backtrader.feed module

Data Feed Module - Financial data feed implementations.

This module provides the base classes and implementations for data feeds in backtrader. Data feeds are the source of price/volume data for strategies and indicators.

Key Classes:

AbstractDataBase: Base class for all data feeds with core functionality. DataBase: Full-featured data feed with replay/resample support. CSVDataBase: Base class for CSV file data feeds. FeedBase: Base for live/real-time data feeds.

Data feeds provide:

• OHLCV (Open, High, Low, Close, Volume) data

• Timeline management and session handling

• Replay and resampling capabilities

• Live data support for trading

Example

Creating a custom data feed: >>> class MyDataFeed(CSVDataBase): … params = ((‘dataname’, ‘data.csv’),)

class backtrader.feed.AbstractDataBase

Bases: OHLCDateTime

Base class for all data feed implementations.

Provides the core functionality for data feeds including: - Data loading and preprocessing - Timeline management - Session handling - Live data support - Notification system for data status changes

States:

CONNECTED, DISCONNECTED, CONNBROKEN, DELAYED, LIVE, NOTSUBSCRIBED, NOTSUPPORTED_TF, UNKNOWN

Params:

dataname: Data source identifier (filename, URL, etc.). name: Display name for the data feed. compression: Timeframe compression factor. timeframe: TimeFrame period (Days, Minutes, etc.). fromdate: Start date for data filtering. todate: End date for data filtering. sessionstart: Session start time. sessionend: Session end time. filters: List of data filters to apply. tz: Output timezone. tzinput: Input timezone. qcheck: Timeout in seconds for live event checking. calendar: Trading calendar to use.

Example

>>> data = AbstractDataBase(dataname='data.csv')
>>> cerebro.adddata(data)

CONNECTED = 0

DISCONNECTED = 1

CONNBROKEN = 2

DELAYED = 3

LIVE = 4

NOTSUBSCRIBED = 5

NOTSUPPORTED_TF = 6

UNKNOWN = 7

__init__(*args, **kwargs)

Initialize a LineMultiple instance.

Sets up the internal state for managing multiple lines, including line type indicator, lines collection, clock reference, and line iterator tracking.

Initializes:

_ltype: Line type indicator (None for base LineMultiple). lines: Collection of line objects (creates if not exists). _clock: Clock reference for synchronization. _lineiterators: Dictionary tracking registered lineiterators. _minperiod: Minimum period requirement (defaults to 1).

resampling = 0

replaying = 0

date2num(dt)

Convert datetime to internal numeric format.

Parameters:

dt – datetime object to convert.

Returns:

Internal numeric representation of the datetime.

Return type:

float

num2date(dt=None, tz=None, naive=True)

Convert internal numeric format to datetime.

Parameters:

• dt – Numeric datetime value (uses current if None).

• tz – Timezone to use (uses feed tz if None).

• naive – Return naive datetime if True.

Returns:

Converted datetime object.

Return type:

datetime

haslivedata()

Check if this data feed has live data.

Returns:

False for base class, override for live data feeds.

Return type:

bool

do_qcheck(onoff, qlapse)

Calculate wait interval for queue checking.

Parameters:

• onoff – Whether queue checking is enabled.

• qlapse – Time elapsed since last check.

islive()

If this returns True, Cerebro will deactivate preload and runonce because a live data source must be fetched tick by tick (or bar by bar)

put_notification(status, *args, **kwargs)

Add arguments to notification queue

get_notifications()

Return the pending “store” notifications

getfeed()

Get the parent feed object.

Returns:

The parent feed instance.

Return type:

FeedBase or None

qbuffer(savemem=0, replaying=False)

Apply queued buffering to all lines.

Parameters:

• savemem – Memory saving mode.

• replaying – Whether replaying is active.

start()

Start the data feed.

Resets internal queues and sets initial status to CONNECTED.

stop()

Stop the data feed.

Override in subclasses for cleanup.

clone(**kwargs)

Create a clone of this data feed.

Parameters:

**kwargs – Additional keyword arguments for the clone.

Returns:

A cloned data feed.

Return type:

DataClone

copyas(_dataname, **kwargs)

Copy the data feed with a different name.

Parameters:

• _dataname – New name for the data feed.

• **kwargs – Additional keyword arguments.

Returns:

A cloned data feed with the new name.

Return type:

DataClone

setenvironment(env)

Keep a reference to the environment

getenvironment()

Get the cerebro environment reference.

Returns:

The cerebro environment instance.

addfilter_simple(f, *args, **kwargs)

Add a simple filter wrapper to this data feed.

Parameters:

• f – Filter function to apply.

• *args – Positional arguments for the filter.

• **kwargs – Keyword arguments for the filter.

addfilter(p, *args, **kwargs)

Add a filter to this data feed.

Parameters:

• p – Filter class or instance.

