오더북 재구성을 위한 NASDAQ ITCH 메시지 파싱

# 01_itch_parser.py


```py
# ---
# jupyter:
#   jupytext:
#     formats: ipynb,py:percent
#     text_representation:
#       extension: .py
#       format_name: percent
#       format_version: '1.3'
#       jupytext_version: 1.19.3
#   kernelspec:
#     display_name: Python 3 (ipykernel)
#     language: python
#     name: python3
# ---

# %% [markdown]
# # NASDAQ TotalView-ITCH: Order Book Data Parsing
#
# **Chapter 3: Market Microstructure**
#
# **Docker image**: `ml4t`
#
# ## Purpose
#
# This notebook demonstrates how to parse NASDAQ's TotalView-ITCH binary protocol.
# Understanding MBO (message-by-order) data is foundational for microstructure-based ML features.
#
# ## Learning Objectives
#
# After completing this notebook, you will be able to:
# - Read one ITCH v5.0 message out of a binary file: find where it starts, how long it is,
#   and which of the twenty-odd message types it is.
# - Unpack a message's fields with Python's `struct` module, and convert the two encodings
#   ITCH uses - a nanosecond offset from midnight, and a price held as an integer with four
#   implied decimal places.
# - Run the parse over a full trading day, writing each message type to its own Parquet
#   partition so that a day that does not fit in memory still lands on disk.
# - Load an already-parsed day and check that every message type reads back.
# - Say when the Python parser is the right tool and when the Rust one is.
#
# ## Book reference
#
# Section §3.3, *From raw messages to the limit order book* - the binary-parsing-at-scale
# subsection and Table 3.1. §3.2 places ITCH among the other feeds.
#
# ## Cross-References
#
# - **Downstream**: `02_itch_lob_reconstruction` (builds order book from these messages)
# - **Related**: `09_databento_mbo_analysis` (alternative MBO data source)
#
# ## Data Requirements
#
# ITCH sample data can be downloaded using:
# ```bash
# python data/equities/market/microstructure/nasdaq_itch_download.py --list   # List available files
# python data/equities/market/microstructure/nasdaq_itch_download.py          # Download default sample
# ```
#
# Files are ~5GB compressed from: https://emi.nasdaq.com/ITCH/

# %% [markdown]
# ## ITCH Message Types
#
# The [ITCH v5.0 specification](https://www.nasdaqtrader.com/content/technicalsupport/specifications/dataproducts/NQTVITCHSpecification.pdf) defines 20+ message types:
#
# | Type | Name | Description |
# |------|------|-------------|
# | **S** | System Event | Market open/close events |
# | **R** | Stock Directory | Ticker information and characteristics |
# | **H** | Trading Action | Trading halts, pauses, and resumptions |
# | **Y** | Reg SHO Restriction | Short sale price test restrictions |
# | **L** | Market Participant | Market maker positions |
# | **V** | MWCB Decline Level | Market-wide circuit breaker levels |
# | **W** | MWCB Status | Circuit breaker breach status |
# | **A** | Add Order | New limit order enters the book |
# | **F** | Add Order (MPID) | Same as A, with market participant ID |
# | **E** | Order Executed | Partial/full execution against standing order |
# | **C** | Order Executed w/Price | Execution at different price (hidden orders) |
# | **X** | Order Cancel | Partial cancellation |
# | **D** | Order Delete | Full removal from book |
# | **U** | Order Replace | Modify price/size (cancel + add) |
# | **P** | Trade | Non-displayed execution |
# | **Q** | Cross Trade | Opening/closing cross |
# | **B** | Broken Trade | Trade cancellation |
# | **I** | NOII | Net Order Imbalance Indicator (auction) |
# | **J** | LULD Auction Collar | Limit up-limit down price bands |
# | **K** | IPO Quoting Period | IPO quotation timing |
#
# By combining these messages chronologically, we can reconstruct the order book at any point in time.

# %%
"""NASDAQ TotalView-ITCH: Order Book Data Parsing — parse ITCH binary protocol into structured messages."""

