Migrating from Classic Zipline

Ziplime keeps the event-driven strategy model of classic Zipline, but the runtime is not a drop-in replacement. The biggest differences are async strategy functions, Polars data frames, explicit execution styles, and a data-bundle runner built around Ziplime services. Use this page when converting an old Zipline algorithm into a Ziplime algorithm file.

What still feels familiar

The core strategy shape is the same:

def initialize(context):
    ...


def handle_data(context, data):
    ...

In Ziplime, the main functions become async:

async def initialize(context):
    ...


async def handle_data(context, data):
    ...

The same mental model still applies:

Store persistent strategy state on context.
Use handle_data for per-bar trading logic.
Use schedule_function for periodic rebalances.
Use record for custom metrics.
Use target order helpers for rebalancing.

Quick conversion table

Classic Zipline pattern	Ziplime pattern
`from zipline.api import symbol`	Prefer `await context.symbol(...)` inside `initialize`.
`asset = symbol("AAPL")`	`asset = await context.symbol("AAPL")`
`symbols("AAPL", "MSFT")`	`[await context.symbol(s) for s in ["AAPL", "MSFT"]]`
`data.current(asset, "close")`	`(await data.current(assets=[asset], fields=["close"]))["close"][0]`
`data.history(asset, "close", 20, "1d")`	`await data.history(assets=[asset], fields=["close"], bar_count=20)`
`order(asset, 100)`	`await context.order(asset, 100, style=MarketOrder())`
`order_target_percent(asset, 0.5)`	`await context.order_target_percent(asset, 0.5, style=MarketOrder())`
Pandas results from data access	Polars data frames from `data.current` and `data.history`.
Synchronous `handle_data`	`async def handle_data(...)` and `await` async API calls.
`run_algorithm(...)` with in-memory data	`run_simulation(...)` with loaded Ziplime bundles and services.

Imports

Classic Zipline algorithms often import many API functions:

from zipline.api import (
    symbol,
    order_target_percent,
    record,
    schedule_function,
    date_rules,
    time_rules,
)

In Ziplime, prefer direct context calls and import only helper classes:

from ziplime.finance.execution import MarketOrder
from ziplime.utils.events import date_rules, time_rules

The Zipline-style ziplime.api namespace exists for compatibility, but direct context calls are easier to read in async code:

asset = await context.symbol("AAPL")
await context.order_target_percent(asset, 0.25, style=MarketOrder())
context.record(weight=0.25)

Lifecycle functions

`initialize`

Classic Zipline:

def initialize(context):
    context.asset = symbol("AAPL")

Ziplime:

async def initialize(context):
    context.asset = await context.symbol("AAPL")

Use initialize to look up assets, set strategy parameters, register scheduled callbacks, attach pipelines, and configure controls.

`handle_data`

Classic Zipline:

def handle_data(context, data):
    price = data.current(context.asset, "close")
    if price > 100:
        order_target_percent(context.asset, 1.0)

Ziplime:

from ziplime.finance.execution import MarketOrder


async def handle_data(context, data):
    current = await data.current(assets=[context.asset], fields=["close"])
    price = current["close"][0]

    if price > 100:
        await context.order_target_percent(
            context.asset,
            1.0,
            style=MarketOrder(),
        )

`before_trading_start`

Classic Zipline used a synchronous before_trading_start. Ziplime currently does the same.

def before_trading_start(context, data):
    context.traded_today = False

Do not define it as async def in the current Ziplime runtime. Do not place orders in this hook.

`analyze`

Classic Zipline usually passed a performance DataFrame to analyze. Ziplime also calls analyze synchronously with the final performance table.

def analyze(context, perf):
    print(perf.tail())

Market data differences

Classic Zipline allowed scalar-style calls:

price = data.current(asset, "close")
history = data.history(asset, "close", 20, "1d")

Ziplime expects lists and returns a Polars DataFrame:

current = await data.current(
    assets=[asset],
    fields=["close", "volume"],
)

price = current["close"][0]
volume = current["volume"][0]

Historical windows are also async:

history = await data.history(
    assets=[asset],
    fields=["close"],
    bar_count=20,
)

mean_close = history["close"].mean()

If your old strategy uses pandas operations, convert the Ziplime result explicitly:

pandas_history = history.to_pandas()

Prefer Polars-native expressions for new code.

Order differences

Classic Zipline often allowed:

order(asset, 100)
order_target_percent(asset, 0.5)

Ziplime order calls are async and require an execution style:

from ziplime.finance.execution import MarketOrder, LimitOrder

await context.order(asset, 100, style=MarketOrder())
await context.order_target_percent(asset, 0.5, style=MarketOrder())
await context.order(asset, 100, style=LimitOrder(limit_price=150.0))

Target order helpers do not account for still-open orders. If the old strategy repeatedly submits target orders, add an open-order guard:

if not context.get_open_orders(asset):
    await context.order_target_percent(asset, 0.5, style=MarketOrder())

Symbol lookup

Classic Zipline usually resolved symbols from an asset database with the simulation lookup date. Ziplime resolves symbols through its asset service:

asset = await context.symbol("AAPL")
asset = await context.symbol("AAPL", mic="XNGS")
asset = await context.symbol("AAPL@XNGS")

When a ticker can exist on multiple exchanges, pass mic or use the SYMBOL@MIC form. For named universes:

context.q100us = await context.symbols_universe("Q100US")

Scheduling

Classic Zipline:

def initialize(context):
    schedule_function(
        rebalance,
        date_rule=date_rules.month_end(),
        time_rule=time_rules.market_close(minutes=30),
    )


def rebalance(context, data):
    order_target_percent(context.asset, 1.0)

Ziplime:

from ziplime.finance.execution import MarketOrder
from ziplime.utils.events import date_rules, time_rules


async def initialize(context):
    context.schedule_function(
        rebalance,
        date_rule=date_rules.month_end(),
        time_rule=time_rules.market_close(minutes=30),
    )


async def rebalance(context, data):
    await context.order_target_percent(
        context.asset,
        1.0,
        style=MarketOrder(),
    )

In daily simulations, time rules are effectively ignored because the clock emits one bar per session. In minute simulations, time rules matter.

