dynalog for Python

dynalog is a structured logging library for Python. It provides colored console logs, named events, contextual fields, JSON and logfmt output, filtering, redaction, enrichment, custom highlights, and test-friendly clocks through one default dynalog instance.

[D7MFXJ01RE3W4 info] Server started port=3000 service=api

Console records use compact, sortable LSB32 timestamps. File and memory records retain the original nanosecond timestamp and contain no ANSI color sequences.

Features

• Pythonic: Module-level logging functions and configurable Dynalog instances.
• Structured: Immutable events with named JSON-compatible payloads.
• Contextual: Shared fields with per-call overrides.
• Readable: Level colors, semantic indicators, and custom highlights.
• Flexible: Plain text, JSON, and compact logfmt output.
• Composable: Console, file, memory, fan-out, and custom destinations.
• Controlled: Minimum-level filtering, recursive redaction, and enrichment.
• Precise: Nanosecond event timestamps and 13-character LSB32 display IDs.
• Testable: Injectable clocks and in-memory capture.

Installation

pip install dynalog

Dynalog requires Python 3.10 or newer.

Quick Start

import dynalog
from dynalog import Level

dynalog.setup(
    level=Level.INFO,
    context={"service": "api"},
)

dynalog.info("Server started", context={"port": 3000})
dynalog.event("user.login", {"id": 42, "username": "Alice"})

The default destination is Console. Call setup once during application startup to select a minimum level, output format, shared context, destinations, or custom clock.

Logging Levels

dynalog.trace("Resolving configuration")
dynalog.debug("Cache connected")
dynalog.info("Server started")
dynalog.warn("Queue depth is elevated", context={"depth": 82})
dynalog.error("Request failed", context={"status": 503})
dynalog.fatal("Worker stopped")

Set a minimum level to discard lower-priority records:

from dynalog import Level

dynalog.setup(level=Level.WARN)

Use allows before computing expensive diagnostic fields:

if dynalog.allows(Level.DEBUG):
    dynalog.debug("Cache snapshot", context=collect_cache_details())

Structured Events

Use named events for records intended for search, analytics, metrics, alerts, or automated processing.

dynalog.event(
    "payment.completed",
    {
        "order": "ORD-1042",
        "amount": 129.95,
        "currency": "USD",
    },
)

Messages and structured events use the same timestamp, filtering, context, formatting, and destination configuration.

Events and Sources

Construct an event directly when a caller needs explicit timestamps, context, or source information:

from dynalog import Event, Level, Source

event = Event.message(
    "Server started",
    timestamp=1_767_225_600_123_456_789,
    level=Level.INFO,
    context={"port": 3000},
    source=Source("app.py", 12, "app"),
)

dynalog.write(event)

Event values are immutable. Transforming an event returns a replacement instead of changing the original value.

Context

Shared context is merged into every new log record. Per-call context overrides matching shared values without changing later events.

dynalog.setup(
    context={"service": "checkout", "region": "west"},
)

dynalog.info(
    "Order accepted",
    context={"request": "req-42", "region": "central"},
)

Context values can be strings, integers, floats, booleans, or nested mappings.

Output Formats

Format	Use case
Output.PLAIN	Human-readable terminal output
Output.JSON	Files, collectors, and structured ingestion
Output.COMPACT	Dense single-line logfmt output

from dynalog import Output

dynalog.setup(output=Output.JSON)

Format an event without writing it to a destination:

from dynalog import Event, Level

event = Event.message(
    "Cache connected",
    level=Level.INFO,
    context={"host": "cache.internal"},
)

plain = dynalog.text(event, "plain")
json = dynalog.text(event, "json")
compact = dynalog.text(event, "compact")

format returns bytes; text returns a decoded string.

Destinations

Send the same record to multiple destinations by listing them in sinks. Every destination is attempted, and failures are reported together after dispatch.

import dynalog
from dynalog import Console, File, Memory, Output

memory = Memory()

dynalog.setup(
    output=Output.JSON,
    sinks=[Console(), File("app.jsonl"), memory],
)

dynalog.info("Ready", context={"port": 3000})

assert len(memory.events) == 1
assert len(memory.records) == 1

Memory.clear() removes captured events and records.

