exonware-xwsystem 0.9.0.27

Enterprise-grade Python framework with AI-powered performance optimization, 24 serialization formats, military-grade security, automatic memory leak prevention, circuit breakers, and production monitoring - replaces 50+ dependencies

xwsystem

One install instead of 50+. Serialization (24+ formats), caching, security, validation, HTTP, IPC, monitoring - same APIs everywhere. The base every other eXonware package builds on.

Company: eXonware.com · Author: eXonware Backend Team · Email: [e-mail]

---

Install

Requires Python 3.12+. (See pyproject.toml requires-python.)

Install	What you get	When to use
pip install exonware-xwsystem	Lite - core only, zero optional deps	Minimal footprint, or you install what you need.
pip install exonware-xwsystem[lazy]	Lazy - core + xwlazy; missing deps install on first import	Development; optional formats without pre-installing everything.
pip install exonware-xwsystem[full]	Full - core + common optionals pre-installed	Production or CI when you want all formats and features up front.

Same package; [lazy] and [full] are extras. You can also install as xwsystem (same thing).

---

Quick start

Serialization (same API for every format):

from exonware.xwsystem import JsonSerializer

s = JsonSerializer()
text = s.dumps({"name": "Alice", "age": 30})
data = s.loads(text)

Lazy install: With [lazy], use normal imports; if a format backend is missing, xwlazy installs it on first use. No try/except import blocks.

Windows console UTF-8:

from exonware.xwsystem.console.cli import ensure_utf8_console
ensure_utf8_console()

Config: xwconfig.toml at project root; in code: from exonware.xwsystem import settings then settings.current().

---

What you get

Area	What’s in it
One dependency	Replaces dozens of imports (json, yaml, msgpack, crypto, requests, path validation, caching, …) with one API surface.
24+ serialization formats	Text (JSON, YAML, TOML, XML, CSV, INI, properties, .env), binary (MsgPack, BSON, CBOR, Pickle), database (Sqlite3, Dbm, Shelve), tabular (DataFrame, Excel, CSV dialects). Enterprise/scientific (~10 formats: Avro, Protobuf, Parquet, ORC, Arrow, etc.) integrate via xwformats.
20+ caching variants	LRU, LFU, TTL (sync + async), read-through/write-through/write-behind, two-tier (memory+disk), secure/encrypted caches, observable caches, Prometheus export.
Security	Path validation, file security, audit helpers, crypto utilities, validators, policies.
Validation	Pydantic-style XModel, schema discovery.
Runtime	Config (performance modes, logging), console (CLI, colors, progress, tables), monitoring (memory, performance, tracing), IPC (process fabric, queues, pipes, shared memory), HTTP client (HTTP/2, retries, streaming).
Patterns	Circuit breakers, retry, object pool, context managers. Plus: operations (diff, merge, patch), structures (tree walker, circular ref detection), threading (locks, thread-safe factory), utils (datetime, paths, string, web), data structures (trie, union-find), grammar/syntax (50+ languages, Monaco), plugins, query registry.
Indexing	Pluggable index backends (io/indexing): InMemory, FileBacked, BTreeIndexBackend (ordered, range_scan), FullTextIndexBackend and FileBackedFullTextIndexBackend (term index, persistent). See docs/REF_15_API.md (Indexing module).

Lite = zero optional deps. Lazy = xwlazy installs format backends on first use. Full = common optionals pre-installed.

---

Why xwsystem? (Highlights)

Automatic format detection (no guessing) - One API for any file. AutoSerializer and detect_format pick the right codec from file extension, magic bytes, or content patterns (confidence-scored). No need to know whether it’s JSON, YAML, or MessagePack; pass a path or bytes and get the right serializer.

from exonware.xwsystem import AutoSerializer, detect_format
from pathlib import Path

# By path: extension tells us the format
fmt = detect_format(Path("data.json"))   # → JSON

# One serializer for everything: format from path (extension)
auto = AutoSerializer()
data = auto.load_file(Path("config.yaml"))
auto.save_file(data, Path("out.toml"))

JSON: 3x+ faster than stdlib - Pluggable parsers: orjson (3-4x faster), optional hybrid (msgspec for read, orjson for write). Measured: 1MB serialize <45ms, deserialize <52ms (see REF_54_BENCH).

Caching: O(1), benchmark-backed - LRU/LFU/TTL with sub-microsecond get/set (10K items). Default uses the fastest Python cache baseline (~33k GET ops/sec, ~3.3k MIXED in our benchmarks). SLAs and NFRs in REF_54_BENCH.

Serialization benchmark (1MB data) - From REF_15_API / REF_54_BENCH:

Format	Serialize	Deserialize
MessagePack	23 ms	28 ms
JSON	45 ms	52 ms
YAML	123 ms	145 ms

Benchmarks - Campaigns live in benchmarks/ (per GUIDE_54_BENCH). Run from project root; full index: benchmarks/INDEX.md.

Benchmarks summary (10-Feb-2026) - Condensed from latest runs. All campaigns: JSON, caching, serialization, atomic I/O, operations, data structures, object pool, validation, threading locks, async fabric.

