plotaviz 0.1.0__tar.gz → 0.2.0__tar.gz

{plotaviz-0.1.0 → plotaviz-0.2.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: plotaviz
-Version: 0.1.0
+Version: 0.2.0
 Summary: Cross-platform desktop app that auto-selects and renders the right chart for any dataset — rules + scoring engine with optional LLM assist.
 Author: PlotaViz contributors
 License: MIT
@@ -21,7 +21,7 @@ Requires-Python: >=3.11
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: PySide6>=6.6
-Requires-Dist: pandas>=2.1
+Requires-Dist: pandas<4,>=2.1
 Requires-Dist: numpy>=1.26
 Requires-Dist: polars>=0.20
 Requires-Dist: pyarrow>=14
@@ -55,6 +55,7 @@ Dynamic: license-file
 # PlotaViz
 
 [![CI](https://github.com/saadibnainan/plotaviz/actions/workflows/ci.yml/badge.svg)](https://github.com/saadibnainan/plotaviz/actions/workflows/ci.yml)
+[![PyPI](https://img.shields.io/pypi/v/plotaviz)](https://pypi.org/project/plotaviz/)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue)](https://www.python.org/downloads/)
 [![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
 [![Qt: LGPLv3](https://img.shields.io/badge/Qt-LGPLv3-orange)](THIRD_PARTY_LICENSES.md)
@@ -101,7 +102,14 @@ between them in one click. The parts that usually go wrong are treated as first-
 
 Python 3.11 or newer.
 
-### From source (all platforms)
+### From PyPI (all platforms)
+
+```bash
+pip install plotaviz
+plotaviz
+```
+
+### From source (contributing / latest main)
 
 ```bash
 git clone https://github.com/saadibnainan/plotaviz.git
@@ -118,7 +126,7 @@ Apple Silicon and Intel are both supported.
 
 ```bash
 brew install python@3.11          # if you need it
-pip install -e .
+pip install plotaviz
 ```
 
 A prebuilt `.app` is attached to each [release](https://github.com/saadibnainan/plotaviz/releases).
@@ -146,7 +154,7 @@ sudo dnf install python3-devel libglvnd-egl xcb-util-cursor libxkbcommon-x11
 sudo pacman -S python libxkbcommon-x11 xcb-util-cursor
 ```
 
-Then either `pip install -e .`, or grab the **AppImage** from the releases page:
+Then either `pip install plotaviz`, or grab the **AppImage** from the releases page:
 
 ```bash
 chmod +x PlotaViz-*.AppImage
@@ -159,9 +167,9 @@ To build one: `make build-appimage`. Recipes for `.deb`, `.rpm`, and an **AUR PK
 ### Optional extras
 
 ```bash
-pip install -e ".[llm]"     # Anthropic / OpenAI / Gemini / Ollama providers
-pip install -e ".[fast]"    # pyqtgraph fast path for very large numeric plots
-pip install -e ".[dev]"     # tests, ruff, mypy, pre-commit
+pip install "plotaviz[llm]"     # Anthropic / OpenAI / Gemini / Ollama providers
+pip install "plotaviz[fast]"    # pyqtgraph fast path for very large numeric plots
+pip install -e ".[dev]"         # tests, ruff, mypy, pre-commit — source checkout only
 ```
 
 ---
@@ -173,13 +181,21 @@ plotaviz                                  # open the app
 plotaviz samples/sales_timeseries.csv     # open the app with a file
 ```
 
-1. Drag a dataset in, or use **File → Open**.
-2. Read the **cleaning report** — what was renamed, filled, deduplicated, flagged.
+1. Drag a dataset in, or use **File → Open**. For a semicolon file, a non-UTF-8 export, a title
+   row above the header, or a specific Excel sheet, use **File → Open with options…**.
+2. Read the **cleaning report** — what was renamed, deduplicated, flagged. Change how rows are
+   cleaned in the **Cleaning** tab; missing values are left alone unless you choose otherwise.
 3. Check the **Types** tab and correct anything inference got wrong.
-4. Pick a chart from **Charts**. The reasoning for each is right there.
+4. Pick a chart from **Charts** — the reasoning for each is right there — or build one in the
+   chart row above it (works offline). Adjust title, labels, scales, and colours under
+   **Customize**.
 5. Narrow the data in **Filters** — per-column widgets, or a raw pandas query.
-6. **File → Export** for the image, the interactive HTML, or the Python script.
-7. **File → Save session** writes a `.pviz` file that restores all of it later.
+   **Edit → Undo** reverts any filter, type, or cleaning change.
+6. Several charts worth keeping? **Pin to dashboard** (Ctrl+D) and look at them together on
+   the **Dashboard** tab.
+7. **File → Export** for the image, the interactive HTML, or the Python script — of the chart,
+   or of the whole dashboard.
+8. **File → Save session** writes a `.pviz` file that restores all of it later.
 
