motedb-python 0.15.2__tar.gz → 0.16.0__tar.gz

motedb_python-0.16.0/.github/FUNDING.yml

@@ -0,0 +1,11 @@
+# MoteDB is maintained by one person, on evenings and weekends.
+# If MoteDB saved you time, or if you depend on it to run a robot in
+# production, a small GitHub Sponsors contribution is what keeps it going.
+#
+# See the community page for everything else:
+# https://github.com/motedb/motedb/blob/main/community/README.md
+
+github: [motedb]
+# Once a second funding route exists (or if you prefer not to use GitHub
+# Sponsors), replace the line above with:
+# custom: ["<your-funding-url>"]

{motedb_python-0.15.2 → motedb_python-0.16.0}/.github/workflows/ci.yml

@@ -32,6 +32,15 @@ jobs:
       - name: cargo clippy (advisory — warnings do not fail the job)
         run: cargo clippy --all-targets --all-features --no-deps
         continue-on-error: true
+      # e3ebb24 shipped a mid-match wildcard in BOTH bindings: every concrete
+      # arm after it became unreachable (a warn-only lint) and Python returned
+      # None for every value. The root-crate clippy above never sees the
+      # binding crates — they are not workspace members. Deny the lint here.
+      - name: bindings check — unreachable_patterns denied
+        run: |
+          for m in bindings/python/Cargo.toml bindings/node/Cargo.toml; do
+            RUSTFLAGS="-D unreachable_patterns" cargo check --manifest-path "$m" --quiet
+          done
 
   # ── Supply-chain: vulnerability + license gate (cargo-deny) ──────────────
   # Fails on any NEW RustSec advisory (vulnerability/unsound). Four known
@@ -322,7 +331,7 @@ jobs:
           args: --release
       - name: Install wheel
         run: pip install bindings/python/target/wheels/*.whl
-      - name: Run E2E workload (11 suites)
+      - name: Run E2E workload (12 suites)
         env:
           MOTE_CLI: ${{ github.workspace }}/target/release/motedb-cli
         run: python bindings/python/e2e/e2e_workload.py --root /tmp/e2e