Area	Best / representative result
JSON	JsonSerializer (hybrid/hyperjson): small ~798k decode / 506k encode ops/s; with hyperjson installed matches hyperjson. orjson/msgspec in same tier.
Caching	PylruCache (default): get 2.99M/s, put 2.83M/s, mixed 2.49M/s. CacheboxCache get 3.72M/s. TwoTierCache now included (put/get contract).
Serialization	MsgPackSerializer small: 862k decode / 301k encode. YamlSerializer medium: 15 decode / 39 encode. PickleSerializer small: 422k decode / 357k encode. Competitive with msgpack/PyYAML/stdlib pickle.
Atomic I/O	AtomicFileWriter: 205-357 writes/s (1KB-100KB). Plain write: 504-743 writes/s. Atomic adds ~2-3x latency for crash-safe commit.
Operations	diff (structural/content/full) small: 35k-48k ops/s. merge (deep/shallow/overwrite) small: 59k-70k ops/s. patch_apply small: 64k ops/s.
Data structures	TrieNode lookup 899k ops/s; dict lookup 26M ops/s. UnionFind (make_set+union+find) 573 runs/s. trie insert batch 49k ops/s.
Object pool	ObjectPool get+release (thread_safe=True) 350k ops/s; (thread_safe=False) 287k ops/s. direct SimpleObj() 2.93M ops/s. Pool overhead ~8x for reuse.
Validation	check_data_depth shallow 1.11M ops/s, deep 229k ops/s. validate_path_input (safe) 2.58M ops/s.
Locks	EnhancedRLock (track_stats=True) 1.53M ops/s; (track_stats=False) 3.37M ops/s. threading.RLock 7.14M ops/s. Use track_stats=False for fast path.
Async fabric	Submit 1k tasks ~3.27k ops/s; queue latency ~0.02 ms; shared memory reuse read ~0.006 ms (see benchmarks/20260210-benchmark xwsystem async fabric).

Using these results: create_cache() already defaults to PylruCache when pylru is installed (~2.5-3M mixed ops/s). For hot-path locking without timeout or stats, use fast_lock() or EnhancedRLock(track_stats=False) (~3.4M ops/s). Use TwoTierCache when you need memory+disk; use ObjectPool when object creation cost outweighs pool overhead. See REF_54_BENCH trade-offs table.

---

Rust and the rest of the stack

Rust: xwsystem is the reference implementation. Multi-language (TypeScript, Rust, Go) is planned via contracts/specs. Right now Python performance is comparable to what we would expect from a Rust port on the hot paths, so no conversion yet. Getting here took several rewrites; the current design is the one we kept.

Ecosystem: xwsystem is the foundation for 12+ eXonware projects (xwstorage, xwformats, xwjson, xwnode, xwdata, xwauth, xwquery, xwchat, xwui, *-server). They get one dependency instead of 50+, same APIs everywhere, and Lite/Lazy/Full so each package can ship lean or full. One place to fix bugs and add features for the whole stack.

---

Full feature list and examples

For a complete feature tour, code samples, and platform notes see README_LONG.md.

---

Docs and tests

Content in this README is aligned with the project REFs and docs/GUIDE_01_USAGE.md (per GUIDE_63_README).

• Start: docs/INDEX.md - doc index and quick links.
• Use it: docs/GUIDE_01_USAGE.md - installation, codecs, caching, runtime, production.
• Requirements and status: docs/REF_01_REQ.md, docs/REF_22_PROJECT.md.
• API and design: docs/REF_15_API.md, docs/REF_13_ARCH.md.
• DX and quality: docs/REF_14_DX.md, docs/REF_50_QA.md, docs/REF_54_BENCH.md, docs/REF_51_TEST.md.
• Ideas and planning: docs/REF_12_IDEA.md, docs/REF_21_PLAN.md.
• Compliance: docs/compliance/. Evidence: docs/logs/ (changes, tests, benchmarks, plans).
• Benchmarks: benchmarks/INDEX.md - run scripts in benchmarks/<campaign>/scripts/ (JSON, caching, serialization, atomic I/O, operations, data structures, object pool, validation, locks). See REF_54_BENCH.

Tests: 4-layer suite (0.core, 1.unit, 2.integration, 3.advance). Run via project test runner or pytest. See docs/REF_51_TEST.md.

---

Where xwsystem fits

xwsystem provides the shared infrastructure layer for other eXonware packages (xwstorage, xwformats, xwjson, xwnode, xwdata, xwauth, xwquery, xwchat, xwui, *-server). They depend on its serializers, caches, security helpers, IPC layer, and runtime services instead of re-implementing them.

Approximate surface area (as of 2026-02-07, see REFs and benchmarks for exact lists):

• Formats: 24+ core serialization formats in this package, plus ~10 enterprise/scientific formats via xwformats.
• Caching: 20+ cache variants built from core implementations and wrappers (LRU/LFU/TTL, two-tier, observable, secure, async, read-through/write-through/write-behind).
• Grammar/syntax: 50+ language grammars with Monaco integration (see grammar/ and REF_13_ARCH).
• IPC: 4+ IPC primitives (process pool, message queue, shared memory, async fabric facade) wired into a single Async Process Fabric.
• Benchmarks: 10+ benchmark campaigns (JSON, caching, serialization, atomic I/O, operations, data structures, object pool, validation, locks, async fabric) tracked in benchmarks/ and REF_54_BENCH.

Downstream libraries consume these services via stable APIs exported from exonware.xwsystem, so fixes and performance improvements in xwsystem flow through the rest of the ecosystem.

---

License and links

MIT - see LICENSE.

• Homepage: https://exonware.com
• Repository: https://github.com/exonware/xwsystem

Part of the eXonware ecosystem - one foundation for all of it. Version: 0.9.0.27 | Updated: 30-Mar-2026

Built with ❤️ by eXonware.com - Revolutionizing Python Development Since 2025