 ### Command line
 
@@ -206,6 +222,18 @@ plotaviz --input data.csv --chart box --x region --y revenue \
 
 `--json` makes `--describe` and `--recommend` machine-readable.
 
+```bash
+# Every CSV in a folder: one chart and one report per file
+plotaviz --input "data/*.csv" --output-dir out --export .png --report .html
+
+# A specific Excel sheet, or all sheets stacked with a 'sheet' column
+plotaviz --input book.xlsx --sheet Data --auto --export chart.pdf
+plotaviz --input book.xlsx --sheet '*' --report book.html
+
+# Your own chart rules instead of the shipped rules.yaml
+plotaviz --input data.csv --rules my_rules.yaml --recommend
+```
+
 ---
 
 ## How it decides
@@ -271,8 +299,10 @@ Both return a **structured chart spec (JSON), never executable code.** The spec
 against your real column names before anything renders; a hallucinated column is a clear error
 message, not a crash.
 
-**What gets sent:** column names, summary statistics, and up to 5 sample rows. Never the dataset.
-You are asked for consent before the first remote request.
+**What gets sent:** column names, summary statistics, and up to five example and most-frequent
+values per column — withheld entirely for identifier and free-text columns, which is where names
+and e-mail addresses live. Never the dataset. Nothing goes to a remote provider until you tick
+that provider's consent box in Settings; consent is per provider.
 