motedb_python-0.16.0/CAPABILITIES.md

@@ -0,0 +1,178 @@
+# Capability Manifest
+
+> This file is the **single source of truth** for what MoteDB can do.
+> The website, the README, blog posts and any external material should cite
+> this manifest rather than restating feature claims.
+> The point: every line below should map to an implementation and a test.
+
+**Version baseline:** v0.15.2 · **Last verified:** 2026-10-24
+
+Legend: `✅ supported` · `🟡 partial (see note)` · `❌ not supported` · `🟢 differentiator`
+
+> If you find a line here that does not match actual behaviour, please open an
+> issue — **those are the ones we triage first**.
+
+---
+
+## 1. SQL engine
+
+| Capability | Status | Notes |
+|---|---|---|
+| `CREATE TABLE` / `CREATE INDEX` / `DROP TABLE [IF EXISTS]` / `DROP INDEX` | ✅ | Column, vector, text, spatial, timestamp index variants |
+| `ALTER TABLE` | ✅ | `ADD COLUMN`, `AUTO_INCREMENT = N` |
+| `INSERT` (incl. `INSERT ... SELECT`) | ✅ | |
+| `UPDATE` / `DELETE` | ✅ | |
+| `SELECT` with `WHERE` | ✅ | |
+| `JOIN` — INNER / LEFT / RIGHT / FULL | ✅ | |
+| Subqueries in `WHERE`, derived tables (`FROM (SELECT ...)`) | ✅ | |
+| `GROUP BY` | ✅ | Columns, expressions (`id % 5`), and SELECT aliases |
+| `HAVING` | ✅ | Incl. aggregate aliases |
+| `ORDER BY` | ✅ | Multi-key, expressions, non-projected columns, `NULLS FIRST/LAST` |
+| `LIMIT` / `OFFSET` | ✅ | |
+| `DISTINCT` (rows) / `COUNT(DISTINCT col)` | ✅ | |
+| Aggregates: `COUNT` `SUM` `AVG` `MIN` `MAX` `STDDEV` `VARIANCE` | ✅ | |
+| `UNION` / `UNION ALL` | ✅ | |
+| `CASE WHEN ... THEN ... ELSE ... END` | ✅ | |
+| `WITH` / CTEs, incl. `WITH RECURSIVE` | ✅ | Semi-naive fixed-point iteration, `UNION ALL` bodies, chained CTEs |
+| Window functions: `ROW_NUMBER` `RANK` `DENSE_RANK` `NTILE` `LAG` `LEAD` `FIRST_VALUE` `LAST_VALUE` + aggregates | ✅ | `OVER (PARTITION BY … ORDER BY …)` |
+| Explicit window frames (`ROWS` / `RANGE BETWEEN <bound> AND <bound>`) | ✅ 🟢 | ROWS = physical rows; RANGE = peer groups, or value intervals (`RANGE BETWEEN 10 PRECEDING AND 10 FOLLOWING`) on a numeric/timestamp order key. Default frame is the RANGE peer-group frame. Tested in `tests/test_window_frames.rs` |
+| Scalar functions: `UPPER` `LOWER` `LENGTH` `TRIM` `SUBSTR` `REPLACE` `CONCAT` `INSTR` `COALESCE` `ROUND` `ABS` + arithmetic | ✅ | `CONCAT` skips NULL args (SQLite/Postgres semantics); `\|\|` propagates NULL; `ROUND` is binary-exact decimal rounding |
+| `EXPORT TABLE t TO 'x.parquet'` / `IMPORT PARQUET 'x' INTO t` | ✅ 🟢 | Columnar dump / batch load through the insert pipeline; pyarrow / pandas / DuckDB round-trip |
+| `EXPORT LEROBOT <table> TO 'dir/' [FPS n] [TASKS col]` | ✅ 🟢 | Writes a LeRobot v3 dataset on disk that the official `LeRobotDataset` (lerobot ≥ 0.6) loads directly |
+| `DECIMAL` / `DATE` types | ❌ | Planned — see [ROADMAP](./ROADMAP.md) |
+| Cross-statement server-side cursors | ❌ | Planned — see [ROADMAP](./ROADMAP.md) |
+
+---
+
+## 2. Multimodal types
+
+| Type | Status | Notes |
+|---|---|---|
