import update_nowcast, news_dfm, para_const
+
+# Backward compatibility aliases
+ModelConfig = DFMConfig  # Deprecated: ModelConfig merged into DFMConfig
+ModelSpec = DFMConfig  # Deprecated: use DFMConfig
+load_spec = load_config  # Deprecated: use load_config
+
+__all__ = [
+    'DFMConfig', 'AppConfig',
+    # Backward compatibility
+    'ModelConfig',  # Alias for DFMConfig
+    'load_config', 'load_config_from_yaml', 'load_config_from_csv',
+    'load_data', 'transform_data',
+    'DFMResult', 'dfm',
+    'run_kf', 'skf', 'fis', 'miss_data',
+    'update_nowcast', 'news_dfm', 'para_const',
+    # Deprecated aliases
+    'ModelSpec', 'load_spec',
+]
+

dfm_python/config.py

@@ -0,0 +1,435 @@
+"""Configuration models for DFM nowcasting using OmegaConf and Hydra."""
+
+import numpy as np
+from typing import List, Optional, Dict, Any
+from pathlib import Path
+import warnings
+from dataclasses import dataclass, field
+
+try:
+    from hydra.core.config_store import ConfigStore
+    from omegaconf import DictConfig, OmegaConf
+    HYDRA_AVAILABLE = True
+except ImportError:
+    HYDRA_AVAILABLE = False
+    ConfigStore = None
+    DictConfig = None
+    OmegaConf = None
+
+# Valid frequency codes
+_VALID_FREQUENCIES = {'d', 'w', 'm', 'q', 'sa', 'a'}
+
+# Valid transformation codes
+_VALID_TRANSFORMATIONS = {
+    'lin', 'chg', 'ch1', 'pch', 'pc1', 'pca', 
+    'cch', 'cca', 'log'
+}
+
+# Transformation to readable units mapping
+_TRANSFORM_UNITS_MAP = {
+    'lin': 'Levels (No Transformation)',
+    'chg': 'Change (Difference)',
+    'ch1': 'Year over Year Change (Difference)',
+    'pch': 'Percent Change',
+    'pc1': 'Year over Year Percent Change',
+    'pca': 'Percent Change (Annual Rate)',
+    'cch': 'Continuously Compounded Rate of Change',
+    'cca': 'Continuously Compounded Annual Rate of Change',
+    'log': 'Natural Log'
+}
+
+
+def validate_frequency(frequency: str) -> str:
+    """Validate frequency code."""
+    if frequency not in _VALID_FREQUENCIES:
+        raise ValueError(f"Invalid frequency: {frequency}. Must be one of {_VALID_FREQUENCIES}")
+    return frequency
+
+
+def validate_transformation(transformation: str) -> str:
+    """Validate transformation code."""
+    if transformation not in _VALID_TRANSFORMATIONS:
+        warnings.warn(f"Unknown transformation code: {transformation}. Will use untransformed data.")
+    return transformation
+
+
+@dataclass
+class SeriesConfig:
+    """Configuration for a single time series.
+    
+    This is a generic DFM configuration - no API or database-specific fields.
+    For API/database integration, implement adapters in your application layer.
+    """
+    # Required fields (no defaults)
+    frequency: str
+    transformation: str
+    blocks: List[int]
+    # Optional fields (with defaults - must come after required fields)
+    series_id: Optional[str] = None  # Auto-generated if None: "series_0", "series_1", etc.
+    series_name: Optional[str] = None  # Optional metadata for display
+    units: str = ""  # Optional metadata
+    category: str = ""  # Optional metadata
+    
+    def __post_init__(self):
+        """Validate fields after initialization."""
+        self.frequency = validate_frequency(self.frequency)
+        self.transformation = validate_transformation(self.transformation)
+        # Auto-generate series_id if not provided
+        if self.series_id is None:
+            # Will be set when SeriesConfig is created in a list
+            pass
+        # Auto-generate series_name if not provided
+        if self.series_name is None and self.series_id:
+            self.series_name = self.series_id
+
+
+@dataclass
+class DFMConfig:
+    """Unified DFM configuration - model structure + estimation parameters.
