Source code for backtrader.indicator

#!/usr/bin/env python
"""Backtrader Indicator Module.

This module provides the base Indicator class and related infrastructure
for creating and managing technical analysis indicators. It replaces the
metaclass-based approach with explicit inheritance and registration.

The Indicator class serves as the foundation for all technical indicators
in backtrader, managing line data, minimum periods, and calculation logic.
"""

from .lineiterator import IndicatorBase, LineIterator
from .lineseries import Lines
from .metabase import AutoInfoClass, OwnerContext
from .utils.py3 import range


[docs] class IndicatorRegistry: """Registry to manage indicator classes and provide caching functionality. This class replaces the metaclass-based indicator registration and caching mechanism from the original backtrader implementation. """ _indcol: dict = {} _icache: dict = {} _icacheuse = False
[docs] @classmethod def register(cls, name, indicator_cls): """Register an indicator class in the registry. Args: name: Name of the indicator class indicator_cls: The indicator class to register """ if not name.startswith("_") and name != "Indicator": cls._indcol[name] = indicator_cls
[docs] @classmethod def cleancache(cls): """Clear the indicator cache.""" cls._icache = {}
[docs] @classmethod def usecache(cls, onoff): """Enable or disable indicator caching. Args: onoff: If True, enable caching; if False, disable it """ cls._icacheuse = onoff
[docs] @classmethod def get_cached_or_create(cls, indicator_cls, *args, **kwargs): """Get cached indicator instance or create new one. Args: indicator_cls: The indicator class to instantiate *args: Positional arguments for the indicator **kwargs: Keyword arguments for the indicator Returns: Cached indicator instance if available and caching enabled, otherwise a new indicator instance """ if not cls._icacheuse: return indicator_cls(*args, **kwargs) # Implement a cache to avoid duplicating lines actions ckey = (indicator_cls, tuple(args), tuple(kwargs.items())) # tuples hashable try: return cls._icache[ckey] except TypeError: # something is not hashable return indicator_cls(*args, **kwargs) except KeyError: pass # hashable but not in the cache _obj = indicator_cls(*args, **kwargs) return cls._icache.setdefault(ckey, _obj)
[docs] class Indicator(IndicatorBase): """Base class for all technical indicators in Backtrader. This class provides the foundation for creating custom indicators. It manages line data, minimum periods, and calculation logic. Indicators inherit from IndicatorBase and integrate with the LineIterator system for data flow. Attributes: _ltype: Line type set to IndType (0) for indicators csv: Whether to output this indicator to CSV (default: False) aliased: Whether this indicator has an alias name """ _ltype = LineIterator.IndType csv = False
[docs] def __getitem__(self, ago): """CRITICAL FIX: Forward item access to the first line (e.g., sma line) For indicators with named lines like SMA (which has lines.sma), accessing indicator[0] should return the value from the first line, not the indicator's own array. """ # Use the first line if available if hasattr(self, "lines") and hasattr(self.lines, "lines") and len(self.lines.lines) > 0: return self.lines.lines[0][ago] # Fallback to parent class behavior return super().__getitem__(ago)
# Track if this is an aliased indicator aliased = False
[docs] def __init_subclass__(cls, **kwargs): """Handle subclass registration and initialization. This method is called when a subclass of Indicator is created. It performs: 1. Lines creation using Lines infrastructure 2. Automatic registration in IndicatorRegistry 3. Alias handling for module-level access 4. next/once method setup for calculation modes Args: **kwargs: Additional keyword arguments """ super().__init_subclass__(**kwargs) init = cls.__dict__.get("__init__") if init is not None and not getattr(init, "_bt_owner_context_wrapped", False): def owner_context_init(self, *args, **kwargs): parent_owner = OwnerContext.get_current_owner(LineIterator) with OwnerContext.set_owner(self): result = init(self, *args, **kwargs) if ( parent_owner is not None and parent_owner is not self and hasattr(parent_owner, "addindicator") ): old_owner = getattr(self, "_owner", None) if old_owner is not parent_owner: try: old_lists = getattr(old_owner, "_lineiterators", {}) for indicators in old_lists.values(): while self in indicators: indicators.remove(self) except Exception: # nosec B110 # Best-effort detach from a previous owner; ignore failures. pass self._owner = parent_owner parent_owner.addindicator(self) return result owner_context_init.__name__ = getattr(init, "__name__", "__init__") owner_context_init.