ries 2.0.2

Find algebraic equations given their solution - Rust implementation
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151

# Changelog

All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [Unreleased]

## [2.0.2] - 2026-06-27

### Fixed
- Turbo now mirrors serial generation semantics before parallel matching:
  small batch searches use the same LHS deduplication as serial, while
  streaming-sized searches preserve the raw LHS stream. This fixes a
  level-3 rank-one parity miss where `--turbo` could discard the serial best
  match for `2.506314` before matching.
- Added regressions covering both turbo's small batch-generation parity path
  and the expensive level-3 flat-exact `2.506314` case.

### Changed
- Documented the refined turbo contract: canonical generation semantics plus
  parallel match/Newton work, with the lower-ranked tail still allowed to differ
  from serial.

## [2.0.1] - 2026-06-27

### Fixed
- Turbo and parallel-generation searches now fall back to the canonical serial
  pipeline for order-dependent `--stop-at-exact` and `--stop-below` searches,
  preventing per-worker early exits from violating rank-one parity
- Turbo CLI results are canonically sorted before display so the documented
  best match is always presented at rank one

## [2.0.0] - 2026-06-26

### Added
- Turbo mode (`--turbo`) parallelizes matching and Newton refinement across
  available cores while preserving the same single best match as serial search
- Shared `MatchSummary` conversion for consistent Rust, Python, and WASM match
  representations
- Widest candidate-window diagnostics in JSON output and search stats, recording the
  LHS expression that produced the maximum RHS scan window
- Level-3 benchmark baseline artifacts dated 2026-03-20
- Supported Rust-side and end-to-end Python binding test workflows using `uv`

### Changed
- **Breaking:** `SearchStats::record_candidate_window` now takes the LHS
  `EvaluatedExpr` plus window width (was width only) so widest-window diagnostics
  can be attributed to a specific expression
- Python wheels use the CPython stable ABI for Python 3.8 and newer instead of
  being tied to the single interpreter version used to build them
- Adaptive search radius is capped by the pool strict-gate coarse-error envelope,
  matching `would_accept_strict` value-space bounds; batch and streaming paths now
  share the same radius logic
- Parallel generation OOM preflight uses count-only streaming instead of cloning
  full expression sets, reducing peak RSS during large runs
- `sinpi`/`cospi`/`tanpi` at default π scale snap arguments within `1e-12` of
  integers or half-integers to exact values and derivatives, removing pathological
  near-singular artifacts (e.g. `1Sxn^S`) but also shifting which equations rank
  highly; refreshed level-3 baselines show fewer Newton calls and a lower Newton
  success rate (~92% → ~73%) alongside far fewer candidate pairs scanned
  (~66M → ~8M)
- Optional widest-window JSON fields are omitted when no window has been recorded
  (`#[serde(skip_serializing_if = "Option::is_none")]`)
- Python and WASM APIs reject non-finite targets consistently and document their
  bounded level scale separately from the CLI complexity scale

### Fixed
- PSLQ relation detection, duplicate/scalar relation handling, and rational
  fast-match classification
- Adaptive target-count exponent overflow and profile parsing/merge error handling
- Adaptive-growth behavior and classic streaming result ordering
- Expression deduplication during generation no longer clones when replacing a
  duplicate with a simpler expression in the same quantized bucket
- Trig argument snapping guards `f64`→`i64` conversion against out-of-range values

## [1.1.1] - 2026-03-18

### Added
- A direct live-demo link in the README and on the GitHub repository homepage

### Changed
- Refreshed GitHub Actions workflow dependencies to current Node 24-compatible
  releases to remove deprecation warnings from CI and release automation
- Updated public-facing README copy to make the landing page and standalone demo
  entrypoints explicit

## [1.1.0] - 2026-03-18

### Added
- Automated release-integrity checks for cross-package version sync, canonical
  homepage metadata, release notes presence, and README image assets
- Public-facing v1.1.0 release notes centered on the canonical landing page and
  live demo handoff

### Changed
- Pointed package and citation homepage metadata at the canonical project page
  on `maxwellsantoro.com`
- Clarified release ownership in `RELEASING.md`: this repo verifies release
  artifacts, while the website repo verifies the deployed demo experience
- Kept the crate-level rustdoc command-line example contiguous by moving the
  test-only crate attribute out of the middle of the docs

### Fixed
- Aligned Zenodo/DOI wording across the README, release notes, and release
  process docs so the repo no longer overclaims archival status
- Added an automated guard for the README screenshot asset referenced at the top
  of the project page

## [1.0.1] - 2026-03-16

### Changed
- Automated the tag-driven release workflow to publish to crates.io and PyPI
  before creating the GitHub release
- Added Python source distribution (`sdist`) packaging to the release pipeline
- Made release publish steps rerun-safe when a tagged version is already present
  on crates.io or PyPI

### Fixed
- Resolved `maturin sdist` packaging for `ries-py` by using a package-local
  PyPI readme and explicit Python package metadata

## [1.0.0] - 2026-03-14

### Added
- PSLQ integer relation detection (`--pslq`, `--pslq-extended`, `--pslq-max-coeff`)
- WebAssembly bindings for browser and Node.js usage
- Python bindings with PyO3/maturin
- High-precision verification using arbitrary precision arithmetic
- Stability ladder for impostor detection
- Domain presets for targeted searches (analytic-nt, physics, etc.)
- Run manifest for reproducibility (`--emit-manifest`)
- Deterministic mode for reproducible results (`--deterministic`)
- CITATION.cff for academic citation

## [0.1.0] - 2026-02-18

### Added
- Initial release
- Core equation search with parallel execution
- Automatic differentiation for Newton-Raphson refinement
- User-defined constants and functions via profiles
- Multiple output formats (default, pretty, Mathematica, SymPy)
- Complexity scoring for equation ranking
- Classic mode for compatibility with original RIES
- CLI with comprehensive options

### References
- Based on [RIES by Robert Munafo](https://mrob.com/pub/ries/)
- Compatible with [clsn/ries fork](https://github.com/clsn/ries)