import gzip
import os
import shutil
import struct
from collections import Counter, defaultdict
from datetime import date, datetime
from pathlib import Path
from time import time

import polars as pl
from itch_message_specs import (
    FMT_DICT,
    MESSAGE_SPECS,
    NT_DICT,
    flush_to_parquet,
    parse_price4,
    parse_timestamp,
    print_message_formats,
)
from tqdm.auto import tqdm

from data import load_nasdaq_itch
from utils.paths import display_path

# %% tags=["parameters"]
SKIP_PARSING = False
# Messages to read before stopping. None parses the whole session, which is what the
# shipped output is built from; a small number is how you try the parser out first.
MAX_MESSAGES = None

# %% [markdown]
# The parse reads one directory and writes another. `load_nasdaq_itch(get_base_path=True)`
# resolves the parsed-message root from the repository's data configuration rather than a
# path typed here, so the same notebook runs against a local checkout and inside the Docker
# image. The raw binary sits in a `raw/` directory beside it, which is where the download
# script puts it.
#
# `must_exist=False` is what makes a clean start work. For every other notebook in this
# chapter an absent `messages/` directory is the download instruction and the loader is
# right to refuse; this notebook is the one that creates it, and it has to be told where
# before there is anything there. The directory itself is created in Section 4, where the
# parse is about to write: creating it here would leave an empty one behind on a machine
# with no raw feed, and an empty directory is exactly what stops the loader refusing.

# %%
MESSAGE_DIR = load_nasdaq_itch(get_base_path=True, must_exist=False)
ITCH_RAW_DIR = MESSAGE_DIR.parent / "raw"

print(f"Raw ITCH binary (input):  {display_path(ITCH_RAW_DIR)}")
print(f"Parsed messages (output): {display_path(MESSAGE_DIR)}")
raw_present = ITCH_RAW_DIR.exists() and next(ITCH_RAW_DIR.iterdir(), None) is not None
print(f"Raw binary present: {raw_present}")
if not raw_present:
    print(
        "  Fetch it first: "
        "uv run python data/equities/market/microstructure/nasdaq_itch_download.py"
    )


def message_type_dirs() -> list[Path]:
    """Parsed message-type subdirectories, one uppercase letter each.

    Empty before Section 4 has run: `messages/` itself does not exist on a machine that
    has never parsed, so every cell that lists message types goes through here.
    """
    if not MESSAGE_DIR.exists():
        return []
    return [
        d
        for d in sorted(MESSAGE_DIR.iterdir())
        if d.is_dir() and len(d.name) == 1 and d.name.isupper()
    ]


# %% [markdown]
# ## 1. Message Specifications
#
# Each ITCH message has a fixed binary structure. The format is defined in `itch_message_specs.py`
# using Python's `struct` module. Format codes:
# - `H` = unsigned short (2 bytes)
# - `I` = unsigned int (4 bytes)
# - `Q` = unsigned long long (8 bytes)
# - `s` = char (1 byte), `Ns` = N chars
# - `>` = big-endian byte order

# %%
# Show message formats (loaded from utils/itch_message_specs.py)
print_message_formats()

# %% [markdown]
# ## 2. Binary Parsing Example
#
# Let's demonstrate how binary parsing works by creating and parsing a sample Add Order message.

# %% [markdown]
# Packing a message by hand is the quickest way to see the layout. An Add Order carries the
# format `>HH6sQsI8sI`: big-endian, two unsigned shorts, a six-byte timestamp, an eight-byte
# order reference, a one-character side, the share count, an eight-character padded ticker
# and the price. Unpacking it below returns exactly these fields.

# %%
sample_add_order = struct.pack(
    ">HH6sQsI8sI",
    1234,  # stock_locate
    5678,  # tracking_number
    b"\x00\x00\x00\x00\x00\x01",  # timestamp (1 nanosecond)
    9876543210,  # order_reference_number
    b"B",  # buy_sell_indicator
    100,  # shares
    b"AAPL    ",  # stock (padded to 8 chars)
    1500000,  # price: four implied decimals, so this is $150
)

print(f"Raw Add Order message ({len(sample_add_order)} bytes):")
print(f"  Hex: {sample_add_order.hex()}")

# %%
# Parse the binary data using our struct format
parsed = struct.unpack(FMT_DICT["A"], sample_add_order)
add_order = NT_DICT["A"]._make(parsed)

print("Parsed Add Order Message:")
print("-" * 40)
for field, value in add_order._asdict().items():
    if isinstance(value, bytes):
        value = value.decode("ascii").strip()
    print(f"  {field:25}: {value}")

# %%
# Apply conversions using helper functions from utils.itch_message_specs
ts_ns = parse_timestamp(add_order.timestamp)
price = parse_price4(add_order.price)

print(f"Timestamp: {ts_ns:,} nanoseconds = {ts_ns / 1e9:.9f} seconds after midnight")
print(f"Price: ${price:.4f}")