Portfolio and positions

Classic Zipline commonly used:

amount = context.portfolio.positions[asset].amount
cash = context.portfolio.cash

Ziplime exposes portfolio totals directly, but position storage is nested by exchange and account in the current runtime. Prefer helper methods:

cash = context.portfolio.cash
portfolio_value = context.portfolio.portfolio_value
amount = await context.portfolio.get_asset_positions_amount(asset)
value = await context.portfolio.get_asset_positions_value(asset)

For exact exchange-asset positions:

amount = await context.portfolio.get_exchange_asset_positions_amount(
    asset,
    exchange_name="LIME",
)

Recording metrics

This part is close to Zipline:

context.record(signal=signal, cash=context.portfolio.cash)

Recorded values become columns in the final performance table.

Pipeline migration

Ziplime includes a Zipline-like Pipeline API:

from ziplime.pipeline import Pipeline
from ziplime.pipeline.data import EquityPricing
from ziplime.pipeline.terms.factors import SimpleMovingAverage


def make_pipeline():
    sma_20 = SimpleMovingAverage(
        inputs=[EquityPricing.close],
        window_length=20,
    )
    return Pipeline(columns={"sma_20": sma_20})

Attach the pipeline in initialize:

async def initialize(context):
    context.attach_pipeline(make_pipeline(), "signals")

Read output after initialization:

def before_trading_start(context, data):
    context.signals = context.pipeline_output("signals")

Pipeline support depends on loaders. Pricing data via EquityPricing is the standard path; custom datasets require custom loaders.

Running migrated algorithms

Classic Zipline examples often call run_algorithm(...) directly with pandas data or bundle names. Ziplime’s standard flow is:

Ingest or load data into a Ziplime bundle.
Load the bundle with bundle_service.load_bundle(...).
Pass the loaded data source to run_simulation(...).
Point algorithm_file at the migrated .py strategy file.

result = await run_simulation(
    start_date=start_date,
    end_date=end_date,
    trading_calendar="NYSE",
    algorithm_file="my_migrated_algo.py",
    total_cash=100_000.0,
    market_data_source=market_data,
    custom_data_sources=[],
    emission_rate=datetime.timedelta(days=1),
    benchmark_asset_symbol="AAPL",
    stop_on_error=True,
    asset_service=asset_service,
)

Full before and after

Classic Zipline:

from zipline.api import order_target_percent, record, symbol


def initialize(context):
    context.asset = symbol("AAPL")
    context.window = 20


def handle_data(context, data):
    history = data.history(context.asset, "close", context.window, "1d")
    price = data.current(context.asset, "close")
    mean_price = history.mean()

    if price > mean_price:
        order_target_percent(context.asset, 1.0)
    else:
        order_target_percent(context.asset, 0.0)

    record(price=price, mean_price=mean_price)

Ziplime:

from ziplime.finance.execution import MarketOrder


async def initialize(context):
    context.asset = await context.symbol("AAPL")
    context.window = 20


async def handle_data(context, data):
    history = await data.history(
        assets=[context.asset],
        fields=["close"],
        bar_count=context.window,
    )
    current = await data.current(
        assets=[context.asset],
        fields=["close"],
    )

    price = current["close"][0]
    mean_price = history["close"].mean()

    target = 1.0 if price > mean_price else 0.0

    if not context.get_open_orders(context.asset):
        await context.order_target_percent(
            context.asset,
            target,
            style=MarketOrder(),
        )

    context.record(price=price, mean_price=mean_price, target=target)

Migration checklist

Replace zipline.api imports with direct context calls and Ziplime helper imports.
Make initialize, handle_data, and scheduled callbacks async.
Add await to asset lookup, data access, order placement, and order cancellation.
Pass lists to data.current and data.history.
Update pandas assumptions to Polars or call .to_pandas() explicitly.
Add an execution style such as MarketOrder() to order calls.
Use def before_trading_start, not async def before_trading_start.
Use def analyze, not async def analyze.
Replace direct portfolio.positions[asset] access with portfolio helper methods.
Pass MIC codes for ambiguous symbols.
Move runner setup to run_simulation(...) with Ziplime data bundles.

Introduction

Writing Strategies

Strategy Language

Migrating to Ziplime

Limitations

Migrating from Classic Zipline

What still feels familiar

Quick conversion table

Imports

Lifecycle functions

`initialize`

`handle_data`

`before_trading_start`

`analyze`

Market data differences

Order differences

Symbol lookup

Scheduling

Portfolio and positions

Recording metrics

Pipeline migration

Running migrated algorithms

Full before and after

Migration checklist

​What still feels familiar

​Quick conversion table

​Imports

​Lifecycle functions

​initialize

​handle_data

​before_trading_start

​analyze

​Market data differences

​Order differences

​Symbol lookup

​Scheduling

​Portfolio and positions

​Recording metrics

​Pipeline migration

​Running migrated algorithms

​Full before and after

​Migration checklist

What still feels familiar

Quick conversion table

Imports

Lifecycle functions

`initialize`

`handle_data`

`before_trading_start`

`analyze`

Market data differences

Order differences

Symbol lookup

Scheduling

Portfolio and positions

Recording metrics

Pipeline migration

Running migrated algorithms

Full before and after

Migration checklist