V0.5

.gitignore

@@ -13,6 +13,7 @@ build/
 .venv/
 venv/
 .env
+.uv-cache/
 
 # Testing
 .pytest_cache/

CHANGELOG.md

@@ -0,0 +1,75 @@
+# v0.5.0
+
+First release after v0.2.0. This is a large jump — the reports system has been replaced by a proper plugin architecture, parsing has grown in several directions, and the CLI, LSP, and docs have all been reworked. The notes below group changes by theme rather than by commit.
+
+## Breaking changes
+
+- **`@exercise` blocks are now `@movement` blocks.** The block type, the parsed dataclass (`MovementDefinition`), and the tree-sitter grammar all use the new name. Update any `.ox` logs that use the old directive.
+- **The `run` REPL command has been removed.** Use `plugins` to list available plugins, and invoke a plugin directly by name (e.g. `volume -m squat`). Avoid naming a custom plugin the same as a built-in command (`query`, `tables`, `reload`, `lint`, `plugins`, `help`, `exit`, `quit`) — built-ins win the name lookup.
+- **Reports have been replaced by plugins.** The previous `report` / `generate` CLI commands are gone. The `stats` and `history` plugins were removed; other analyses moved to the new plugin system.
+- **Unit normalization.** `lbs` has been unified to `lb`. The grammar now accepts any pint-compatible mass unit (`g`, `oz`, `stone`, `grain`, `kg`, `lb`, …).
+- **Time tokens are ISO 8601.** Durations are written `PT30M`, `PT1H30M15S`, etc.
+
+## New plugins
+
+Four built-in plugins ship with this release:
+
+- **`e1rm`** — Estimated 1RM progression from sets tagged with `^rm`, with Brzycki and Epley formulas and table/plot output.
+- **`weighin`** — Body-weight tracking with rolling average, trend, multi-scale breakdown, and plot.
+- **`srpe`** — Session RPE training load, AU totals per time bin, and ACWR / monotony / strain output modes.
+- **`wendler531`** — Generates a 4-week Wendler 5/3/1 cycle as planned sessions (`!` flag), with optional `^rm` tagging and configurable start date and unit.
+
+The existing `volume` plugin remains.
+
+## Plugin system
+
+- Plugins are first-class. A plugin exports `register()` and its functions receive `PluginContext(db, log)`, returning `TableResult`, `TextResult`, or `PlotResult`.
+- Load user plugins from your log with `@plugin "path/to/plugin.py"`. Paths resolve relative to the `.ox` file.
+- User plugins loaded via `@plugin` override built-ins with the same name.
+- On startup, the CLI prints the list of user plugins that were loaded.
+- The REPL lists plugins via `plugins` and invokes them by name.
+
+## Parser and language features
+
+- **Movement definitions** — `@movement name … @end` blocks with `equipment`, `tags`, `note`, and `url` fields. Parsed into `MovementDefinition` and exposed on `TrainingLog.movement_definitions`.
+- **Notes are first-class objects** — both single-line `note "…"` entries and in-session `note:` lines flow through `Note` / session notes and into the database.
+- **Stored queries** — `2025-01-10 query "name" "SELECT …"` lines are parsed and surfaced via the `query` command by name.
+- **Weigh-ins** — `date W weight [time] [scale]` lines parse into `WeighIn` dataclasses and populate a `weigh_ins` table.
+- **Implied units in progressive weights** — `160/185/210lb` now parses correctly; each segment inherits the nearest following unit.
+- **`BW` inside progressions** — `BW/24kg/32kg` and similar forms work without lint errors.
+- **Parse diagnostics / linter** — parse errors are collected on load and surfaced via the `lint` command and through the LSP.
+- **SQLite `REGEXP`** — available in `query` expressions.
+- **Short flags** on plugin parameters (e.g. `-m`, `-b`).
+
+## CLI
+
+- `plugins` — new command to list available plugins.
+- Plugins are invoked directly by name. Running a plugin with no args prints its usage.
+- `reload` — re-parse the current log from disk without leaving the REPL. Reprints parse diagnostics and re-announces loaded user plugins.
+- `tables -h` — show column details alongside the table/view list.
+- `query name` — recall a stored query by name, or run inline SQL with `query SELECT …`.
+- `--version` reads from `pyproject.toml` so there is a single source of truth.
+
+## Plots
+
+All built-in plots now route through a small `plot` facade over `plotext`, giving consistent axes, markers, and legends across `e1rm`, `weighin`, and `srpe`. The earlier hand-rolled ASCII plots are gone.
+
+## Editor support
+
+- **VSCode** — syntax highlighting updated to cover `@movement`, `@session`, `@template`, `@plugin`, `@include`, `note` entries, `query` entries, and the `equipment`/`tags`/`note`/`url` fields inside movement definitions. Comment folding is fixed.
+- **LSP** — movement-name autocomplete is populated from `@movement` blocks in the parsed log; diagnostics surface parse errors and invalid `@include` paths; comment folding is supported.
+
+## Documentation
+
+- `docs/getting-started.md` covers movement definitions, session-level notes, stored queries, and `@plugin` loading.
+- `docs/plugins.md` documents every built-in plugin, loading rules, the plugin API, and reserved names.
+- `docs/api-reference.md` now lists every dataclass (including `MovementDefinition`), the full `TrainingLog` surface, and the plugin result types.
+- `docs/cli-reference.md` reflects the current command set.
+- `docs/editor-support.md` clarifies that VSCode is the only shipped extension; Neovim and Helix users can wire up the grammar and `ox-lsp` directly.
+
+## Fixes
+
+- `e1rm` only considers completed sessions.
+- Progressive weights with `BW` no longer trigger lint errors.
+- Weekly time bins default to Sunday dates.
+- Single-line sessions now use the movement name as the session name.