# %% [markdown]
# ## 3. What the Pipeline Holds Right Now
#
# Before parsing anything, take stock: the raw binary is the input, the per-message-type
# Parquet directories are the output, and either can be absent. On a first run the parsed
# side is empty, and it stays empty until Section 4 writes it. The parsed messages are
# loaded in Section 5, after there is something to load.

# %%
# Check what data is available locally
print("ITCH Data Pipeline Status:")
print("-" * 50)

# Step 1: Raw binary from download
raw_files = []
if ITCH_RAW_DIR.exists():
    raw_files = list(ITCH_RAW_DIR.glob("*.gz")) + list(ITCH_RAW_DIR.glob("*.bin"))
print(f"Raw binary files: {len(raw_files)}")
for f in raw_files:
    print(f"  {f.name} ({f.stat().st_size / 1e9:.2f} GB)")

# Step 2: Parsed messages (single uppercase letter = message type)
parsed_types = message_type_dirs()
parsed_with_data = [d for d in parsed_types if list(d.glob("*.parquet"))]
print(f"Parsed message types: {len(parsed_with_data)}")
for msg_dir in parsed_with_data:
    name = MESSAGE_SPECS.get(msg_dir.name, {}).get("name", "Unknown")
    n_files = len(list(msg_dir.glob("*.parquet")))
    print(f"  {msg_dir.name} ({name}): {n_files} files")

# %% [markdown]
# ## 4. Full Parser Implementation
#
# The parser below is written to be read. It processes one message at a time in Python, which
# is what makes each step visible and also what makes a full trading day take about twenty
# minutes. Section 6 covers the Rust parser, which emits the same Parquet schema and is what
# you would run over many days.


# %% [markdown]
# ### Parser Helpers
#
# The parse splits into three functions. `read_frame` takes the next message off the file
# and hands back its type and its raw bytes; `decode_message` turns those bytes into a
# dictionary of Python values; and `parse_itch_file` runs the loop, buffering decoded
# messages and writing them out in batches.


# %%
def read_frame(f, pbar) -> tuple[str, bytes] | None:
    """Read one ITCH message frame: 2-byte length + 1-byte type + payload.

    Returns (msg_type, payload) on success, or None on EOF/truncation.
    """
    # 2-byte big-endian length prefix (message size including type byte)
    length_bytes = f.read(2)
    if len(length_bytes) < 2:
        return None
    pbar.update(2)

    msg_size = int.from_bytes(length_bytes, "big")

    # 1-byte message type
    msg_type_byte = f.read(1)
    if len(msg_type_byte) < 1:
        print(f"\nWarning: Truncated message at byte {f.tell()}, expected type byte")
        return None
    pbar.update(1)

    msg_type = msg_type_byte.decode("ascii")

    # Payload (msg_size includes type byte, so payload is msg_size - 1)
    payload = f.read(msg_size - 1)
    if len(payload) < msg_size - 1:
        print(f"\nWarning: Truncated payload for message type {msg_type}")
        return None
    pbar.update(msg_size - 1)

    return msg_type, payload


# %% [markdown]
# Decode binary payload into a Python dict, converting raw timestamp bytes
# to nanosecond integers and byte strings to stripped ASCII.


# %%
def decode_message(msg_type: str, payload: bytes) -> dict | None:
    """Unpack binary payload into a dict, converting timestamps and strings.

    Returns parsed message dict, or None on struct error.
    """
    try:
        parsed = struct.unpack(FMT_DICT[msg_type], payload)
        msg = NT_DICT[msg_type]._make(parsed)._asdict()
    except struct.error:
        return None

    # Convert timestamp: nanoseconds since midnight
    if "timestamp" in msg:
        msg["timestamp"] = int.from_bytes(msg["timestamp"], "big")

    # Decode string fields
    for field, value in msg.items():
        if isinstance(value, bytes):
            msg[field] = value.decode("ascii").strip()

    return msg


# %% [markdown]
# The main parser reads the binary file sequentially, buffering decoded messages
# and flushing to Parquet periodically to bound memory usage.


# %%
def parse_itch_file(
    itch_file: Path,
    trading_day: date,
    output_dir: Path,
    max_buffered_messages: int = 10_000_000,
    max_messages: int | None = None,
) -> dict[str, int]:
    """Parse ITCH binary file and store messages as Parquet.