+    
+    This is the single configuration class for the DFM module, combining:
+    - Model structure (what series, blocks, factors)
+    - Estimation parameters (how to run EM algorithm)
+    """
+    # ========================================================================
+    # Model Structure (WHAT - defines the model)
+    # ========================================================================
+    series: List[SeriesConfig]  # Series specifications
+    block_names: List[str]  # Block names (e.g., ["Global", "Consumption", "Investment"])
+    factors_per_block: Optional[List[int]] = None  # Number of factors per block. If None, defaults to 1 per block
+    
+    # ========================================================================
+    # Estimation Parameters (HOW - controls the algorithm)
+    # ========================================================================
+    ar_lag: int = 1  # Number of lags in AR transition equation (lookback window)
+    threshold: float = 1e-5  # EM convergence threshold
+    max_iter: int = 5000  # Maximum EM iterations
+    nan_method: int = 2  # Missing data handling method (1-5)
+    nan_k: int = 3  # Spline parameter for NaN interpolation
+    
+    # ========================================================================
+    # Internal cache (not user-configurable)
+    # ========================================================================
+    _cached_blocks: Optional[np.ndarray] = field(default=None, init=False, repr=False)
+    
+    def __post_init__(self):
+        """Validate blocks structure and consistency."""
+        if not self.series:
+            raise ValueError("At least one series must be specified")
+        
+        # Auto-generate series_id if not provided
+        for i, s in enumerate(self.series):
+            if s.series_id is None:
+                s.series_id = f"series_{i}"
+            if s.series_name is None:
+                s.series_name = s.series_id
+        
+        # Extract blocks matrix
+        n_series = len(self.series)
+        n_blocks = len(self.block_names)
+        
+        # Check all series have same number of blocks
+        for i, s in enumerate(self.series):
+            if len(s.blocks) != n_blocks:
+                raise ValueError(
+                    f"Series {i} ({s.series_id}) has {len(s.blocks)} blocks, "
+                    f"but expected {n_blocks} (from block_names)"
+                )
+        
+        # Check first column (global block) is all 1s
+        for i, s in enumerate(self.series):
+            if s.blocks[0] != 1:
+                raise ValueError(
+                    f"Series {i} ({s.series_id}) must load on global block "
+                    f"(first block must be 1)"
+                )
+        
+        # Validate factors_per_block if provided
+        if self.factors_per_block is not None:
+            if len(self.factors_per_block) != n_blocks:
+                raise ValueError(
+                    f"factors_per_block length ({len(self.factors_per_block)}) must match "
+                    f"number of blocks ({n_blocks})"
+                )
+            if any(f < 1 for f in self.factors_per_block):
+                raise ValueError("factors_per_block must contain positive integers (>= 1)")
+    
+    # Convenience properties for backward compatibility
+    @property
+    def SeriesID(self) -> List[str]:
+        """Backward compatibility: SeriesID property."""
+        return [s.series_id if s.series_id is not None else f"series_{i}" 
+                for i, s in enumerate(self.series)]
+    
+    @property
+    def SeriesName(self) -> List[str]:
+        """Backward compatibility: SeriesName property."""
+        return [s.series_name if s.series_name is not None else (s.series_id or f"series_{i}")
+                for i, s in enumerate(self.series)]
+    
+    @property
+    def Frequency(self) -> List[str]:
+        """Backward compatibility: Frequency property."""
+        return [s.frequency for s in self.series]
+    
+    @property
+    def Units(self) -> List[str]:
+        """Backward compatibility: Units property."""
+        return [s.units for s in self.series]
+    
+    @property
+    def Transformation(self) -> List[str]:
+        """Backward compatibility: Transformation property."""
+        return [s.transformation for s in self.series]
+    
+    @property
+    def Category(self) -> List[str]:
+        """Backward compatibility: Category property."""
+        return [s.category for s in self.series]
+    
+    @property
+    def Blocks(self) -> np.ndarray:
+        """Backward compatibility: Blocks property as numpy array (cached)."""
+        if self._cached_blocks is None:
+            blocks_list = [s.blocks for s in self.series]
+            self._cached_blocks = np.array(blocks_list, dtype=int)
+        return self._cached_blocks
+    
+    @property
+    def BlockNames(self) -> List[str]:
+        """Backward compatibility: BlockNames property (alias for block_names)."""
+        return self.block_names
+    
+    @property
+    def UnitsTransformed(self) -> List[str]:
+        """Backward compatibility: UnitsTransformed property."""
+        return [_TRANSFORM_UNITS_MAP.get(t, t) for t in self.Transformation]
+    
+    @classmethod
+    def _from_legacy_dict(cls, data: Dict[str, Any]) -> 'DFMConfig':
+        """Convert legacy format (separate lists) to new format (series list)."""
+        series_list = []
+        n = len(data.get('SeriesID', data.get('series_id', [])))
+        
+        # Handle Blocks - can be numpy array or list of lists
+        blocks_data = data.get('Blocks', data.get('blocks', []))
+        if isinstance(blocks_data, np.ndarray):
+            blocks_data = blocks_data.tolist()
+        elif not isinstance(blocks_data, list):
+            blocks_data = []
+        
+        # Helper to get list value with index fallback
+        def get_list_value(key: str, index: int, default=None):
+            """Get value from list, handling both camelCase and snake_case keys."""