CLAUDE.md

@@ -20,47 +20,52 @@ uv run ruff format src/ tests/ # format
 
 ```
 src/ox/
-  parse.py    - Tree-sitter node → data structures (the core parser)
-  data.py     - Dataclasses: TrainingSet, Movement, TrainingSession, TrainingLog, Note, WeighIn, StoredQuery, Diagnostic
-  db.py       - In-memory SQLite layer: create_db(log) → Connection
-  reports.py  - Reports: volume, matrix; get_all_reports() merges builtin + plugin reports
-  plugins.py  - Plugin discovery and registry (report + generator types)
-  units.py    - Pint unit registry (shared instance)
-  cli.py      - Click CLI with interactive REPL (stats, history, report, generate, query, tables, lint, reload)
-  lsp.py      - LSP server: diagnostics, movement completion, comment folding
-  lint.py     - Parse error collection for CLI lint command and LSP
+  parse.py      - Tree-sitter node → data structures (the core parser)
+  data.py       - Dataclasses: TrainingSet, Movement, TrainingSession, TrainingLog, Note, WeighIn, StoredQuery, Diagnostic
+  db.py         - In-memory SQLite layer: create_db(log) → Connection
+  plugins.py    - Plugin discovery, registry, PluginContext, result types (TableResult, TextResult, PlotResult)
+  sql_utils.py  - SQL helper utilities for plugins (parse_plugin_args, plugin_usage, _weight_sql_expr, _time_bin_expr)
+  units.py      - Pint unit registry (shared instance)
+  cli.py        - Click CLI with interactive REPL (run, query, tables, lint, reload)
+  lsp.py        - LSP server: diagnostics, movement completion, comment folding
+  lint.py       - Parse error collection for CLI lint command and LSP
   builtins/
-    e1rm.py        - Estimated 1RM report (Brzycki/Epley)
-    weighin.py     - Weigh-in stats/plot report (rolling average, trend, multi-scale)
-    wendler531.py  - Wendler 5/3/1 cycle generator
+    volume.py      - Volume over time plugin
+    e1rm.py        - Estimated 1RM plugin (Brzycki/Epley)
+    weighin.py     - Weigh-in stats/plot plugin (rolling average, trend, multi-scale)
+    srpe.py        - Session RPE training load plugin (ACWR, monotony, strain)
+    wendler531.py  - Wendler 5/3/1 cycle generator plugin
 tests/
   conftest.py        - Shared fixtures (simple_log_*, weight_edge_cases, log_with_query_*, log_with_weigh_ins_*, weigh_in_multi_scale_*, simple_db, example_db)
   test_parse.py      - Weight/rep parsing
   test_data.py       - Data structures
   test_db.py         - SQLite schema, loading, views, queries
-  test_reports.py    - Reports, arg parsing, registry
+  test_reports.py    - SQL utils, volume plugin, arg parsing, plugin registry
   test_plugins.py    - Plugin registration, loading, builtins
   test_integration.py - End-to-end parsing
-  test_weighin.py    - Weigh-in report (rolling avg, trend, table/plot/stats)
+  test_weighin.py    - Weigh-in plugin (rolling avg, trend, table/plot/stats)
   test_notes.py      - Note parsing, session notes, DB population
+  test_srpe.py       - sRPE plugin (training load, ACWR, monotony, strain)
   test_lint.py       - Diagnostic collection
 tree-sitter-ox/
   grammar.js  - Tree-sitter grammar definition for .ox format
 editors/
   vscode/     - VSCode extension for .ox syntax highlighting
 examples/
-  plugins/    - Example plugin scripts (wendler531.py)
+  plugins/            - Example plugin scripts (wendler531.py)
+  plugin_template.py  - Template for writing user plugins
 docs/         - MkDocs documentation source
-example/
-  example.ox  - Reference training log with all supported formats
+examples/
+  example.ox   - Reference training log with all supported formats
+  advanced.ox  - sRPE tracking example with 8 weeks of training data
 ```
 