    Args:
        itch_file: Path to binary ITCH file (.bin, not .gz).
        trading_day: Trading date for timestamp construction.
        output_dir: Directory for Parquet output (one subdir per message type).
        max_buffered_messages: Flush threshold (total buffered messages).
        max_messages: Optional limit for testing.

    Returns:
        Dictionary with message type counts.
    """
    # Midnight timestamp for the trading day (ITCH timestamps are nanoseconds offset)
    base_ts = datetime(trading_day.year, trading_day.month, trading_day.day)
    file_counters: dict[str, int] = defaultdict(int)

    buffers = defaultdict(list)
    counts = Counter()
    file_size = itch_file.stat().st_size
    start_time = time()

    with (
        itch_file.open("rb") as f,
        tqdm(total=file_size, desc="Parsing ITCH", unit="B", unit_scale=True) as pbar,
    ):
        while True:
            if max_messages and sum(counts.values()) >= max_messages:
                print(f"\nLimit reached: {max_messages:,} messages")
                break

            frame = read_frame(f, pbar)
            if frame is None:
                break
            msg_type, payload = frame
            counts[msg_type] += 1

            if msg_type not in FMT_DICT:
                continue

            msg = decode_message(msg_type, payload)
            if msg is None:
                continue

            # Check for end of messages
            if msg_type == "S" and msg.get("event_code") == "C":
                print("\nEnd of Messages")
                flush_to_parquet(buffers, output_dir, base_ts, file_counters)
                break

            buffers[msg_type].append(msg)

            # Periodic flush
            if sum(len(v) for v in buffers.values()) >= max_buffered_messages:
                flush_to_parquet(buffers, output_dir, base_ts, file_counters)

    # Final flush
    if any(buffers.values()):
        flush_to_parquet(buffers, output_dir, base_ts, file_counters)

    elapsed = time() - start_time
    total = sum(counts.values())
    print(f"Parsed {total:,} messages in {elapsed:.1f}s ({total / elapsed:,.0f} msg/s)")

    return dict(counts)


# %%
# Locate ITCH data file (skip if SKIP_PARSING is set)
if SKIP_PARSING:
    print("SKIP_PARSING=True: skipping ITCH binary parsing (uses pre-parsed data)")
    itch_file = None
    counts = {}
else:
    # Clear any existing parsed data to avoid schema conflicts
    # (Different parser versions may produce different schemas)
    # Set ITCH_KEEP_EXISTING=1 to skip cleanup and use existing data
    clear_existing = os.environ.get("ITCH_KEEP_EXISTING", "0") != "1"
    if clear_existing and MESSAGE_DIR.exists() and list(MESSAGE_DIR.glob("*/part-*.parquet")):
        print(f"Clearing existing parsed data in {MESSAGE_DIR}")
        print("  (Set ITCH_KEEP_EXISTING=1 to keep existing data)")
        shutil.rmtree(MESSAGE_DIR)
        MESSAGE_DIR.mkdir(parents=True, exist_ok=True)

    # Find ITCH file (compressed or uncompressed)
    gz_files = list(ITCH_RAW_DIR.glob("*.gz")) if ITCH_RAW_DIR.exists() else []
    bin_files = list(ITCH_RAW_DIR.glob("*.bin")) if ITCH_RAW_DIR.exists() else []

    if not gz_files and not bin_files:
        raise FileNotFoundError(
            f"No raw ITCH binary found at {ITCH_RAW_DIR}.\n"
            "Download first:\n"
            "  uv run python data/equities/market/microstructure/nasdaq_itch_download.py"
        )

# %%
# Decompress if needed and extract trading date
if not SKIP_PARSING and (gz_files or bin_files):
    # Prefer uncompressed, otherwise decompress
    if bin_files:
        itch_file = bin_files[0]
        print(f"Found uncompressed: {itch_file.name}")
    else:
        gz_file = gz_files[0]
        itch_file = gz_file.with_suffix(".bin")

        if not itch_file.exists():
            print(f"Decompressing {gz_file.name}...")
