Jarvis-PLOT

Jarvis-PLOT is a lightweight, Python/Matplotlib-based plotting framework developed for Jarvis-HEP,
but it can also be used as a standalone scientific plotting tool.

It provides a simple command-line interface (CLI) to generate publication-quality figures from YAML configuration files, with most layout and style decisions handled by predefined profiles and style cards.

Current source candidate: 2.1.11 (jplot -v reports the installed version). Release history: CHANGELOG.md. Design documents and specifications: Jarvis-Books Docs/Jarvis-PLOT/.

Installation

pip install JarvisPLOT

The PyPI / distribution name is JarvisPLOT (normalized wheel name: jarvisplot). The product name remains Jarvis-PLOT. The Python import package and entrypoint remain unchanged:

import jarvisplot

jplot path/to/config.yaml

If you have an older environment that still uses a historical package name, replace:

pip uninstall jarvisplot Jarvis-PLOT JarvisPLOT
pip install JarvisPLOT

Command-Line Usage

Human help uses the same fixed-column Rich panels as Jarvis-HEP V2 / Portal; jplot -h prints the verb map, and jplot <verb> -h opens one verb.

Render with a bare path — there is no jplot run:

jplot path/to/config.yaml
jplot path/to/config.yaml --rebuild-cache
jplot path/to/config.yaml --print

Everything else is discovery and judgement, in three groups.

Discover — find out what is legal before writing anything:

jplot data describe samples.csv     # columns, dtypes, ranges (HDF5 tree too)
jplot data head samples.csv -n 5    # real sample rows
jplot data eval 'exp(LogL)' --data samples.csv    # sandbox an expr
jplot data suggest-axes samples.csv # scale / lim hints for frame.ax
jplot cap                           # section index
jplot cap methods                   # all | methods | transforms | types
jplot cap styles                    #   | styles | cmaps | funcs | cli
jplot man                           # manual index
jplot man workflow                  # topic page
jplot man scatter                   # one method's contract
jplot explain JP-COL-001            # a diagnostic code

Draft & edit — scaffold or address a YAML by name:

jplot template list                 # template kinds
jplot template show posterior_2d    # one template + its slots
jplot suggest --data samples.csv --kind posterior_2d
jplot config paths plot.yaml        # named addresses in this file
jplot config get plot.yaml 'Figures[p].style'
jplot config set plot.yaml 'Figures[p].frame.ax.xlim' '[0, 10]' --write
jplot config expand plot.yaml --figure p --write   # type: → layers

Judge — check without rendering:

jplot validate plot.yaml            # schema/contracts + source paths/column headers; no transforms
jplot dryrun plot.yaml              # load data + transforms, row ledger, JP-VIZ health
jplot doctor plot.yaml              # validate + dryrun in one pass

Every one of these takes --json.

Default fonts

The bundled font card recommends Helvetica for sans text and STIXGeneral for serif text, with STIX MathText formulas. Normal YAML rendering applies the card automatically; explicit YAML fonts and sizes take precedence. When Helvetica is absent, Jarvis-PLOT selects the first installed similar family: Arial, Liberation Sans, Nimbus Sans, then DejaVu Sans. check and resolve report the actual font and substitution reason.

jplot fonts                         # role card: title, labels, ticks, legend, text, annotations, caption
jplot fonts show title --json       # current defaults for one role
jplot fonts show --recommended --json
jplot fonts list --filter Helvetica --json
jplot fonts list --refresh --json   # rescan after installing fonts
jplot fonts resolve Helvetica --weight bold --json
jplot fonts check --json            # actual installed faces used by the card
jplot fonts set roles.title.family sans
jplot fonts set roles.legend.family serif
jplot fonts set roles.title.size_pt 9
jplot fonts set roles.title.size_pt null  # inherit the plot style's size
jplot fonts reset roles.title.size_pt
jplot fonts reset                   # remove all personal overrides
jplot man fonts --json

Personal overrides persist in ~/.jarvis/plot-fonts.json (JARVIS_HOME overrides the directory). Queries never create that file. size_pt: null preserves the plot style's canvas-specific size; recommended_size_pt is drafting advice. Font discovery uses Matplotlib's cached system and bundled font index, while TTC face selection preserves the requested weight and slant. $...$ uses MathText; an external LaTeX compiler has its own font configuration.

Design-Reference Overlay

Set debug: true on a figure and Jarvis-PLOT draws a dimension overlay on top of it: every axes box labelled with its rect and its size in centimetres, page margins, the colorbar gap, and an "axes layout" panel. It is meant for checking a layout, so a figure with layers: [] is a legitimate use.

Figures:
  - name: layout_check
    debug: true
    style: [a4paper_2x1, rectcmap]
    layers: []

The style card's Debug block decides what the annotations look like and which of them appear; a per-figure debug: mapping overrides the card for that figure only. See jplot man debug-overlay and STYLE_SCHEMA.md.