 ## .ox File Format
 
 ```
 # Comments start with #
 
-# Single-line entry: date flag exercise: weight reps "note"
+# Single-line entry: date flag movement: weight reps "note"
 2025-01-10 * pullups: BW 5x10
 
 # Session block
@@ -82,9 +87,24 @@ kb-oh-press: 24kg 5/5/5
 # Include another file
 @include "other.ox"
 
+# Movement definition
+@movement squat
+equipment: barbell
+tags: squat, lower
+note: back squat
+@end
+
+# Template block
+@template "my-template"
+movement: details
+@end
+
+# Load a plugin
+@plugin "my_plugin.py"
+
 # Flags: * = completed, ! = planned, W = weigh-in
 # Weight units: kg, lb, g, oz, stone, grain, and more (any pint-compatible mass unit)
-# Weight formats: 24kg, BW, 24kg+32kg (combined), 24kg/32kg/48kg (progressive)
+# Weight formats: 24kg, BW, 24kg+32kg (combined), 24kg/32kg/48kg (progressive), 160/185/210lb (implied unit)
 # Rep formats: 5x5 (sets x reps), 5/5/5 (per-set reps)
 # Duration: ISO 8601 (PT30M, PT1H30M15S)
 # Distance: numeric + unit (m, km, ft, mi, etc.)
@@ -95,9 +115,9 @@ kb-oh-press: 24kg 5/5/5
 - Python 3.12, dependencies managed with uv
 - Frozen dataclasses with `slots=True` for data structures
 - `pint.Quantity` for all weight values (never raw numbers)
-- Exercise names are hyphenated lowercase (e.g. `kb-oh-press`, `bench-press`)
+- Movement names are hyphenated lowercase (e.g. `kb-oh-press`, `bench-press`)
 - `to_ox()` methods serialize back to .ox format (round-trip support)
 - Tree-sitter nodes are processed in `parse.py`; data structures live in `data.py` — keep this separation
+- All analysis features are plugins (builtins or user-defined). Plugins receive `PluginContext(db, log)` and return `TableResult`, `TextResult`, or `PlotResult`
+- CLI commands: `plugins` to list available plugins, `query` for raw SQL. Plugins are invoked by name directly (e.g. `volume -m squat`)
 
-## Known Issues
-- Progressive weights for the same movement require explicit units, this is a known bug and applicable tests are skipped.

README.md

@@ -1,6 +1,6 @@
 # Ox
 
-Plain text training log format and toolchain. Write workouts in `.ox` files, parse into structured data, analyze progress over time.
+Plain text training log format and toolchain. Record training in `.ox` files, parse into structured data, analyze progress over time.
 
 Inspired by [Beancount](https://github.com/beancount/beancount) (plain text accounting, but for training). Named after [Milo of Croton](https://en.wikipedia.org/wiki/Milo_of_Croton).
 
@@ -35,7 +35,7 @@ Full docs at [konnerhorton.github.io/ox](https://konnerhorton.github.io/ox):
 - [CLI Reference](https://konnerhorton.github.io/ox/cli-reference/) — commands and usage
 - [Reports & Plugins](https://konnerhorton.github.io/ox/plugins/) — built-in reports, plugin system
 - [API Reference](https://konnerhorton.github.io/ox/api-reference/) — Python library
-- [Editor Support](https://konnerhorton.github.io/ox/editor-support/) — VSCode, Neovim, Helix
+- [Editor Support](https://konnerhorton.github.io/ox/editor-support/) — VSCode extension, LSP, tree-sitter grammar
 
 ## Syntax Overview
 
@@ -48,7 +48,7 @@ Full docs at [konnerhorton.github.io/ox](https://konnerhorton.github.io/ox):
 2025-01-15 * Lower Body
 squat: 135lb 5x5
 deadlift: 185lb 3x5
-note: easy day
+note: "easy day"
 @end
 
 # Weigh-in
@@ -59,19 +59,29 @@ note: easy day
 
 # Include another file
 @include "other.ox"
+
+# Movement definition
+@movement squat
+equipment: barbell
+tags: squat, lower
+note: back squat
+@end
+
+# Load a plugin
+@plugin "plugins/my_plugin.py"
 ```
 