            with gzip.open(gz_file, "rb") as f_in, open(itch_file, "wb") as f_out:
                shutil.copyfileobj(f_in, f_out)
            print(f"Created: {itch_file.name} ({itch_file.stat().st_size / 1e9:.1f} GB)")
        else:
            print(f"Found: {itch_file.name}")

    # Extract trading date from filename (format: MMDDYYYY.NASDAQ_ITCH50.bin)
    date_str = itch_file.stem.split(".")[0]
    if len(date_str) == 8 and date_str.isdigit():
        trading_day = date(int(date_str[4:8]), int(date_str[:2]), int(date_str[2:4]))
    else:
        trading_day = date(2020, 1, 30)  # Fallback

    print(f"Trading day: {trading_day}")

    # Parse ITCH file (full-day parse: ~22 min on the reference machine)
    if itch_file and itch_file.exists():
        # Created here, not where the path is resolved: an empty messages/ directory
        # left behind by a run that found no raw binary would stop load_nasdaq_itch
        # raising for every later notebook in the chapter.
        MESSAGE_DIR.mkdir(parents=True, exist_ok=True)
        counts = parse_itch_file(
            itch_file=itch_file,
            trading_day=trading_day,
            output_dir=MESSAGE_DIR,
            max_messages=MAX_MESSAGES,
        )

        print("\nMessage counts:")
        for msg_type, count in sorted(counts.items(), key=lambda x: -x[1]):
            name = MESSAGE_SPECS.get(msg_type, {}).get("name", "Unknown")
            print(f"  {msg_type} ({name:25}): {count:>10,}")

# %% [markdown]
# ## 5. Message Type Analysis
#
# The parse has written one Parquet directory per message type, so the store can now be
# read back. The cell below asserts on trade messages (`P`) because that is the type it
# goes on to load. A full session always contains them, so an empty `P` directory means
# the parse did not finish rather than that this day happened to have no trades. That
# reasoning holds only for a full parse: a run with `MAX_MESSAGES` set can stop before
# the first trade prints, so there the absence is reported and the remaining cells work
# with whichever types the trial did write.

# %%
trade_dir = MESSAGE_DIR / "P"
trade_files = sorted(trade_dir.glob("*.parquet")) if trade_dir.exists() else []

if not trade_files and MAX_MESSAGES is not None:
    print(
        f"No trade messages under {display_path(trade_dir)}: the parse stopped after "
        f"{MAX_MESSAGES:,} messages and a `P` need not appear that early. "
        "Set MAX_MESSAGES to None to read the whole session."
    )
else:
    assert trade_files, (
        f"No parsed trade messages at {trade_dir}.\n"
        "Section 4 writes them from the raw binary. If SKIP_PARSING is True, it was skipped\n"
        "and the store has to have been written by an earlier run or by the Rust parser of\n"
        "Section 6. If the raw binary is missing, fetch it first:\n"
        "  uv run python data/equities/market/microstructure/nasdaq_itch_download.py"
    )

    trades = pl.read_parquet(trade_dir / "*.parquet")
    print(f"Loaded {len(trades):,} trade messages")
    print(f"Columns: {trades.columns}")

# %%
# Message type distribution — use lazy scan to count without loading all data
written = [d for d in message_type_dirs() if list(d.glob("*.parquet"))]
print("Parsed Message Types:")
print("-" * 50)
if not written:
    print("  Nothing parsed yet - Section 4 writes these from the raw binary.")
for msg_dir in written:
    count = pl.scan_parquet(msg_dir / "*.parquet").select(pl.len()).collect().item()
    name = MESSAGE_SPECS.get(msg_dir.name, {}).get("name", "Unknown")
    print(f"  {msg_dir.name} ({name:25}): {count:>12,} messages")

# %%
# Schema compatibility check — verify we can read each message type
written = [d for d in message_type_dirs() if list(d.glob("*.parquet"))]
print("Schema Compatibility Check:")
print("-" * 50)
if not written:
    print("  Nothing parsed yet - Section 4 writes these from the raw binary.")