Figures[].debug is not the jplot --debug flag — the flag only raises the log level.

For Coding Agents

Jarvis-PLOT is built so a coding agent can drive it without ever seeing a rendered figure. The rule is the agent writes the YAML itself; the CLI's job is to say what is legal and to judge what was written. There is no verb that generates a finished config for you.

Every discovery and judging verb prints one JSON envelope on stdout, with humans and logs on stderr:

{"api_version": 1, "kind": "cap.methods", "ok": true,
 "data": {...}, "diagnostics": [], "error": null}

--json turns it on, and it is already the default when stdout is not a TTY — piping is enough.

The loop jplot man prints is:

Before drafting, inspect jplot fonts show --json and jplot fonts check --json for the typography defaults and actual installed font choices.

1. jplot data describe <file> --json — column names from the real file. Never invent one; a wrong name is a hard error, not a warning.
2. jplot cap methods|styles|cmaps|funcs|transforms|types --json — the legal strings, read live from the registries rather than from documentation.
3. Edit the YAML in an editor (or jplot config set … --write). Reach for a type: macro first, drop to explicit layers when the macro cannot say it.
4. jplot doctor <file> --json — validate + dryrun in one pass. Diagnostics carry JP-* codes; jplot explain JP-XXX-NNN expands any of them.
5. jplot <file> — render, only when the figure itself is needed.

jplot man --json is the machine-readable manual: a topic index, plus live pages for every registered layer method and transform step (jplot man scatter, jplot man transform.profile, jplot man PDF1D). Method contracts come from the same registry jplot cap methods reads, so a man page cannot drift from the code.

For compact legends and keys composed from real layers, use jplot man legend --json. Declare legend: {object, label, role} on each drawing layer; components with the same object regroup automatically using card templates such as band_line, line_errorbar and error_bounds. The role is optional for a single-component object. frame.<axes>.legend owns card selection, order, ncols, position and spacing. JSON cards provide reusable defaults, and YAML can override them. See the legend contract and the self-contained example.

For continuous one-dimensional distributions from raw samples, use PDF1D (CDF plus PDF) or CDF1D in a data transform. coordinates.x owns expr, lim, and grid; optional repeat computes independent batch means and standard deviations. See the contract and the runnable example. Reconstruction defaults to adaptive true CDF anchors and monotone_c2 interpolation (C2 CDF / C1 PDF). Set anchors: all and interpolation: pchip to reproduce the original all-node reconstruction.

Two things worth knowing before trusting a config:

• jplot validate checks shape and contracts and never touches the data; jplot dryrun loads the data and reports the row ledger and JP-VIZ render health. Passing the first says nothing about the second.
• AGENT_DATA_API.md describes a different, still-frozen bridge (a numeric digest channel for Jarvis-Agent). The surface above is what ships today.

Flowchart Rendering

Render a Jarvis-HEP flowchart scene JSON (produced by HEP scan tooling — this is not general plot YAML):

jplot flowchart path/to/scene.json
# optional output path:
jplot flowchart path/to/scene.json -o path/to/out.png

Library form:

from jarvisplot import render_flowchart, render_flowchart_file

Project Workdir and Cache

• You can set project.workdir in YAML.
• If output.dir is omitted, Jarvis-PLOT defaults to <workdir>/plots/.
• Data cache is stored in <workdir>/.cache/.
• --print accepts the current settings and renders final figures without the logo. It announces that, after all enabled figures are saved successfully, cache payloads used or created by this run will be deleted (pipeline data and metadata, named/shared data, summaries, and HDF5 materialization slots). Rendering failures or a run with no enabled figures keep the cache. Unrelated cache entries and the lightweight fingerprint manifest are retained; later renders rebuild deleted entries as needed.
• Profiling pipelines are prebuilt once and reused from cache when source fingerprint and profile settings are unchanged.
• Profiling uses a fast two-stage grid reduction (pregrid + render bin) for large datasets.

Plot caches are automatically registered in ~/.jarvis/plot.json (or $JARVIS_HOME/plot.json). The registry groups caches by workdir and records associated YAML paths and last use; payloads stay in their original directories. This includes caches created by data describe and default dryrun --with-data exports. Sizes are calculated when listing, rather than stored in the registry.

jplot cache list
jplot cache scan ~/Jarvis-Workshop               # register existing plot caches
jplot cache clean --workdir /path/to/project    # this directory and all nested projects; repeatable
jplot cache clean --all --older-than 30         # unused for at least 30 days
jplot cache clean --all --dry-run               # preview; no deletion/index changes
jplot cache clean --all                         # clean all registered inactive caches
jplot cache -h
jplot man cache

--workdir PATH selects registered caches in PATH itself and all of its subdirectories, even when PATH has no cache of its own. Repeat it to select multiple directory trees; overlapping selections are deduplicated. Use cache scan PATH first for old caches that have not been registered.