+| `VECTOR(n)` | ✅ 🟢 | Index-accelerated KNN; `ORDER BY emb <-> '[…]'` (L2) or `<~>` (cosine) |
+| `BLOB` | 🟡 | Byte-exact round-trip, **16 MiB cap per value** |
+| `TIMESTAMP` | ✅ 🟢 | Microsecond precision; `TIMESERIES(ts)` adds time-series semantics |
+| `TEXT` | ✅ | Built-in tokenizer (jieba and others configurable) |
+| Spatial coordinates | ✅ 🟢 | R-Tree index; `ST_POINT` / `ST_WITHIN` / `ST_DISTANCE` / `ST_KNN`, and `loc <-> ST_POINT(x, y)` distance ordering |
+| Array types (`TEXT[]` / `ARRAY<FLOAT>`) | ❌ | Use BLOB, a separate table, or TEXT + JOIN |
+
+### ⚠️ Known BLOB gaps (unchanged since v0.13)
+
+- Scalar functions over BLOB columns (e.g. `LENGTH(payload)`) are **not wired through all executor paths**
+- BLOB output under `ORDER BY` is **not wired through all executor paths**
+- **Verified working:** bare projection, equality (`WHERE payload = X'…'`), time-window scans, parameter binding
+
+---
+
+## 3. Indexes
+
+| Index | Status | Notes |
+|---|---|---|
+| Column index | ✅ 🟢 | Equality / range |
+| Vector index | ✅ 🟢 | **DiskANN** (FreshDiskANN + rerank). **Not** HNSW, not IVF |
+| Full-text index | ✅ 🟢 | BM25 ranking + snippet; `MATCH(col) AGAINST('q')`; multi-word defaults to AND (SQLite FTS5-compatible), explicit `a OR b` unions |
+| Spatial index | ✅ 🟢 | R-Tree |
+| Time-series index | ✅ 🟢 | Range scan + compaction + `TIME_BUCKET` aggregation + TTL |
+
+> 📌 **Correction notice.** Older material describing the vector index as
+> "HNSW / IVF" is out of date. The current implementation is **DiskANN**.
+> This file and `docs/08-vector-index.md` are authoritative.
+
+---
+
+## 4. Transactions & consistency
+
+| Capability | Status | Notes |
+|---|---|---|
+| `BEGIN` / `COMMIT` / `ROLLBACK` | ✅ | |
+| Savepoints | ✅ | |
+| Read-your-writes inside a transaction | ✅ | Scans, aggregates, point lookups |
+| MVCC | ✅ | |
+| WAL group commit | ✅ 🟢 | |
+| `LATEST BY` on tables > 1M rows | 🟡 | Performance not yet optimized (large-table merge path) |
+
+---
+
+## 5. Retrieval
+
+| Capability | Status | Notes |
+|---|---|---|
+| Vector KNN | ✅ 🟢 | `ORDER BY emb <-> '[…]'`, index-accelerated |
+| Full-text search | ✅ 🟢 | BM25 ranking |
+| **Hybrid search (BM25 × KNN, RRF fusion)** | ✅ 🟢 | One call — you don't implement score calibration yourself |
+| Time-series window aggregation | ✅ 🟢 | `TIME_BUCKET` + TTL |
+| Spatial range / KNN | ✅ 🟢 | `ST_WITHIN` / `ST_DISTANCE` / `ST_KNN` |
+
+---
+
+## 6. Data shapes & deployment
+
+| Capability | Status | Notes |
+|---|---|---|
+| Collection: high-performance directory | ✅ 🟢 | Atomic publish via temp + fsync + rename, plus WAL group commit |
+| Transport: single-file snapshot | ✅ 🟢 | `EXPORT DB TO '…'`; per-file CRC32, optional zstd, incremental re-export |
+| Deployment: read-only container | ✅ 🟢 | All write paths rejected, multiple concurrent readers |
+| MCAP import | ✅ 🟢 | |
+| Index self-healing | ✅ 🟢 | Rebuilds a corrupt index from source data |
+| Crash recovery | ✅ 🟢 | WAL replay + segment migration + orphan segment adoption |
+| `for_edge()` low-memory profile | ✅ | Tuned for small-memory devices |
+
