WARNING: This project is no longer actively maintained.

It has been a few years since I have supported FIX connections in production, and I am no longer making use of this codebase myself anymore.

We have an issue open to look for a new maintainer. Please comment there if you are interested in becoming involved.

WTF(ix)

The Pythonic Financial Information eXchange (FIX) client that you have been looking for.

Background

WTF(ix) is a simple and easy to use FIX client for Python. It can also be adapted to plug into various crypto exchanges that do not adhere to the formal FIX (or any other) specification.

WTF(ix) does not require you to alter your development stack, deploy a database, compile a C++ codebase and figure out weird API bindings, or wrangle with arcane XML file definitions.

Project Highlights and Goals

Batteries included - comes with everything that you need to connect to a FIX server and start sending and receiving messages in minutes. Provides default implementations for:
- Authentication
- Maintaining a heartbeat
- Sequence number management and resend requests
- Message storage and retrieval
Fast, easy to understand message processing pipeline based on a modern async and await implementation.

Easily extendable architecture - modular 'apps' can be added to the pipeline stack to add custom message processing routines or new application features.

PIPELINE_APPS = [
    "wtfix.apps.utils.PipelineTerminationApp", # Post-processing / cleanup
    "my_app.apps.SecretAlgoTradingRecipe",     # <-- Your application logic
    "wtfix.apps.api.RESTfulServiceApp",        # REST API for sending messages
    "wtfix.apps.brokers.RedisPubSubApp",       # Redis Pub/Sub broker for sending / receiving messages
    "wtfix.apps.admin.HeartbeatApp",           # Heartbeat monitoring and maintenance
    "wtfix.apps.admin.AuthenticationApp",      # Login / logout handling
    "wtfix.apps.admin.SeqNumManagerApp",       # Message gap detection and filling
    "wtfix.apps.store.MessageStoreApp",        # Store messages (caching or persistence)
    "wtfix.apps.utils.InboundLoggingApp",      # Log inbound messages
    "wtfix.apps.parsers.RawMessageParserApp",  # Message parsing: Logon (A): {BeginString (8): FIX.4.4 | BodyLength (9): 99 | MsgType (35): A | MsgSeqNum (34): 1 | SenderCompID (49): SENDER | SendingTime (52): 20190305-08:45:45.979 | TargetCompID (56): TARGET | EncryptMethod (98): 0 | HeartBtInt (108): 30 | Username (553): USERNAME | Password (554): PASSWORD | ResetSeqNumFlag (141): Y | CheckSum (10): 94}
    "wtfix.apps.utils.OutboundLoggingApp",     # Log outbound messages
    "wtfix.apps.wire.WireCommsApp",            # Raw message encoding / decoding: b'8=FIX.4.4\x019=99\x0135=A\x0134=1\x0149=SENDER\x0152=20190305-08:42:32.793\x0156=TARGET\x0198=0\x01108=30\x01553=USERNAME\x01554=PASSWORD\x01141=Y\x0110=081\x01'
    "wtfix.apps.sessions.ClientSessionApp",    # HTTP session management
]

Messages can be cached in memory or saved to a Redis message store for later retrieval. Alternatively you can add your own message storage solution using the provided interfaces.
Send messages directly from the pipeline, via 3rd party applications using a REST API, or by publishing them to a Redis Pub/Sub channel for immediate delivery.

Provides a convenient @on decorator for fine-grained control over which apps will respond to which types of messages:

from wtfix.apps.base import MessageTypeHandlerApp, on
from wtfix.conf import settings

logger = settings.logger

class SecretAlgoTradingRecipe(MessageTypeHandlerApp):

    @on(context.protocol.MsgType.Logon)  # Only invoked when 'Logon (type A)' messages are received.
    def on_logon(self, message):
        self.send_security_definition_request()
        return message

    def on_receive(self, message):  # Invoked for every type of message.
      logger.info(f"Received message {message}!")

Provides custom Field and FieldMap types for working with FIX tags and field values. These types are 'pythonic', implementing many of the standard protocols, and behave as expected when you integrate them in existing Python code.

A simple message tag syntax, with various convenience methods, for quick access to commonly used message attributes.

>>> from wtfix.message import admin

# Instantiate a new 'Logon' message
>>> logon_msg = admin.LogonMessage("my_username", "my_password", heartbeat_int=30)

# Short, concise string representation
>>> str(logon_msg)
'A: {(35, A) | (98, 0) | (108, 30) | (553, my_username) | (554, my_password)}'

# Pretty print tag names by using the 't' formatting option
>>> f"{logon_msg:t}"
'Logon (A): {MsgType (35): A | EncryptMethod (98): 0 | HeartBtInt (108): 30 | Username (553): my_username | Password (554): my_password}'

# Example of getting the message type
>>> logon_msg.type
'A'

# Example of getting the message type name
>>> logon_msg.name
'Logon'

# Look up the sequence number
>>> logon_msg.seq_num
1

# Various ways for accessing the different fields that make up the message. Fields are just
# (tag, value) namedtuples.
>>> logon_msg[108]  # Using old school tag number
Field(108, '30')

>>> logon_msg[context.protocol.Tag.HeartBtInt]  # Using the tag name as per the FIX specification
Field(108, '30')

>>> logon_msg.HeartBtInt  # Using tag name shortcut
Field(108, '30')