__doc__ = getattr(init, "__doc__", None) owner_context_init._bt_owner_context_wrapped = True cls.__init__ = owner_context_init # CRITICAL FIX: Handle lines creation for indicators like LineSeries does # This ensures that lines tuples are converted to Lines instances lines = cls.__dict__.get("lines", ()) extralines = cls.__dict__.get("extralines", 0) # Ensure lines is a tuple (it might be a class type) if not isinstance(lines, (tuple, list)): if hasattr(lines, "_getlines"): lines = lines._getlines() or () else: lines = () else: lines = tuple(lines) # Ensure it's a tuple # Create lines class using the proper Lines infrastructure if lines or extralines: # Use the LineSeries mechanism to create the lines class from .lineseries import Lines cls.lines = Lines._derive("lines", lines, extralines, ()) # NOTE: __init__ patching for _finalize_minperiod disabled as it's handled elsewhere # The minperiod calculation is now done explicitly in indicators that need it (like MACD) # Register subclasses automatically if not cls.aliased and cls.__name__ != "Indicator" and not cls.__name__.startswith("_"): IndicatorRegistry.register(cls.__name__, cls) # Handle aliases - register them to the indicators module if hasattr(cls, "alias") and cls.alias: import sys indicators_module = sys.modules.get("backtrader.indicators") if indicators_module: # Set the main class name setattr(indicators_module, cls.__name__, cls) # Set all aliases - handle both tuple and list formats aliases = cls.alias if isinstance(aliases, (list, tuple)): for alias in aliases: if isinstance(alias, str): setattr(indicators_module, alias, cls) # Check if next and once have both been overridden # Define default methods if they don't exist if not hasattr(cls, "next"): cls.next = lambda self: None if not hasattr(cls, "once"): cls.once = lambda self, start, end: None next_over = getattr(cls, "next", None) != getattr(Indicator, "next", None) once_over = getattr(cls, "once", None) != getattr(Indicator, "once", None) # CRITICAL FIX: Also check if once() is the no-op from LineRoot # If once is inherited from LineRoot (which is just 'pass'), treat it as not overridden # This handles indicators that only set up line bindings without defining next/once from .lineroot import LineRoot if hasattr(LineRoot, "once") and getattr(cls, "once", None) == getattr( LineRoot, "once", None ): # LineRoot.once is a no-op, so always use once_via_next cls.once = cls.once_via_next cls.preonce = cls.preonce_via_prenext cls.oncestart = cls.oncestart_via_nextstart elif next_over and not once_over: # No -> need pointer movement to once simulation via next cls.once = cls.once_via_next cls.preonce = cls.preonce_via_prenext cls.oncestart = cls.oncestart_via_nextstart
# Cache related methods - moved from metaclass
[docs] @classmethod def cleancache(cls): """Clear the indicator cache""" IndicatorRegistry.cleancache()
[docs] @classmethod def usecache(cls, onoff): """Enable or disable caching""" IndicatorRegistry.usecache(onoff)
def _finalize_minperiod(self): """CRITICAL FIX: Finalize minimum period calculation after indicator __init__ completes. This method is called after the subclass's __init__ has finished creating sub-indicators and line bindings. It ensures that the minimum periods from all data sources, lines and sub-indicators are properly propagated to this indicator's _minperiod. """ # Step 0: Calculate minperiod from data sources first # This is critical for indicators applied to other indicators/lines try: if hasattr(self, "datas") and self.datas: data_minperiods = [getattr(d, "_minperiod", 1) for d in self.datas if d is not None] if data_minperiods: data_max = max(data_minperiods) if data_max > self._minperiod: self._minperiod = data_max except (AttributeError, TypeError): # No usable datas to derive a minperiod from; keep current value. pass # Step 1: Calculate minperiod from lines try: if hasattr(self, "lines") and self.lines is not None: line_minperiods = [] for line in self.lines: mp = getattr(line, "_minperiod", 1) line_minperiods.append(mp) if line_minperiods: lines_max = max(line_minperiods) if lines_max > self._minperiod: self._minperiod = lines_max except (AttributeError, TypeError): # Lines not iterable yet; keep current minperiod. pass # Step 2: Calculate minperiod from sub-indicators try: if hasattr(self, "_lineiterators"): indicators = self._lineiterators.get(LineIterator.IndType, []) if indicators: ind_minperiods = [getattr(ind, "_minperiod", 1) for ind in indicators] if ind_minperiods: ind_max = max(ind_minperiods) if ind_max > self._minperiod: self._minperiod = ind_max except (AttributeError, TypeError): # No sub-indicator registry available; keep current minperiod. pass # Step 3: Update minperiod on all lines try: if hasattr(self, "lines") and self.lines is not None: for line in self.lines: if hasattr(line, "updateminperiod"): line.updateminperiod(self._minperiod) except (AttributeError, TypeError): # Lines not iterable; minperiod propagation is best-effort. pass