• *args – Positional arguments for filter creation.

• **kwargs – Keyword arguments for filter creation.

compensate(other)

Call it to let the broker know that actions on this asset will compensate open positions in another

advance_peek()

Peek at the datetime of the next bar.

Returns:

Numeric datetime of next bar, or inf if unavailable.

Return type:

float

advance(size=1, datamaster=None, ticks=True)

Advance the data feed by the specified size.

Parameters:

• size – Number of bars to advance (default: 1).

• datamaster – Master data feed for synchronization.

• ticks – Whether to process tick data.

next(datamaster=None, ticks=True)

Move to the next bar.

Parameters:

• datamaster – Master data feed for synchronization.

• ticks – Whether to process tick data.

Returns:

True if a bar is available, False otherwise.

Return type:

bool

preload()

Preload all available data from the data feed.

Loads all bars and resets position to the beginning.

load()

Load the next bar from the data feed.

Returns:

True if a bar was loaded, False if no more data.

Return type:

bool

This method handles:

• Forwarding the data pointer

• Processing filters

• Checking date boundaries

resample(**kwargs)

Add a resampling filter to this data feed.

Resampling converts data to a different timeframe (e.g., minutes to days).

Parameters:

**kwargs – Arguments for the Resampler filter.

replay(**kwargs)

Add a replay filter to this data feed.

Replay filters process tick data into bars with precise control.

Parameters:

**kwargs – Arguments for the Replayer filter.

frompackages = ()

packages = ()

class backtrader.feed.DataBase

Bases: AbstractDataBase

Full-featured data feed class.

Inherits all functionality from AbstractDataBase. This is the standard data feed class for most use cases.

frompackages = ()

packages = ()

class backtrader.feed.FeedBase

Bases: object

Base class for feed containers.

Manages multiple data feeds and provides parameter processing without using metaclasses.

__init__(**kwargs)

Initialize the feed base.

Parameters:

**kwargs – Keyword arguments for parameters.

start()

Start all managed data feeds.

stop()

Stop all managed data feeds.

getdata(dataname, name=None, **kwargs)

Get or create a data feed and add it to the managed datas.

Parameters:

• dataname – Data source identifier (filename, URL, etc.).

• name – Display name for the data feed.

• **kwargs – Additional parameters for the data feed.

Returns:

The created or retrieved data feed instance.

Return type:

DataBase

class backtrader.feed.CSVDataBase

Bases: DataBase

Base class for classes implementing CSV DataFeeds

The class takes care of opening the file, reading the lines and tokenizing them.

Subclasses do only need to override:

• _loadline(tokens)

The return value of _loadline (True/False) will be the return value of _load which has been overriden by this base class

f = None

__init__(*args, **kwargs)

Initialize a LineMultiple instance.

Sets up the internal state for managing multiple lines, including line type indicator, lines collection, clock reference, and line iterator tracking.

Initializes:

_ltype: Line type indicator (None for base LineMultiple). lines: Collection of line objects (creates if not exists). _clock: Clock reference for synchronization. _lineiterators: Dictionary tracking registered lineiterators. _minperiod: Minimum period requirement (defaults to 1).

start()

Start the CSV data feed.

Opens the CSV file and optionally skips headers.

stop()

Stop the CSV data feed.

Closes the CSV file if open.

preload()

Preload all data from the CSV file.

Loads all available data and closes the file handle.

frompackages = ()

packages = ()

class backtrader.feed.CSVFeedBase

Bases: FeedBase

Base class for CSV feed containers.

Manages CSV data feeds with support for base path prefixing.

__init__(basepath='', **kwargs)

Initialize the CSV feed base.

Parameters:

• basepath – Base path to prepend to data file names.

• **kwargs – Additional keyword arguments for parameters.

class backtrader.feed.DataClone

Bases: AbstractDataBase

Clones an existing data feed.

Creates a new data feed that references an existing data feed. Useful for creating multiple views of the same data with different parameters or filters.

__init__(*args, **kwargs)

Initialize a LineMultiple instance.

Sets up the internal state for managing multiple lines, including line type indicator, lines collection, clock reference, and line iterator tracking.

Initializes:

_ltype: Line type indicator (None for base LineMultiple). lines: Collection of line objects (creates if not exists). _clock: Clock reference for synchronization. _lineiterators: Dictionary tracking registered lineiterators. _minperiod: Minimum period requirement (defaults to 1).

start()

Start the data clone.

Initializes internal tracking variables.

preload()

Preload data from the source data feed.

After preloading, resets the source data’s position.

advance(size=1, datamaster=None, ticks=True)

Advance the data clone by the specified size.

Parameters:

• size – Number of bars to advance.

• datamaster – Master data feed for synchronization (unused).

• ticks – Whether to process tick data.

frompackages = ()

packages = ()