 **Where keys live:** your OS keychain, via `keyring`. Never a config file, never a session file,
 never the repo. Keys are redacted from logs and error messages.
@@ -285,8 +315,9 @@ ollama pull llama3.1
 ollama serve
 ```
 
-In **Settings → LLM provider**, choose **Ollama**. No API key, no network, no consent prompt —
-this is the right choice for sensitive data.
+In **Settings → LLM provider**, choose **Ollama**. With the default `localhost` host there is no
+API key, no network, and no consent prompt — this is the right choice for sensitive data. Point
+it at another machine and it is treated like any remote provider: consent is required.
 
 Hosted providers: Anthropic (Claude), OpenAI, and Google Gemini. Install with
 `pip install -e ".[llm]"`, then paste a key in Settings.
@@ -295,18 +326,35 @@ Hosted providers: Anthropic (Claude), OpenAI, and Google Gemini. Install with
 
 ## Features
 
-- **Loading** — CSV, TSV, Excel, JSON, NDJSON, Parquet. Delimiter and encoding sniffing. Files
-  over a configurable threshold read through polars and profile on a sample.
+- **Loading** — CSV, TSV, Excel, JSON, NDJSON, Parquet. Delimiter, encoding, and European
+  decimal-comma detection, with manual import options (delimiter, encoding, decimal/thousands
+  marks, header line, Excel sheet). Files over a configurable threshold read through polars and
+  profile on a sample.
 - **Cleaning** — an ordered, replayable step list: name normalization, type coercion,
   missing-value strategies, IQR/z-score outlier flagging (non-destructive by default),
-  deduplication. Every step reports what it changed.
+  deduplication. Every step reports what it changed; settings and individual steps are editable
+  in the Cleaning tab, and every change can be undone.
 - **Type override** — inferred roles shown and editable before analysis proceeds.
 - **Filters** — range widgets, multi-selects, date ranges, and a raw pandas-query bar. Filters are
   pipeline steps, so they serialize into sessions and generated code for free.
 - **Charts** — 15 types, interactive via Plotly in a WebEngine view, with a matplotlib canvas
   fallback when WebEngine is unavailable.
-- **Export** — PNG, SVG, PDF, JPEG, WebP at up to 1200 DPI; self-contained interactive HTML; or a
-  standalone Python script in matplotlib or Plotly flavour.
+- **Panels** — split a chart into one small chart per category (Customize → Panels by), when a
+  colour split has too many series to tell apart.
+- **Join** — File → Join another dataset adds columns from a second file matched on key columns,
+  with a match-rate preview; undoable, saved in sessions, replayed by generated code.
+- **Report** — File → Export report writes one offline HTML page or a PDF: summary, cleaning,
+  every column, missing values, correlations, recommended charts with their reasoning, and the
+  dashboard. Headless: `plotaviz --input data.csv --report report.html`.
+- **Preview** — sort by any column (the whole dataset, not just the visible rows), search all
+  rows, and see a column's statistics.
+- **Appearance** — View → Theme: follow the system, light, or dark. Exports always stay light.
+- **Dashboard** — pin several charts (Ctrl+D) to see a dataset's trend, breakdown, and
+  distribution side by side. Tiles redraw from the current filters and cleaning, are saved in the
+  session, and export together as one PNG/SVG/PDF grid, one offline HTML page, or one script.
+  Headless: `plotaviz --input data.csv --dashboard 4 --export overview.pdf`.
+- **Export** — PNG, SVG, PDF, JPEG, WebP at up to 1200 DPI; self-contained interactive HTML
+  (plotly.js embedded, works offline); or a standalone Python script in matplotlib or Plotly flavour.
 - **Sessions** — `.pviz` files store the recipe, not the data, and warn if the source changed.
 - **Guardrails** — automatic downsampling with a visible notice, top-N category capping,
   virtualized preview, debounced re-render.
@@ -387,9 +435,8 @@ See [CONTRIBUTING.md](CONTRIBUTING.md) for adding selector rules and LLM provide
 
 ## Roadmap
 
-Designed for, not built yet: auto EDA report export (HTML/PDF), a multi-chart dashboard grid, a
-chart customization panel (titles, palettes, axis scales), undo/redo across the step stack,
-colorblind-safe palettes and OS-following dark mode, and joining two datasets on a key.
+Everything from the original roadmap is built — see [docs/AUDIT.md](docs/AUDIT.md) for what was
+done and what is deliberately left open.
 
 Explicitly out of scope for v1: plugin architecture, localization, real-time streaming, and live
 database connectors.

{plotaviz-0.1.0 → plotaviz-0.2.0}/README.md

@@ -1,6 +1,7 @@
 # PlotaViz
 
 [![CI](https://github.com/saadibnainan/plotaviz/actions/workflows/ci.yml/badge.svg)](https://github.com/saadibnainan/plotaviz/actions/workflows/ci.yml)
+[![PyPI](https://img.shields.io/pypi/v/plotaviz)](https://pypi.org/project/plotaviz/)
 [![Python 3.11+](https://img.shields.io/badge/python-3.11%2B-blue)](https://www.python.org/downloads/)
 [![License: MIT](https://img.shields.io/badge/license-MIT-green)](LICENSE)
 [![Qt: LGPLv3](https://img.shields.io/badge/Qt-LGPLv3-orange)](THIRD_PARTY_LICENSES.md)
@@ -47,7 +48,14 @@ between them in one click. The parts that usually go wrong are treated as first-
 
 Python 3.11 or newer.
 
-### From source (all platforms)
+### From PyPI (all platforms)
+
+```bash
+pip install plotaviz
+plotaviz
+```
+
+### From source (contributing / latest main)
 
 ```bash
 git clone https://github.com/saadibnainan/plotaviz.git
@@ -64,7 +72,7 @@ Apple Silicon and Intel are both supported.
 
 ```bash
 brew install python@3.11          # if you need it
-pip install -e .
+pip install plotaviz
 ```
 
 A prebuilt `.app` is attached to each [release](https://github.com/saadibnainan/plotaviz/releases).
@@ -92,7 +100,7 @@ sudo dnf install python3-devel libglvnd-egl xcb-util-cursor libxkbcommon-x11
 sudo pacman -S python libxkbcommon-x11 xcb-util-cursor
 ```
 
-Then either `pip install -e .`, or grab the **AppImage** from the releases page:
+Then either `pip install plotaviz`, or grab the **AppImage** from the releases page:
 