A pragmatic unicode sandwich based approach to encoding / decoding messages mean that you never need to deal with byte sequences directly.

from wtfix.message.field import Field
from wtfix.message.message import generic_message_factory

# Create a new Message from a byte sequence received over the wire
>>> fields = Field.fields_frombytes(b"35=A\x0198=0\x01108=30\x01553=my_username\x01554=my_password\x01")
>>> logon_msg = generic_message_factory(*fields)

>>> str(logon_msg)
'A: {(35, A) | (98, 0) | (108, 30) | (553, my_username) | (554, my_password)}'

# Fields are tuples of (tag, value) pairs
>>> username = logon_msg.Username

>>> username.tag
553

>>> username.value
"my_username"

# Fields behave just like Python's built-in types, and most operations can be performed directly
# on a field's 'value' attribute.
>>> username += "_123"
>>> username
Field(553, 'my_username_123')

Access the underlying byte sequence when you need it by casting the message back to bytes:

>>> bytes(logon_msg)
b'35=A\x0198=0\x01108=30\x01553=my_username\x01554=my_password\x01'

It is easy to add Fields to a Message: simply assign the tag value:

>>> logon_msg.PossDupFlag = "Y"
>>> f"{logon_msg:t}"
'Logon (A): {MsgType (35): A | EncryptMethod (98): 0 | HeartBtInt (108): 30 | Username (553): my_username | Password (554): my_password | PossDupFlag (43): Y}'

# Most FIX field types can be cast to equivalent Python built-in types to make processing easier
>>> bool(logon_msg.PossDupFlag) is True
True

A very forgiving approach to parsing repeating groups of message tags; with or without pre-defined templates:

from wtfix.protocol.common import Tag, MsgType
from wtfix.message.message import generic_message_factory

# If you provide a group template, then messages are stored in an 'OrderedDict' for fast lookups
>>> msg = generic_message_factory((Tag.MsgType, MsgType.ExecutionReport), (Tag.NoMiscFees, 2), (Tag.MiscFeeAmt, 10.00), (Tag.MiscFeeType, 2), (Tag.MiscFeeAmt, 20.00), (Tag.MiscFeeType, "A"), group_templates={Tag.NoMiscFees: [Tag.MiscFeeAmt, Tag.MiscFeeType,]})
>>> msg.data
OrderedDict([(35, Field(35, '8')), (136, Group(Field(136, '2'), Field(137, '10.0'), Field(139, '2'), Field(137, '20.0'), Field(139, 'A')))])

# Get 'NoMiscFees' group
>>> group = msg.NoMiscFees
>>> f"{group:t}"
'[NoMiscFees (136): 2] | [MiscFeeAmt (137): 10.0 | MiscFeeType (139): 2] | [MiscFeeAmt (137): 20.0 | MiscFeeType (139): A]'

# Determine the number of instances in the group
>>> group.size
2

# Retrieve the second group instance
>>> group.instances[1]
FieldList(Field(137, '20.0'), Field(139, 'A'))

# Without a pre-defined group template, WTFIX falls back to using a (slightly slower) list structure for representing message fields internally
>>> msg = generic_message_factory((Tag.MsgType, MsgType.ExecutionReport), (Tag.NoMiscFees, 2), (Tag.MiscFeeAmt, 10.00), (Tag.MiscFeeType, 2), (Tag.MiscFeeAmt, 20.00), (Tag.MiscFeeType, "A"))
>>> msg.data
[Field(35, '8'), Field(136, '2'), Field(137, '10.0'), Field(139, '2'), Field(137, '20.0'), Field(139, 'A')]

Getting Started

Install the project's dependencies (e.g. pip install -r requirements/local.txt), preferably in a Python virtual environment that has been created specifically for that purpose.
Run the test suite with pytest to verify the installation.

Create a .env file in the project's root directory that contains at least the following configuration settings:

# Supports different configuration settings for local development, staging, or production environments.
WTFIX_SETTINGS_MODULE=config.settings.local

HOST=             # Required. The FIX server hostname or IP address
PORT=             # Required. The port on the FIX server to connect to

SENDER=           # Required. SENDER_COMP_ID (tag 49).
TARGETD=          # Required. TARGET_COMP_ID (tag 56).

USERNAME=         # Required. Username to use for Logon messages (tag 553).
PASSWORD=         # Required. Password to use for logon messages (tag 554).

PYTHONASYNCIODEBUG=0  # Set to '1' for detailed debugging messages.

Start the FIX client with python runclient.py. The default implementation will log in to the FIX server and maintain a steady heartbeat.
Use Ctrl-C to quit. This will trigger a Logout message to be sent before the pipeline is terminated.

Project Resources

Need More?

WTF(ix) is open source, free, and fully functional. However if you are a corporate intending to make use of WTF(ix) commercially, then you may need more support and assurance than can be offered by the community here.

Commercial support is available - get in touch if you need help with:

Getting started and development best practices
Integrating WTF(ix) with the exchange of your choice (FIX, websocket stream, or REST API)
Sample code (including persisting messages and off-loading tasks to background processing libraries)
Code reviews
Running unmonitored instances in production
Problem investigation and identification

Inspired By

slowbreak's message processing pipeline and @on decorator
simplefix's approach to raw message parsing

jcass77 / WTFIX