[docs] def advance(self, size=1): """Advance indicator lines when data length is less than clock length. Also advances sub-indicators so that during _oncepost() replay every level of the indicator tree stays in sync (fixes runonce ATR/SMMA index mismatch when an indicator uses sub_ind[0] in next()). Args: size: Number of steps to advance (default: 1) """ # Prefer the concrete secondary-feed clock resolved in # Strategy._periodset() for indicators that follow a non-primary feed # (e.g. SMA over an H1 LinesOperation inside an M15 strategy). Their # _clock may point at a feed whose len() is correct, but when an # explicit secondary clock was pinned we use it so the indicator # advances in lockstep with that feed. See # docs/DEV_REGRESSION_FAILURES.md. adv_clock = getattr(self, "_resolved_secondary_clock", None) or self._clock if len(self) < len(adv_clock): self.lines.advance(size=size) for ind in self._lineiterators.get(LineIterator.IndType, []): ind.advance(size)
[docs] def preonce_via_prenext(self, start, end): """Implement preonce using prenext for batch calculation. This is a generic implementation if prenext is overridden but preonce is not. It loops through the range and calls prenext for each step. Args: start: Starting index end: Ending index """ # Generic implementation if prenext is overridden but preonce is not for i in range(start, end): # Advance all data feeds for data in self.datas: data.advance() # Advance all sub-indicators for indicator in self._lineiterators[LineIterator.IndType]: indicator.advance() # CRITICAL FIX: Directly advance lines instead of using self.advance() self.lines.advance() # Call prenext self.prenext()
[docs] def oncestart_via_nextstart(self, start, end): """Implement oncestart using nextstart for batch calculation. This is used when nextstart is overridden but oncestart is not. Args: start: Starting index end: Ending index """ # nextstart has been overridden, but oncestart has not - call the overridden nextstart for i in range(start, end): for data in self.datas: data.advance() for indicator in self._lineiterators[LineIterator.IndType]: indicator.advance() # CRITICAL FIX: Directly advance lines instead of using self.advance() self.lines.advance() self.nextstart()
[docs] def once_via_next(self, start, end): """Implement once using next for batch calculation. This is used when next is overridden but once is not. It loops through the range and calls next for each step. Args: start: Starting index end: Ending index """ # Not overridden, next must be there ... # Simple implementation matching master branch - just advance and call next for i in range(start, end): for data in self.datas: data.advance() for indicator in self._lineiterators[LineIterator.IndType]: indicator.advance() # CRITICAL FIX: Directly advance lines instead of using self.advance() # self.advance() checks len(self) < len(self._clock) which fails when # _clock is MinimalClock (always returns 0) or when _clock is not properly # synchronized. In once_via_next, we always need to advance. self.lines.advance() self.next()
[docs] class LinePlotterIndicatorBase(Indicator.__class__): """Base class for indicators that plot multiple lines. Note: These classes are not currently used in the project. They are kept for compatibility with the original backtrader. """
[docs] def donew(cls, *args, **kwargs): """Create a new LinePlotterIndicator instance. Args: *args: Positional arguments **kwargs: Keyword arguments, must include 'name' Returns: tuple: (created_object, args, kwargs) """ # Get line name lname = kwargs.pop("name") # Get class name name = cls.__name__ # Get cls lines, or return Lines if not present lines = getattr(cls, "lines", Lines) # Derive lines with the new line cls.lines = lines._derive(name, (lname,), 0, []) # Derive plotlines plotlines = AutoInfoClass newplotlines: dict = {} newplotlines.setdefault(lname, {}) cls.plotlines = plotlines._derive(name, newplotlines, [], recurse=True) # Create the object and set the params in place _obj, args, kwargs = super().donew(*args, **kwargs) # Set _obj owner attribute _obj.owner = _obj.data.owner._clock # Add another linebuffer _obj.data.lines[0].addbinding(_obj.lines[0]) # Return the object and arguments to the chain return _obj, args, kwargs
[docs] class LinePlotterIndicator(Indicator, LinePlotterIndicatorBase): """Indicator that plots multiple lines. Note: This class is not currently used in the project. """