+---
+
+## 7. Interop & SDKs
+
+| Surface | Status | Notes |
+|---|---|---|
+| Rust crate | ✅ | `cargo add motedb` (Rust ≥ 1.87) |
+| Python | ✅ | `pip install motedb-python` — imported as `motedb`. The `motedb` name on PyPI is **not yet claimed**; use `motedb-python` |
+| Arrow | ✅ 🟢 | `query_arrow()` → canonical `pyarrow.Table`, vector columns as `FixedSizeList<float32>` |
+| pandas | ✅ | `query_pandas()` |
+| numpy bulk load | ✅ | `insert_arrays()` |
+| Node.js | 🟡 | npm package **is published** (`motedb@0.15.2`). Prebuilt binaries cover macOS (arm64/x64) and Linux (x64/arm64, glibc); **musl and Windows prebuilds are not published yet**, so `npm install motedb` on Windows currently yields no native binary |
+| Go | 🟡 | Binding lives in `bindings/go` (with `go.mod`), but is **not published as a tagged Go module** yet — the module proxy only resolves a pseudo-version off the default branch |
+
+---
+
+## 8. Engineering quality
+
+| Practice | Status | Notes |
+|---|---|---|
+| Crash-injection tests | ✅ 🟢 | Resident in CI |
+| Differential fuzzing (SQLite as oracle) | ✅ 🟢 | Permanent regression suite, including JOIN diffing |
+| Corruption-recovery tests | ✅ 🟢 | |
+| Machine-independent performance gate | ✅ 🟢 | A complexity regression turns CI red |
+| `cargo fmt --check` strict gate | ✅ 🟢 | Whole tree is fmt-clean |
+| cargo-deny supply-chain audit | ✅ 🟢 | Vulnerabilities + licenses |
+| clippy | 🟡 advisory | CI runs clippy in advisory mode (`continue-on-error`); a backlog of style lints is being cleared. **New commits must not introduce new warnings** |
+| Prebuilt binaries | 🟡 | macOS (arm64/x64) and Linux (x64/arm64) prebuilt. Windows works through `cargo add motedb` (source build); a Windows prebuild is on the way |
+
+---
+
+## 9. Explicitly not supported
+
+> Written down so you don't have to discover it the hard way.
+
+| Capability | Why |
+|---|---|
+| **Built-in inference engine / ONNX** | MoteDB is a storage and query layer, not an inference runtime. We invest in **Arrow interop** and leave inference in your own stack |
+| **Distributed / multi-node replication** | The current target is single-machine edge |
+| **Server mode / multi-process concurrent writes** | Contradicts the "embedded, zero-ops" positioning |
+| **Multi-user permission system** | In embedded deployments the host application owns that |
+| **Array types** | Use BLOB, a separate table, or TEXT + JOIN |
+| **`DECIMAL` / `DATE` types** | Planned |
+| **Cross-statement server-side cursors** | Planned |
+
+---
+
+## 10. Performance methodology
+
+Full benchmark data: [`BENCHMARK.md`](./BENCHMARK.md),
+[`PERFORMANCE_REPORT.md`](./PERFORMANCE_REPORT.md), and
+[`docs/benchmark_methodology.md`](./docs/benchmark_methodology.md).
+
+> ⚠️ **Citation rule.** Every performance number published anywhere must carry
+> its **machine configuration, data scale, comparison baseline, and measurement
+> method**. Numbers without that context should not be released.
+
+---
+
+*Found something here that does not match real behaviour? Open an issue.*

