# The book's example data — provenance and terms

These CSVs are the tables every chapter of *gog: A Grammar of Graphics* draws
from. They are published so that a reader can run the manual's examples in R,
Python, Julia or JavaScript, rather than only read them.

They are built by `book/R/make-data.R` and read by `book/R/data.R`. The ruling
that put them here is spec §20, "The cast is fetched, not shipped".

Three tiers, and they are not under one license.

## Original to this book — Apache License 2.0

Twenty-one frames are written out as literals or generated from a fixed seed by
this project's author, and carry the same license as the rest of the code:

`actuals` · `cashflow` · `census` · `commutes` · `day_cycle` · `forecast` ·
`gdp_threshold` · `life_bands` · `listening` · `medals` · `milestones` ·
`quarterly` · `recessions` · `ripples` · `score_band` · `sessions` ·
`six_weeks` · `speed_target` · `spending` · `target_band` · `tide` ·
`thermals` · `thermal_marks` · `winds`

Illustrative rather than authoritative. `census` is two plausible city age
profiles, not a census; `medals` is a medal table's shape, not a record of any
particular games. Do not cite them as data about the world.

## Derived from gapminder — CC0 1.0 (public domain)

Eleven frames are cuts of the `gapminder` R package's table, with three columns
renamed for readability (`gdpPercap` → `gdp`, `lifeExp` → `life`,
`pop` → `population`):

`gm_all` · `gapminder_2007` · `gapminder_asia` · `gm_continents` · `gm_eras` ·
`gm_europe` · `gdp_rug` · `life_rug`

The `gapminder` package is released under **CC0 1.0**, a public domain
dedication, so no permission or attribution is required. It is credited anyway:
the package is by Jennifer Bryan, and the underlying figures are the Gapminder
Foundation's — <https://www.gapminder.org/data/>.

## Derived from R's `datasets` package — GPL-2 | GPL-3

Three frames are reshaped from tables that ship with R itself:

| Frame | From | Reshaping |
|---|---|---|
| `iris_flowers` | `datasets::iris` | columns renamed to bare words |
| `maunga_whau` | `datasets::volcano` | matrix unrolled to long form, every second row and column |
| `quakes_fiji` | `datasets::quakes` | depth negated to an elevation, and cut into 90 km bands |

**These are GPL, and that is why nothing here is shipped inside a package.** R's
`datasets` is licensed GPL-2 | GPL-3, which is copyleft and cannot be absorbed
into an Apache-2.0 wheel or tarball. Hosting them beside the book is ordinary
distribution, which the GPL permits when its terms travel with the files — this
note is that. A reader who uses these three frames is using GPL data and should
treat it accordingly.

The underlying observations are old and public: Edgar Anderson's iris
measurements published by R. A. Fisher in 1936, a 1967 topographic survey of
Maungawhau in Auckland, and seismic events near Fiji. The GPL attaches to R's
compilation of them, not to the facts.

## Derived from Natural Earth — public domain

One frame carries the world's coastlines and borders:

| Frame | From | Reshaping |
|---|---|---|
| `world_borders` | Natural Earth `ne_110m_admin_0_countries` | polygons unrolled to one row per vertex, rings numbered into `piece`, simplified, Antarctica dropped |

Natural Earth places its data in the **public domain** and asks for no
permission and no attribution: *"All versions of Natural Earth raster and vector
map data found on this website are in the public domain. You may use the maps in
any manner, including modifying the content and design, electronic
dissemination, and offset printing."* It is credited here because saying where a
map came from is good practice, not because the terms require it.

The reshaping is what makes it a table rather than a shapefile, and it is the
same shape any boundary data has to arrive in: **one row per vertex**, with a
column naming the ring each vertex belongs to. A country is not always one
closed shape — islands are separate rings, and a country lying wholly inside
another is a ring of its own — so `piece` counts rings, not countries.

Simplified at roughly a third of a degree, which is invisible at the size the
book draws and is what keeps the file small. Antarctica is dropped because at
this resolution its coastline is a ragged strip cut off at the bottom of the
data rather than a shape, and a reader would fairly read that as a bug.