 ```bash
 chmod +x PlotaViz-*.AppImage
@@ -105,9 +113,9 @@ To build one: `make build-appimage`. Recipes for `.deb`, `.rpm`, and an **AUR PK
 ### Optional extras
 
 ```bash
-pip install -e ".[llm]"     # Anthropic / OpenAI / Gemini / Ollama providers
-pip install -e ".[fast]"    # pyqtgraph fast path for very large numeric plots
-pip install -e ".[dev]"     # tests, ruff, mypy, pre-commit
+pip install "plotaviz[llm]"     # Anthropic / OpenAI / Gemini / Ollama providers
+pip install "plotaviz[fast]"    # pyqtgraph fast path for very large numeric plots
+pip install -e ".[dev]"         # tests, ruff, mypy, pre-commit — source checkout only
 ```
 
 ---
@@ -119,13 +127,21 @@ plotaviz                                  # open the app
 plotaviz samples/sales_timeseries.csv     # open the app with a file
 ```
 
-1. Drag a dataset in, or use **File → Open**.
-2. Read the **cleaning report** — what was renamed, filled, deduplicated, flagged.
+1. Drag a dataset in, or use **File → Open**. For a semicolon file, a non-UTF-8 export, a title
+   row above the header, or a specific Excel sheet, use **File → Open with options…**.
+2. Read the **cleaning report** — what was renamed, deduplicated, flagged. Change how rows are
+   cleaned in the **Cleaning** tab; missing values are left alone unless you choose otherwise.
 3. Check the **Types** tab and correct anything inference got wrong.
-4. Pick a chart from **Charts**. The reasoning for each is right there.
+4. Pick a chart from **Charts** — the reasoning for each is right there — or build one in the
+   chart row above it (works offline). Adjust title, labels, scales, and colours under
+   **Customize**.
 5. Narrow the data in **Filters** — per-column widgets, or a raw pandas query.
-6. **File → Export** for the image, the interactive HTML, or the Python script.
-7. **File → Save session** writes a `.pviz` file that restores all of it later.
+   **Edit → Undo** reverts any filter, type, or cleaning change.
+6. Several charts worth keeping? **Pin to dashboard** (Ctrl+D) and look at them together on
+   the **Dashboard** tab.
+7. **File → Export** for the image, the interactive HTML, or the Python script — of the chart,
+   or of the whole dashboard.
+8. **File → Save session** writes a `.pviz` file that restores all of it later.
 
 ### Command line
 
@@ -152,6 +168,18 @@ plotaviz --input data.csv --chart box --x region --y revenue \
 
 `--json` makes `--describe` and `--recommend` machine-readable.
 
+```bash
+# Every CSV in a folder: one chart and one report per file
+plotaviz --input "data/*.csv" --output-dir out --export .png --report .html
+
+# A specific Excel sheet, or all sheets stacked with a 'sheet' column
+plotaviz --input book.xlsx --sheet Data --auto --export chart.pdf
+plotaviz --input book.xlsx --sheet '*' --report book.html
+
+# Your own chart rules instead of the shipped rules.yaml
+plotaviz --input data.csv --rules my_rules.yaml --recommend
+```
+
 ---
 
 ## How it decides
@@ -217,8 +245,10 @@ Both return a **structured chart spec (JSON), never executable code.** The spec
 against your real column names before anything renders; a hallucinated column is a clear error
 message, not a crash.
 
-**What gets sent:** column names, summary statistics, and up to 5 sample rows. Never the dataset.
-You are asked for consent before the first remote request.
+**What gets sent:** column names, summary statistics, and up to five example and most-frequent
+values per column — withheld entirely for identifier and free-text columns, which is where names
+and e-mail addresses live. Never the dataset. Nothing goes to a remote provider until you tick
+that provider's consent box in Settings; consent is per provider.
 
 **Where keys live:** your OS keychain, via `keyring`. Never a config file, never a session file,
 never the repo. Keys are redacted from logs and error messages.
@@ -231,8 +261,9 @@ ollama pull llama3.1
 ollama serve
 ```
 
-In **Settings → LLM provider**, choose **Ollama**. No API key, no network, no consent prompt —
-this is the right choice for sensitive data.
+In **Settings → LLM provider**, choose **Ollama**. With the default `localhost` host there is no
+API key, no network, and no consent prompt — this is the right choice for sensitive data. Point
+it at another machine and it is treated like any remote provider: consent is required.
 