{motedb_python-0.15.2 → motedb_python-0.16.0}/CHANGELOG.md

@@ -1,5 +1,75 @@
 # Changelog
 
+## [0.16.0] — 2026-10-30
+
+> 28 轮系统性猎虫与验证后的正确性版本。**42 个缺陷修复**（含 1 个数据
+> 损坏级、多个"静默错误结果"类），SQL 兼容性大幅提升（隐式别名、复合
+> 主键验证、中文全文检索、时间戳算术），四层 API（Rust/Node/Go/Python）
+> 与 MCP 全部在当前引擎上端到端验证。全面性能 A/B 对照确认无回归
+> （ORDER BY 快 ~28%，查询内存省 18%，其余持平）。
+
+### 静默错误结果类（最关键）
+
+- **🔴🔴 数据损坏**: 向量列在 UPDATE+DELETE+段合并后读成 65535 维全零
+  向量 — 列段合并的墓碑/ALTER 占位写裸 [0xFFFF]，逐行 [dim:u16] 编码里
+  是合法非空维度且无载荷 → 顺序走读者错位、同列后续全雪崩。占位改
+  [0x0000] (dim 0 = NULL)。回归: KNN 与暴力扫描全阶段逐 id 一致
+- WHERE 未知列被静默吞掉 — 单表 SELECT/UPDATE/DELETE 拼错列名返回
+  0 行/affected_rows: 0 无提示；语句入口 schema 绑定校验立即报
+  ColumnNotFound
+- SHOW TABLES / DESCRIBE 流式分发丢弃结果行（db.query 返回 []、
+  MCP list_tables 空）——行数据流经完整路径
+- 日期函数对 DATE 列返回 NULL（YEAR/MONTH/DAY_OF_MONTH/DATEDIFF）
+- CAST(x AS TEXT) 泄漏 Rust Debug 串（`Date { days: 19753 }`）——
+  现按 SQL 文本形态渲染（新增 Timestamp::to_iso）
+- ISO 时间戳小数秒被丢弃（'…12:00:00.5' 的 .5 消失，schema-order
+  INSERT 路径）——解析到 µs，双实现去重
+- 中文 MATCH 恒不命中 — FTS 索引与逐行 MATCH 都硬编码空白分词器；
+  新增 AutoTokenizer（CJK→Jieba，其余→空白）两路径统一
+- TIMESTAMP ± INTEGER(µs) 算术 TypeError 被吞 → WHERE 静默空
+- TIMESTAMP 列 BETWEEN 文本字面量恒空（vector_exec 预解析缺口，
+  `>= AND <=` 却正常）——同一 col_lit 预解析
+- GROUP BY 表达式 + ORDER BY 同表达式乱序（快路径丢键 + full_row
+  zip 错位配对）——表达式键读分组投影值，不可解析则明确报错
+- NULLS FIRST/LAST 被四条排序路径中的三条静默忽略——全路径统一
+- 0x 十六进制字面量静默返回 0（词法器缺十六进制 + 隐式别名吞噬）
+- MCP describe 工具返回 DESCRIBE 表头而非 schema（行被丢弃）
+
+### SQL 兼容性
+
+- **SELECT 列表隐式别名**（`SELECT v x`、`SELECT COUNT(*) c`）——此前
+  全部死于误导性解析错误；别名可在 ORDER BY/GROUP BY/HAVING 引用
+- 验证钉住: 复合 PRIMARY KEY（重复拒绝/双键点查/复合 upsert/重开强制）、
+  列 DEFAULT 值、NOT NULL 双路径强制、LIMIT ?/OFFSET ? 绑定参数
+
+### 语义钉住（文档化，防止引擎间歧义）
+
+- 除零: 整型/浮点除数一律 DivisionByZero（PostgreSQL 语义）
+- LIKE 与 = 一致大小写敏感；表名大小写敏感/列名不敏感
+- CONCAT 忽略 NULL（PG 风格），|| 按标准传播 NULL；DATEDIFF = a−b 秒
+- 同进程二次 open() 附着同一引擎（共享句柄注册表，docs/14）；跨进程
+  flock 互斥
+- DDL 非事务（CREATE 留存+行回滚，事务内 DROP 被拒）
+- 圆整 ROUND(x,-n) half-away；DATE 函数表（docs/13）；已知限制:
+  WHERE 中除零静默 0 行（v0.16+ 候选修复）
+
+### 质量与验证
+
+- clippy 双 feature 集零警告（默认集 + node 最小集——后者此前从未
+  被门禁覆盖）
+- 绑定层全面验证: Node smoke 双绿、Go 6/6 首跑、Python 13/13、
+  MCP stdio e2e 6/6（round-20/22 修复的端到端确认）
+- 性能 A/B（round-19 基线 vs 0.16.0，同机背靠背）: 无回归；ORDER BY
+  快 ~28%，查询后 RSS 省 18%，其余持平
+- 新增回归: test_bug_hunt_v110~v116（57 项）、MCP e2e
+
+### 此前积累（0.15.2 后、猎虫前）
+
+- 向量索引旋钮 SQL 化: CREATE VECTOR INDEX ... WITH (metric=, max_degree=,
+  build_search_list=, search_list=)；构建期精确 f32 快照
+- ORDER BY emb <-> '带引号向量' 三拼写回归；CREATE INDEX WITH 子句
+  死代码复活
+
 ## [0.15.2] — 2026-10-22
 