 **Flags:** `*` completed, `!` planned, `W` weigh-in
 
-**Weights:** `24kg`, `135lb`, `BW`, `24kg+32kg` (combined), `24kg/32kg/48kg` (progressive)
+**Weights:** `24kg`, `135lb`, `BW`, `24kg+32kg` (combined), `24/32/48kg` (progressive, with implied units)
 
 **Reps:** `5x5` (sets x reps), `5/3/1` (per-set)
 
 **Duration:** ISO 8601 (`PT30M`, `PT1H30M15S`)
 
 **Distance:** `5km`, `3mi`, `400m`
 
-**Exercise names:** no spaces, hyphenated lowercase (`kb-oh-press`, `bb-back-squat`)
+**Movement names:** no spaces (`kb-oh-press`, `bb-back-squat`)
 
 ## Installation
 

SPEC.md

@@ -27,10 +27,10 @@ Developers and power users comfortable with text files and CLIs.
 - **Python parser** (`src/ox/parse.py`) — tree-sitter nodes → dataclasses
 - **Data model** (`src/ox/data.py`) — `TrainingSet`, `Movement`, `TrainingSession`, `TrainingLog`, `Note`, `WeighIn`, `StoredQuery`, `Diagnostic`
 - **SQLite query layer** (`src/ox/db.py`) — in-memory DB with `sessions`, `movements`, `sets`, `notes`, `session_notes`, `weigh_ins`, `queries` tables and `training` view
-- **Plugin system** (`src/ox/plugins.py`) — discovery from `~/.ox/plugins/`, entry points, and builtins; report and generator types
+- **Plugin system** (`src/ox/plugins.py`) — built-in plugins plus user plugins loaded via `@plugin` directives in `.ox` files
 - **Built-in reports** (`src/ox/reports.py`) — `volume` (volume over time) and `matrix` (session count per movement)
 - **Built-in plugins** — `e1rm` (estimated 1RM via Brzycki/Epley), `weighin` (weight tracking with stats/plot/rolling average), `wendler531` (5/3/1 cycle generator)
-- **CLI** (`src/ox/cli.py`) — interactive REPL with `stats`, `history`, `report`, `generate`, `query`, `tables`, `lint`, `reload` commands and tab completion
+- **CLI** (`src/ox/cli.py`) — interactive REPL with `report`, `generate`, `query`, `tables`, `lint`, `reload` commands and tab completion
 - **LSP** (`src/ox/lsp.py`) — diagnostics (syntax errors + include validation), movement name completion, comment folding ranges
 - **Weigh-in tracking** — full pipeline: parse → `WeighIn` dataclass → DB → builtin report with table/plot/stats output
 - **Notes** — standalone and session-level notes, parse → `Note` dataclass → DB, `to_ox()` round-trip
@@ -45,22 +45,21 @@ Developers and power users comfortable with text files and CLIs.
 ### What's incomplete
 
