gtviz

Publication-quality survey data visualization, refactored from the GivingPulse quarterly-report codebase into a clean, survey-agnostic library.

The brand theme: defaults that override matplotlib everywhere

The point of gtviz: call any chart with data only and get the published report look. gtviz.theme.use("report") applies the brand rcParams globally, and every function's styling defaults were audited line-by-line from the production report code. Highlights (full table: Design defaults & override policy):

All palette tokens live in gtviz.theme.palette — change a hex once, every chart and table follows.

API structure

gtviz
├── theme        use("report"|"publication", font=...), palette tokens
├── config       set_options(weight_col=, output_dir=, dpi=)
├── charts
│   ├── dots     dot_plot · grouped_dot_plot · trend_dot_plot
│   ├── bars     parallel_bars (baseline vs subgroups, ± diff labels)
│   ├── lines    rolling_trend · split_line_plot · annotated_event_plot
│   ├── civic    contribution_bars · range_dot_plot (dumbbell + benchmark)
│   │            · arrow_range_plot · nested_bars (layered subsets)
│   ├── stacked  stacked_bars (100% band bars) · banded_shares
│   ├── likert   likert_bars (diverging answer distributions)
│   ├── venn     venn · venn_from_counts
│   ├── heatmap  weighted_heatmap
│   ├── funnel   funnel · funnel_from_columns
│   ├── donut    donut
│   └── waffle   waffle  (extra: pip install gtviz[waffle])
├── tables       HtmlTable (publication CSS) · compare_periods · pivot_change_table
├── maps         choropleth_table (FIPS→hex) · scale_bar
├── stats        rolling_summary · period_change · subgroup_summary ·
│                chi_squared_matrix · build_filter · likert utils · aggs
├── io           save (png/svg/pdf/…) · figure_to_html · ReportBuilder (HTML+PDF)
└── pipeline     read_pipeline (Delta/Spark) · process() · sklearn-style steps
                 (ScoreBelonging · ScoreCivicIntent · AssignPew · AssignActivism ·
                  AssignCountyTypes · CivicQuartile)

Every chart accepts ax= and returns (fig, ax); nothing calls plt.show() for you.

Charts: dot plots (single, grouped, trend), parallel bar panels, rolling trend lines, venn diagrams (2/3 set, filtered or from pre-aggregated counts), weighted heatmaps, funnels, donuts, diverging Likert bars. Tables: publication CSS/HTML tables with zebra striping, high/low cell shading, multi-index rollups; period-over-period comparison tables. Maps: county/FIPS choropleth color tables (for SVG map filling) + scale-bar legends. Export: PNG, SVG, PDF, standalone HTML reports (figures embedded as SVG), suitable for websites or print reports.

import gtviz
gtviz.theme.use("report")

fig, ax = gtviz.dot_plot([62, 48, 31], ["Gave money", "Volunteered", "Gave items"],
                         error=[3, 3, 2], title="Generosity in Q2")
gtviz.io.save(fig, "generosity_q2", formats=("png", "svg", "pdf"))

Install

pip install gtviz            # core
pip install gtviz[waffle]    # + waffle charts

Docs

Full documentation, gallery, and migration guide from the original gp_reports repo: https://gtviz.readthedocs.io

Development

pip install -e .[dev,docs]
pytest                                  # unit tests; writes chart images to tests/output/
python examples/generate_gallery.py    # regenerate gallery images

CI runs lint + tests on every push and uploads rendered chart images as build artifacts for human review; a headless job compares rendered images against committed baselines in tests/baseline/. See .github/workflows/.