 ### 功能测试轮修复 (2026-10-22 外部功能测试 24/27 → 全绿)

motedb_python-0.16.0/CODE_OF_CONDUCT.md

@@ -0,0 +1,105 @@
+# Code of Conduct
+
+## Our pledge
+
+We as members, contributors, and maintainers pledge to make participation in
+this project a harassment-free experience for everyone, regardless of age, body
+size, visible or invisible disability, ethnicity, sex characteristics, gender
+identity and expression, level of experience, education, socio-economic status,
+nationality, personal appearance, race, religion, or sexual identity and
+orientation.
+
+We pledge to act and interact in ways that contribute to an open, welcoming,
+diverse, inclusive, and healthy community.
+
+---
+
+## Our standards
+
+### Behaviour that contributes to a positive environment
+
+- Demonstrating empathy and kindness toward other people
+- Being respectful of differing opinions, viewpoints, and experiences
+- Giving and gracefully accepting constructive feedback
+- Accepting responsibility, apologising to those affected by our mistakes, and
+  learning from the experience
+- Focusing on what is best for the overall community
+
+### Behaviour that is unacceptable
+
+- The use of sexualised language or imagery, and sexual attention or advances
+  of any kind
+- Trolling, insulting or derogatory comments, and personal or political attacks
+- Public or private harassment
+- Publishing others' private information, such as a physical or email address,
+  without their explicit permission
+- Other conduct which could reasonably be considered inappropriate in a
+  professional setting
+
+---
+
+## Scope
+
+This Code of Conduct applies within all community spaces, and also applies when
+an individual is officially representing the community in public spaces.
+Community spaces currently include:
+
+- The project's GitHub repository — issues, pull requests, discussions, commits,
+  and code review
+- The project's website and any space created in the project's name
+- Public events where the project is represented
+
+Any space the project opens later — a mailing list, a newsletter, or a chat
+channel — is covered by this Code of Conduct from the day it opens.
+
+---
+
+## Enforcement
+
+Instances of abusive, harassing, or otherwise unacceptable behaviour may be
+reported to the maintainers through any of these channels:
+
+- **GitHub issue or discussion**, mentioning `@meluo`
+- **GitHub private message** to [@meluo](https://github.com/meluo)
+- **Email** to <motedb@126.com> with the subject `[motedb conduct]`
+
+All complaints will be reviewed and investigated promptly and fairly. The
+project maintainer is obligated to respect the privacy and security of the
+reporter of any incident.
+
+**Our commitments:**
+
+- Reports are handled confidentially
+- The identity of the reporter is not disclosed publicly
+- The maintainer responds to reports within 48 hours
+
+---
+
+## Enforcement guidelines
+
+The project maintainers are responsible for clarifying and enforcing these
+standards, and will take appropriate and fair corrective action in response to
+any behaviour they deem inappropriate, threatening, offensive, or harmful.
+Enforcement actions may include:
+
+1. **Correction** — a private, written warning from the maintainers, providing
+   clarity around the nature of the violation and an explanation of why the
+   behaviour was inappropriate
+2. **Warning** — a warning with consequences for continued behaviour, including
+   a period of no interaction with the people involved
+3. **Temporary ban** — a temporary ban from any sort of interaction or public
+   communication with the community
+4. **Permanent ban** — a permanent ban from any sort of public interaction
+   within the community
+
+---
+
+## Attribution
+
+This Code of Conduct is adapted from the
+[Contributor Covenant](https://www.contributor-covenant.org/), version 2.1, and
+also draws on the
+[Rust Code of Conduct](https://www.rust-lang.org/policies/code-of-conduct).
+
+Questions or suggestions about this Code of Conduct belong in a discussion or
+an issue on this repository.