 - Planned sessions (`!` flag) — parsed but ignored in analysis
-- Exercise definitions (`@exercise` blocks) — parsed but not used in analysis
 - Template blocks (`@template`) — grammar exists, no processing
 - Progressive implied weights (e.g. `160/185/210lbs`) — known parsing bug
-- CLI exercise autocompletion (tab-complete exercise names, not just commands)
+- CLI movement autocompletion (tab-complete movement names, not just commands)
 
 ## Direction
 
 ### Richer analysis
 
 - Cycle tracking — micro/meso/macro periodization
-- Exercise definitions feeding into analysis (e.g. grouping by movement pattern)
+- Movement definitions feeding into analysis (e.g. grouping by movement tag)
 - `pint.Quantity` for time/distance — enables derived units like pace and speed
 
 ### Better editor experience
 
-- LSP hover info (exercise definitions, recent history for a movement)
+- LSP hover info (movement definitions, recent history for a movement)
 - LSP completions for session templates
 - Snippets for common entry patterns
 

docs/api-reference.md

@@ -28,6 +28,9 @@ All are frozen dataclasses with `slots=True`.
 |---|---|
 | `sessions` | `tuple[TrainingSession, ...]` |
 | `notes` | `tuple[Note, ...]` |
+| `weigh_ins` | `tuple[WeighIn, ...]` |
+| `queries` | `tuple[StoredQuery, ...]` |
+| `movement_definitions` | `tuple[MovementDefinition, ...]` |
 | `diagnostics` | `tuple[Diagnostic, ...]` |
 
 **Properties:** `completed_sessions`, `planned_sessions`
@@ -65,6 +68,18 @@ All are frozen dataclasses with `slots=True`.
 
 **Properties:** `volume` (`reps × weight`, or `None` for BW)
 
+### MovementDefinition
+
+| Attribute | Type |
+|---|---|
+| `name` | `str` |
+| `equipment` | `str \| None` |
+| `tags` | `tuple[str, ...]` |
+| `note` | `str \| None` |
+| `url` | `str \| None` |
+
+Parsed from `@movement` blocks. Used by the LSP for name completion; queryable directly off the log.
+
 ### WeighIn
 
 | Attribute | Type |
@@ -144,3 +159,35 @@ session.to_ox()     # serialize session to .ox format
 movement.to_ox()    # serialize movement
 note.to_ox()        # serialize note
 ```
+
+## Plugin API
+
+Plugins receive a `PluginContext` and return one of three result types. All are frozen dataclasses in `ox.plugins`.
+
+### PluginContext
+
+| Attribute | Type |
+|---|---|
+| `db` | `sqlite3.Connection` |
+| `log` | `TrainingLog` |
+
+### TableResult
+
+| Attribute | Type |
+|---|---|
+| `columns` | `list[str]` |
+| `rows` | `list[tuple]` |
+
+### TextResult
+
+| Attribute | Type |
+|---|---|
+| `text` | `str` |
+
+### PlotResult
+
+| Attribute | Type |
+|---|---|
+| `lines` | `list[str]` |
+
+See [Plugins](plugins.md) for a walkthrough of writing a plugin and registering it via `register()`.