 Hosted providers: Anthropic (Claude), OpenAI, and Google Gemini. Install with
 `pip install -e ".[llm]"`, then paste a key in Settings.
@@ -241,18 +272,35 @@ Hosted providers: Anthropic (Claude), OpenAI, and Google Gemini. Install with
 
 ## Features
 
-- **Loading** — CSV, TSV, Excel, JSON, NDJSON, Parquet. Delimiter and encoding sniffing. Files
-  over a configurable threshold read through polars and profile on a sample.
+- **Loading** — CSV, TSV, Excel, JSON, NDJSON, Parquet. Delimiter, encoding, and European
+  decimal-comma detection, with manual import options (delimiter, encoding, decimal/thousands
+  marks, header line, Excel sheet). Files over a configurable threshold read through polars and
+  profile on a sample.
 - **Cleaning** — an ordered, replayable step list: name normalization, type coercion,
   missing-value strategies, IQR/z-score outlier flagging (non-destructive by default),
-  deduplication. Every step reports what it changed.
+  deduplication. Every step reports what it changed; settings and individual steps are editable
+  in the Cleaning tab, and every change can be undone.
 - **Type override** — inferred roles shown and editable before analysis proceeds.
 - **Filters** — range widgets, multi-selects, date ranges, and a raw pandas-query bar. Filters are
   pipeline steps, so they serialize into sessions and generated code for free.
 - **Charts** — 15 types, interactive via Plotly in a WebEngine view, with a matplotlib canvas
   fallback when WebEngine is unavailable.
-- **Export** — PNG, SVG, PDF, JPEG, WebP at up to 1200 DPI; self-contained interactive HTML; or a
-  standalone Python script in matplotlib or Plotly flavour.
+- **Panels** — split a chart into one small chart per category (Customize → Panels by), when a
+  colour split has too many series to tell apart.
+- **Join** — File → Join another dataset adds columns from a second file matched on key columns,
+  with a match-rate preview; undoable, saved in sessions, replayed by generated code.
+- **Report** — File → Export report writes one offline HTML page or a PDF: summary, cleaning,
+  every column, missing values, correlations, recommended charts with their reasoning, and the
+  dashboard. Headless: `plotaviz --input data.csv --report report.html`.
+- **Preview** — sort by any column (the whole dataset, not just the visible rows), search all
+  rows, and see a column's statistics.
+- **Appearance** — View → Theme: follow the system, light, or dark. Exports always stay light.
+- **Dashboard** — pin several charts (Ctrl+D) to see a dataset's trend, breakdown, and
+  distribution side by side. Tiles redraw from the current filters and cleaning, are saved in the
+  session, and export together as one PNG/SVG/PDF grid, one offline HTML page, or one script.
+  Headless: `plotaviz --input data.csv --dashboard 4 --export overview.pdf`.
+- **Export** — PNG, SVG, PDF, JPEG, WebP at up to 1200 DPI; self-contained interactive HTML
+  (plotly.js embedded, works offline); or a standalone Python script in matplotlib or Plotly flavour.
 - **Sessions** — `.pviz` files store the recipe, not the data, and warn if the source changed.
 - **Guardrails** — automatic downsampling with a visible notice, top-N category capping,
   virtualized preview, debounced re-render.
@@ -333,9 +381,8 @@ See [CONTRIBUTING.md](CONTRIBUTING.md) for adding selector rules and LLM provide
 
 ## Roadmap
 
-Designed for, not built yet: auto EDA report export (HTML/PDF), a multi-chart dashboard grid, a
-chart customization panel (titles, palettes, axis scales), undo/redo across the step stack,
-colorblind-safe palettes and OS-following dark mode, and joining two datasets on a key.
+Everything from the original roadmap is built — see [docs/AUDIT.md](docs/AUDIT.md) for what was
+done and what is deliberately left open.
 
 Explicitly out of scope for v1: plugin architecture, localization, real-time streaming, and live
 database connectors.

{plotaviz-0.1.0 → plotaviz-0.2.0}/plotaviz/__init__.py

@@ -13,7 +13,7 @@ The package is split into two halves that must not blur together:
 
 from __future__ import annotations
 
-__version__ = "0.1.0"
+__version__ = "0.2.0"
 __app_name__ = "PlotaViz"
 
 __all__ = ["__app_name__", "__version__"]