{motedb_python-0.15.2 → motedb_python-0.16.0}/Cargo.lock

@@ -1117,7 +1117,7 @@ dependencies = [
 
 [[package]]
 name = "motedb"
-version = "0.15.2"
+version = "0.16.0"
 dependencies = [
  "arc-swap",
  "arrow-array",

{motedb_python-0.15.2 → motedb_python-0.16.0}/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "motedb"
-version = "0.15.2"
+version = "0.16.0"
 edition = "2021"
 rust-version = "1.87"
 description = "AI-native embedded multimodal database for embodied intelligence (robots, AR glasses, industrial arms)."

motedb_python-0.16.0/GOVERNANCE.md

@@ -0,0 +1,153 @@
+# Governance
+
+MoteDB is an open-source project maintained by **one person**. This document
+explains **how decisions get made**, so you know where your input goes and who
+is accountable for the outcome.
+
+Being a single-maintainer project is a constraint, not a secret. It means
+turnaround is fast and nobody can quietly outvote you — and it also means the
+project needs your scrutiny more than a large committee would. Everything below
+is written to make that explicit.
+
+---
+
+## Governance model
+
+**The maintainer decides; the community participates.** Concretely:
+
+| Decision | Who decides | How the community participates |
+|---|---|---|
+| Roadmap priorities | Maintainer | Discussion votes + issue proposals |
+| Bug fix ordering | Maintainer | The maintainer explains priority in the issue thread |
+| API design | Maintainer | Open a Discussion first; breaking changes are announced at least 2 weeks ahead |
+| Docs & examples | Anyone can PR directly | — |
+| Day-to-day code review | Maintainer (+ committers, once appointed) | — |
+| Project direction (annual) | Maintainer | Community newsletter, annual AMA |
+
+---
+
+## How decisions are made
+
+### Default: consensus
+
+Most decisions do not need a vote — we try to reach agreement in discussion.
+
+**How to participate:** state your position in an issue or discussion → the
+maintainer integrates it → if there is no substantial objection, it ships.
+
+### Changes that require extra care
+
+The following need the maintainer's explicit, written agreement **and** a
+public comment window of at least 2 weeks:
+
+- Removing or renaming a public API
+- Introducing a breaking change (SQL semantics, file format, configuration keys)
+- Changing the minimum supported version (Rust version, platform support matrix)
+- Changing the license
+- Changing the project's positioning or target audience
+
+Because there is only one maintainer, there is no second approver to
+formally countersign. Instead of pretending otherwise, the maintainer
+**records the decision, the reasoning, and the fact that no second reviewer
+existed** in a tracking issue that stays open and visible. If a suitable
+maintainer or committer exists at the time, their agreement is sought and
+recorded in the same place.
+
+### Disputes
+
+When consensus is not reachable, the maintainer makes the final call and
+**states the reasoning**. Anyone who disagrees may ask for the decision to be
+recorded in an issue so it stays visible long-term — **we do not delete
+recorded decisions.**
+
+---
+
+## Roles
+
+### Contributor
+
+Anyone who files an issue, opens a pull request, or contributes docs or code.
+No application, no ceremony.
+
+### Committer
+
+Someone with write access to the repository who can merge PRs and close
+issues.
+
+**How the role is granted** (any one of these is enough):
+
+- Sustained contribution over at least 3 months, with ≥ 10 merged PRs
+- ≥ 5 high-quality PRs in one critical subsystem (vector index, full-text
+  search, the Python or Node binding), plus a willingness to maintain that
+  subsystem long-term
+- A clear recommendation from an existing committer, approved by the
+  maintainer
+
+**What it grants:** independent review and merge rights within your own
+subsystem, a voice in technical discussion, and a vote in release decisions.
+
+> The first committers will be chosen from the earliest contributors. The
+> maintainer will reach out to people who look like good candidates.
+
+### Champion
+
+Not a maintainer — a core contributor on the community side. Someone who is
+active in Discussions, writes Show & Tell posts, produces content in Chinese
+or English, organises events, and helps newcomers.
+
+**What it grants:** a public badge of recognition, early access to new
+versions and features, and a role in shaping community activities.
+
+---
+
+## Public commitments
+
+### The roadmap is public
+
+Major feature directions are written into [`ROADMAP.md`](./ROADMAP.md) with a
+status: `in progress` / `planned` / `done` / `not planned`.
+
+**"Not planned" is public information too** — it comes with a reason, so the
+community doesn't have to ask the same question twice.
+
+### Breaking changes are announced in advance
+
+Breaking changes are announced in Discussions **at least 2 weeks** before they
+land, including:
+
+- What changes and why
+- How to migrate, with runnable code
+- The tracking issue and the target version
+
+> Example: v0.12 changed the default semantics of multi-word `MATCH` from OR to
+> AND, to match SQLite FTS5. It was announced with a migration recipe — **this
+> is what the announcement mechanism is supposed to look like.**
+
+### Attribution
+
+Every merged PR is credited in the CHANGELOG and in the contributor list on
+the README. Every accepted issue proposal is publicly acknowledged.
+**This is a fixed part of the process, not an optional courtesy.**
+
+---
+
+## Handling disputes
+
+1. Discuss it rationally in the original issue or discussion
+2. If the discussion stalls, @ the maintainer
+3. If it concerns conduct, follow the reporting process in
+   [`CODE_OF_CONDUCT.md`](./CODE_OF_CONDUCT.md)
+4. If it is still unresolved: the maintainer makes the final call and records
+   the reasoning
+
+---
+
+## Contact
+
+Governance questions can go straight to a Discussion tagged `governance`, or
+to the maintainer ([@meluo](https://github.com/meluo)).
+
+---
+
+*This document changes as the community matures. Anyone can propose a change
+to it — open a pull request.*

