Consider `sum()` interface

[hadley]

Unfortunately I can no longer find the bluesky post that pointed this out, but:

(ggplot(mpg, aes(x="displ", y="hwy", colour="class"))
 + geom_point()
 + geom_smooth(method="lm")
 + facet_wrap("drv")
 + theme_minimal()
 + labs(title="Engine Displacement vs Highway MPG"))

Is not much better than

sum(
  ggplot(mpg, aes(x="displ", y="hwy", colour="class")),
  geom_point(),
  geom_smooth(method="lm"),
  facet_wrap("drv"),
  theme_minimal(),
  labs(title="Engine Displacement vs Highway MPG"),
)

So it might be worth considering if overloading + is even worth it.

You can see a bit of historical context at https://github.com/hadley/ggplot1 — ggplot2 only used the plus because I couldn't see a better way to do function composition.

[hadley]

And of course immediately after writing that I found it: https://bsky.app/profile/mikedecr.computer/post/3mklzhr3ux22d

[jefferyUstc]

Thanks @hadley,

Now, we document the four function-style composition idioms in the README that already work out of the box (sum(parts, start=...), sum([iterable]), functools.reduce, and plot + [list]).

One small caveat is that: Python's built-in sum is sum(iterable, /, start=0), so the literal sum(a, b, c, d) from the screenshot would raise TypeError — components need an iterable wrapper, or the 2-line def fnplot(data, *parts): return sum(parts, start=ggplot(data)) helper (credit @mikedecr.computer's Bluesky post).

Suggestions on naming, ergonomics, or further idioms most welcome.