Custom Destinations

A custom destination implements write(record, event):

from dynalog import Event

class Queue:
    def __init__(self) -> None:
        self.records: list[tuple[bytes, Event]] = []

    def write(self, record: bytes, event: Event) -> None:
        self.records.append((bytes(record), event))

dynalog.setup(sinks=[Queue()])

Redaction and Enrichment

Redaction returns a replacement event and leaves the original unchanged. Matching keys are masked recursively.

from dynalog import Event, Level

event = Event.message(
    "Authentication failed",
    timestamp=1_767_225_600_123_456_789,
    level=Level.ERROR,
    context={
        "user": "alice",
        "token": "secret",
        "credentials": {"password": "secret"},
    },
)

safe = dynalog.redact(event, ["password", "token"])
enriched = dynalog.enrich(safe, "api-01")
dynalog.write(enriched)

The default redaction keys are password and token. When no process ID is provided, enrich uses the current Python process.

Console Highlights

Built-in rules emphasize failures, warnings, HTTP status classes, booleans, identifiers, timing values, paths, hosts, and ports. Add literal or compiled regular expression rules for application-specific indicators.

import re
import dynalog
from dynalog import Console, Tone

console = (
    Console(color=True)
    .highlight(re.compile(r"\bAUTH-\d+\b"), Tone.MAGENTA)
    .highlight("payment declined", Tone.RED)
)

dynalog.setup(sinks=[console])
dynalog.error("Payment AUTH-42 failed", context={"status": 503})

Available tones are RED, YELLOW, GREEN, CYAN, MAGENTA, BLUE, WHITE, and GRAY. The first custom rule wins when matches overlap.

Color is selected automatically for a compatible terminal and disabled when NO_COLOR is present. Pass True or False to Console to override automatic detection.

Testable Time

Provide a clock function that returns nanoseconds as an integer:

import dynalog
from dynalog import Memory

memory = Memory()
now = 1_000_000_000

def clock() -> int:
    return now

dynalog.setup(sinks=[memory], clock=clock)
dynalog.info("first")
now += 1_000_000_000
dynalog.info("second")

delta = memory.events[1].timestamp - memory.events[0].timestamp
assert delta == 1_000_000_000

The default clock is time.time_ns.

Public API

Export	Purpose
dynalog	Module-level default logger functions
Dynalog	Independent logger instance
Event	Immutable message or structured event
Source	File, line, and module information
Level	Six ordered logging levels
Output	Plain, JSON, or compact formatting
Tone	Named console highlight color
Console	Colored terminal destination
File	Append-only file destination
Memory	In-memory events and records
Sink	Custom destination protocol

Module-level functions include setup, format, text, allows, redact, enrich, write, event, trace, debug, info, warn, error, fatal, and version.

Requirements

• Python 3.10 or newer.
• A wheel compatible with the target Python version and platform, or a Rust toolchain for source builds.
• A supported terminal for automatic ANSI colors.

Publishing

Upload verified distributions to PyPI:

python -m build
python -m twine check dist/*
python -m twine upload dist/*

PyPI authentication can use an API token through TWINE_USERNAME=__token__ and TWINE_PASSWORD.

The distribution includes the Python API, native extension, runtime artifact, documentation, and both license texts.

Limitations and Compatibility

• High-level log methods omit source fields unless source metadata is supplied explicitly.
• File output is append-only and does not currently rotate.
• Destination writes are synchronous; custom destinations should remain fast.
• Nanosecond timestamps do not guarantee uniqueness when events share an instant.
• LSB32 display IDs require every producer and consumer to use the same epoch.
• Building from source requires a supported Rust toolchain and native compiler.

License

Dynalog is dual-licensed under either of the following, at your option:

• GNU Lesser General Public License v3.0 only
• GNU General Public License v3.0 only

SPDX: LGPL-3.0-only OR GPL-3.0-only