{motedb_python-0.15.2 → motedb_python-0.16.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: motedb-python
-Version: 0.15.2
+Version: 0.16.0
 Classifier: Development Status :: 4 - Beta
 Classifier: Intended Audience :: Developers
 Classifier: Intended Audience :: Science/Research

{motedb_python-0.15.2 → motedb_python-0.16.0}/README.md

@@ -297,13 +297,13 @@ cargo add motedb
 Or in `Cargo.toml`:
 ```toml
 [dependencies]
-motedb = "0.15"
+motedb = "0.16"
 ```
 
 For minimal edge builds (no tokenizer, no parallelism), disable default features:
 ```toml
 [dependencies]
-motedb = { version = "0.15", default-features = false, features = ["jemalloc"] }
+motedb = { version = "0.16", default-features = false, features = ["jemalloc"] }
 ```
 
 ## Configuration
@@ -345,7 +345,10 @@ EXISTS]` / `DROP INDEX`, `ALTER TABLE` (`ADD COLUMN`, `AUTO_INCREMENT = N`),
   fixed-point, UNION ALL bodies, chained CTEs)
 - Window functions: `ROW_NUMBER`, `RANK`, `DENSE_RANK`, `NTILE`, `LAG`,
   `LEAD`, `FIRST_VALUE`, `LAST_VALUE` + aggregates `OVER (PARTITION BY …
-  ORDER BY …)` (RANGE peer-group default frame)
+  ORDER BY … [ROWS|RANGE BETWEEN <bound> AND <bound>])` — explicit frames
+  (ROWS physical rows; RANGE peer groups or value intervals like
+  `RANGE BETWEEN 10 PRECEDING AND 10 FOLLOWING` on the numeric/timestamp
+  order key; default RANGE peer-group frame)
 - Scalar functions: `UPPER`/`LOWER`/`LENGTH`/`TRIM`/`SUBSTR`/`REPLACE`/`CONCAT` (skips NULL args, SQLite/Postgres semantics; `||` propagates NULL)/`INSTR`/`COALESCE`/`ROUND` (binary-exact decimal rounding)/`ABS`/arithmetic
 - Interop: `EXPORT TABLE t TO 'x.parquet'` / `IMPORT PARQUET 'x' INTO t`
   (pyarrow/pandas/DuckDB round-trip), `EXPORT LEROBOT <episode-table> TO