for msg_dir in written:
    try:
        sample = pl.scan_parquet(msg_dir / "*.parquet").head(5).collect()
        name = MESSAGE_SPECS.get(msg_dir.name, {}).get("name", "Unknown")
        print(f"  [OK] {msg_dir.name} ({name:25}): cols={list(sample.columns)[:4]}...")
    except Exception as e:
        print(f"  [FAIL] {msg_dir.name}: {e}")

# %% [markdown]
# ## 6. Production Parsing with Rust
#
# The Python parser above decodes one message per loop iteration, and a trading day holds a
# few hundred million of them. The same protocol parsed in Rust reads the file through a
# memory map and unpacks each message without copying it first, so it neither pays the
# per-message interpreter overhead nor holds the decoded messages in memory.
#
# **Repository**: [github.com/ml4t/itch-parser](https://github.com/ml4t/itch-parser)
#
# The figures in the book's Table 3.1, for the same 13 GB file on one machine, are about
# twenty-three minutes and roughly 8 GB of memory in Python against under five minutes and
# under 500 MB in Rust. Wall-clock timings move with disk and CPU, so read them as an order
# of magnitude rather than a ratio: the gap is large enough that it decides which parser you
# reach for, and not stable enough to quote to a decimal place. The throughput this notebook
# printed above is the Python side of the same comparison, measured on the machine that ran
# it.
#
# ### Installation
#
# ```bash
# # Clone the repository
# git clone https://github.com/ml4t/itch-parser.git
# cd itch-parser
#
# # Build release binary
# cargo build --release
# ```
#
# ### Usage
#
# ```bash
# # Parse ITCH file (works with .gz or uncompressed)
# ./target/release/itch_parser <input_file> <output_dir> <MMDDYYYY>
#
# # Example
# ./target/release/itch_parser data/01302020.NASDAQ_ITCH50.gz ./messages 01302020
# ```
#
# Output is identical Parquet files partitioned by message type, compatible with
# the Python code in this notebook and downstream analysis.
#
# ### When to Use Which
#
# | Use Case | Recommendation |
# |----------|---------------|
# | Learning the protocol | Python (this notebook) |
# | Debugging parse issues | Python |
# | Processing a single day | Either |
# | Multi-day backtesting | **Rust** |
# | Production pipeline | **Rust** |
#
# Both parsers write the same Parquet schema, so a day parsed either way feeds every
# notebook that follows without change.

# %% [markdown]
# ## Key Takeaways
#
# 1. **The protocol is message-by-order.** Every message that changes the book names the
#    individual order it acts on, stamped to the nanosecond, which is what makes book
#    reconstruction possible at all. Session-level messages - `S` for market events, `R`
#    for the stock directory, `Q` for the auction crosses - carry no order reference and
#    describe the venue rather than a single order.
# 2. **Seven message types carry the book.** `A` and `F` add an order, `E` and `C` execute
#    against one, `X` cancels part of one, `D` deletes one, `U` replaces one.
# 3. **Numbers arrive encoded.** Prices are integers with four implied decimal places, and
#    timestamps are nanoseconds since midnight, so both need converting before use.
# 4. **Parse in batches, not in one pass.** Buffering by message type and flushing to
#    Parquet is what keeps a day that does not fit in memory from having to.
# 5. **Which parser depends on how many days you need.** Python reads clearly and takes
#    about twenty minutes per day; Rust emits the same schema in a fraction of that.
#
# ### Known limitations
#
# - The parse covers one venue. NASDAQ ITCH sees NASDAQ-routed activity, not the
#   consolidated tape, so counts here are a venue's share of a symbol's trading rather than
#   all of it.
# - Message types outside `FMT_DICT` are counted and skipped rather than decoded.
# - The Rust timings quoted above were measured elsewhere, on one machine; this notebook
#   times only its own Python parse.
#
# ### Next Steps
#
# - **Order Book Reconstruction**: `02_itch_lob_reconstruction`
# - **Trading Activity Overview**: `05_itch_trading_activity` (includes E/C enrichment)
#
# ---
#
# ## Reference
#
# Bouchaud, J.-P., Bonart, J., Donier, J., & Gould, M. (2018).
# *Trades, Quotes and Prices: Financial Markets Under the Microscope*.
# Cambridge University Press.
# [https://doi.org/10.1017/9781009028943](https://doi.org/10.1017/9781009028943)

```