+            val = data.get(key, data.get(key.lower(), default))
+            if isinstance(val, list) and index < len(val):
+                return val[index]
+            return default
+        
+        for i in range(n):
+            # Extract blocks for this series
+            if blocks_data and i < len(blocks_data):
+                if isinstance(blocks_data[i], (list, np.ndarray)):
+                    series_blocks = list(blocks_data[i]) if isinstance(blocks_data[i], np.ndarray) else blocks_data[i]
+                else:
+                    series_blocks = [blocks_data[i]]
+            else:
+                series_blocks = []
+            
+            freq_val = get_list_value('Frequency', i, 'm')
+            trans_val = get_list_value('Transformation', i, 'lin')
+            units_val = get_list_value('Units', i, '')
+            cat_val = get_list_value('Category', i, '')
+            series_list.append(SeriesConfig(
+                frequency=str(freq_val) if freq_val is not None else 'm',
+                transformation=str(trans_val) if trans_val is not None else 'lin',
+                blocks=series_blocks,
+                series_id=get_list_value('SeriesID', i, None),
+                series_name=get_list_value('SeriesName', i, None),
+                units=str(units_val) if units_val is not None else '',
+                category=str(cat_val) if cat_val is not None else ''
+            ))
+        
+        # Extract estimation parameters if provided
+        return cls(
+            series=series_list,
+            block_names=data.get('BlockNames', data.get('block_names', [])),
+            factors_per_block=data.get('factors_per_block', None),
+            # Estimation parameters
+            ar_lag=data.get('ar_lag', 1),
+            threshold=data.get('threshold', 1e-5),
+            max_iter=data.get('max_iter', 5000),
+            nan_method=data.get('nan_method', 2),
+            nan_k=data.get('nan_k', 3)
+        )
+    
+    @classmethod
+    def from_dict(cls, data: Dict[str, Any]) -> 'DFMConfig':
+        """Create DFMConfig from dictionary.
+        
+        Handles multiple formats:
+        1. Legacy format: {'SeriesID': [...], 'Frequency': [...], 'Blocks': [[...]], ...}
+        2. New format (list): {'series': [{'series_id': ..., ...}], 'block_names': [...]}
+        3. New format (Hydra): {'series': {'series_id': {...}}, 'blocks': {'block_name': {'factors': N}}}
+        
+        Also accepts estimation parameters: ar_lag, threshold, max_iter, nan_method, nan_k
+        """
+        # Detect legacy format (has SeriesID or series_id as lists)
+        if 'SeriesID' in data or ('series_id' in data and isinstance(data.get('series_id'), list)):
+            return cls._from_legacy_dict(data)
+        
+        # New Hydra format: series is a dict (series_id -> config), blocks is a dict (block_name -> {factors: N})
+        if 'series' in data and isinstance(data['series'], dict):
+            # First, parse blocks dict to get block_names (needed for conversion)
+            blocks_dict = data.get('blocks', {})
+            if isinstance(blocks_dict, dict):
+                block_names = list(blocks_dict.keys())
+                factors_per_block = [
+                    blocks_dict[bn].get('factors', 1) if isinstance(blocks_dict[bn], dict) else blocks_dict[bn]
+                    for bn in block_names
+                ]
+            else:
+                block_names = data.get('block_names', [])
+                factors_per_block = data.get('factors_per_block', None)
+            
+            # Parse series dict: {series_id: {frequency: ..., blocks: [block_names], ...}}
+            series_list = []
+            for series_id, series_cfg in data['series'].items():
+                if isinstance(series_cfg, dict):
+                    # Convert block names to binary array (0/1 for each block)
+                    series_blocks_names = series_cfg.get('blocks', [])
+                    if isinstance(series_blocks_names, list) and series_blocks_names:
+                        # Create binary array: 1 if series loads on block, 0 otherwise
+                        # First block (Global) must always be 1
+                        series_blocks = [0] * len(block_names)
+                        series_blocks[0] = 1  # Global block is always 1
+                        for block_name in series_blocks_names:
+                            if block_name in block_names:
+                                block_idx = block_names.index(block_name)
+                                series_blocks[block_idx] = 1
+                    else:
+                        # Default: only global block
+                        series_blocks = [1] + [0] * (len(block_names) - 1)
+                    
+                    series_list.append(SeriesConfig(
+                        series_id=series_id,
+                        series_name=series_cfg.get('series_name', series_id),
+                        frequency=series_cfg.get('frequency', 'm'),
+                        transformation=series_cfg.get('transformation', 'lin'),
+                        blocks=series_blocks,