All cache subcommands accept --json and use the existing CLI envelope and exit codes (0 success/partial, 1 failure, 2 usage). Non-TTY stdout defaults to JSON; human results go to stderr. Active caches are skipped (ok: null when the remaining operation succeeds), and records for missing paths or unmarked .cache directories containing only empty Jarvis-PLOT component directories are pruned during cleanup. Only marked Jarvis-PLOT components are removed; source data, plots, and unrelated .cache content are retained. Failed deletions retain their registry entries; nonempty unmarked directories remain invalid and are not deleted. Scanning adopts legacy caches only when the four pipeline directories and a compatible manifest.json identify them; old unmarked twin-only directories need a new dryrun --with-data run to register. An explicitly supplied --out-dir is an export directory and is not registered. Concurrent registry writes use a lock and atomic replacement. Registration failures warn without preventing rendering.

Example: Ternary Layouts

jplot ./Example/gambit_2x1.yaml

This layout reference uses bundled style cards and renders without external data.

Example: Dynesty Runplot

Jarvis-PLOT includes a reusable dynesty runplot format. With a dataset named dynesty, the figure can use the built-in card without writing axes or layer details:

DataSet:
- name: dynesty
  path: path/to/dynesty_result.csv
  type: csv

Figures:
- name: dynesty_logL_vs_logX
  enable: true
  style:
  - a4paper_2x1
  - dynesty_runplot

See DYNESTY_RUNPLOT.md for the default axes, KDE, scatter overlay, and evidence summary behavior.

Notes

• Figures are saved automatically to the output paths defined in the YAML configuration.
• Common output formats include PNG and PDF (backend-dependent).
• Saved figures include file metadata such as Creator: Jarvis-PLOT, powered by Jarvis-HEP and Jarvis-PLOT version: X.Y.Z; the PNG Description also includes both fields for macOS Finder compatibility.
• Jarvis-PLOT works in headless environments (SSH, batch jobs) without any GUI backend.

Requirements

Python

• Python ≥ 3.10 (tested on 3.10–3.13)

Required Packages

• numpy
• pandas
• polars
• matplotlib
• pyyaml
• jsonschema
• scipy — numerical utilities
• h5py — required for loading HDF5 data files
• shapely
• sympy
• loguru
• deepmerge
• ruamel.yaml>=0.18 — round-trip YAML editing for jplot config set
• rich — the fixed-column help and manual panels
• Jarvis-Operas>=1.1.4

Github Page

https://github.com/Pengxuan-Zhu-Phys/Jarvis-PLOT

Documentation

https://pengxuan-zhu-phys.github.io/Jarvis-Docs/

Project Docs

This repository keeps only README.md and CHANGELOG.md. Design documents, specifications and the roadmap live in Jarvis-Books:

• Doc index
• JARVIS_PLOT_CONTEXT.md - project boundary; read before changing code
• CODE_MAP_JARVIS_PLOT.md - code owner map
• JARVIS_PLOT_FRAMEWORK_LOGIC.md - runtime execution contract
• IMPLEMENTATION_ROADMAP.md - open work
• RELEASING.md - versioning, tools/check_release.py and CI verification
• STYLE_SCHEMA.md - style card contract, including the Debug overlay block
• DEVELOPER_RULES.md - pipeline and cache rules

Read the context docs and roadmap before changing parsing, transforms, rendering, or layout-related behavior.

License

MIT License

Create a continuous forest figure with type: forest, a generic row table, optional points and multiple supplied intervals per row. Native YAML supports class-specific interval lanes, CSV value annotations, explicit x.ticks, opt-in MathText, header notes and a shared-format legend. The default Swiss-column card uses 82.5 mm width and 5.5 mm row pitch; every row shares one numerical x domain. See the forest contract, native examples, or jplot man forest --json.

AI agents: start with jplot man agent-readme --json for the official-package YAML workflow and missing-feature reporting instructions. Feature requests include the installed version from jplot --version; the CLI does not send email automatically.

Releasing

jprel is a maintainer-local script, excluded from Git and distribution archives.

Record changes in CHANGELOG.md, then run from this checkout (replace the example version with the intended release):

./jprel 2.1.11 --dry
./jprel 2.1.11

--dry performs read-only source and release-note checks. The normal command requires a clean Git checkout, builds and checks the wheel and sdist, uploads to PyPI, refreshes the local version with python3 -m pip install -e ., and verifies the installed version. Only then does it commit the version, push the vX.Y.Z tag, and create a GitHub release with changelog notes. Configure Twine and gh authentication before publishing.