+                        units=series_cfg.get('units', ''),
+                        category=series_cfg.get('category', '')
+                    ))
+            
+            # Extract estimation parameters if provided
+            return cls(
+                series=series_list,
+                block_names=block_names,
+                factors_per_block=factors_per_block if factors_per_block else None,
+                # Estimation parameters
+                ar_lag=data.get('ar_lag', 1),
+                threshold=data.get('threshold', 1e-5),
+                max_iter=data.get('max_iter', 5000),
+                nan_method=data.get('nan_method', 2),
+                nan_k=data.get('nan_k', 3)
+            )
+        
+        # New format with series list
+        if 'series' in data and isinstance(data['series'], list):
+            series_list = [
+                SeriesConfig(**s) if isinstance(s, dict) else s 
+                for s in data['series']
+            ]
+            # Extract estimation parameters if provided
+            return cls(
+                series=series_list,
+                block_names=data.get('block_names', []),
+                factors_per_block=data.get('factors_per_block', None),
+                # Estimation parameters
+                ar_lag=data.get('ar_lag', 1),
+                threshold=data.get('threshold', 1e-5),
+                max_iter=data.get('max_iter', 5000),
+                nan_method=data.get('nan_method', 2),
+                nan_k=data.get('nan_k', 3)
+            )
+        
+        # Direct instantiation (shouldn't happen often, but handle it)
+        return cls(**data)
+
+
+# DataConfig removed - data loading is application-specific, not part of generic DFM module
+# Use adapters for data loading configuration
+
+@dataclass
+class AppConfig:
+    """Root application configuration (for backward compatibility only).
+    
+    Note: DataConfig has been removed from generic DFM module.
+    Data loading configuration should be handled by adapters/application layer.
+    """
+    model: DFMConfig  # Renamed from ModelConfig to DFMConfig
+    # data: DataConfig  # Removed - not part of generic DFM module
+    dfm: Optional[DFMConfig] = None  # Deprecated - use model config directly
+
+
+# Register with Hydra ConfigStore following the Structured Config schema pattern
+# This enables validation of YAML config files while keeping our full dataclass
+# with @property methods for runtime use.
+# 
+# Pattern: Schema validation (from Hydra docs)
+# - YAML files extend schemas via defaults list
+# - Schemas provide type checking and validation
+# - Runtime uses full ModelConfig with @property methods
+if HYDRA_AVAILABLE and ConfigStore is not None:
+    try:
+        cs = ConfigStore.instance()
+        if cs is None:
+            raise RuntimeError("ConfigStore.instance() returned None")
+        # Create schema versions without @property methods for Hydra validation
+        # These match our dataclass structure exactly for schema validation.
+        # We'll still use the full ModelConfig/DataConfig/DFMConfig classes with
+        # @property methods at runtime via from_dict() conversion.
+        from dataclasses import dataclass as schema_dataclass
+        
+        @schema_dataclass
+        class SeriesConfigSchema:
+            """Schema for SeriesConfig validation in Hydra."""
+            series_id: str
+            series_name: str
+            frequency: str
+            units: str
+            transformation: str
+            category: str
+            blocks: List[int]
+        
+        @schema_dataclass
+        class DFMConfigSchema:
+            """Schema for unified DFMConfig validation in Hydra."""
+            # Model structure
+            series: List[SeriesConfigSchema]
+            block_names: List[str]
+            factors_per_block: Optional[List[int]] = None
+            # Estimation parameters
+            ar_lag: int = 1
+            threshold: float = 1e-5
+            max_iter: int = 5000
+            nan_method: int = 2
+            nan_k: int = 3
+        
+        # Register schemas in config groups (following Hydra docs pattern)
+        # These can be referenced in YAML defaults lists for validation
+        # Format: defaults: [base_dfm_config, _self_]
+        cs.store(group="dfm", name="base_dfm_config", node=DFMConfigSchema)
+        # Also register as "model" for backward compatibility
+        cs.store(group="model", name="base_model_config", node=DFMConfigSchema)
+        
+        # Also register standalone for direct use
+        cs.store(name="dfm_config_schema", node=DFMConfigSchema)
+        # Backward compatibility alias
+        cs.store(name="model_config_schema", node=DFMConfigSchema)
+        
+    except Exception as e:
+        # If registration fails, continue without schema validation
+        # Configs will still work via from_dict() in scripts
+        warnings.warn(f"Could not register Hydra structured config schemas: {e}. "
+                     f"Configs will still work via from_dict() but without schema validation.")