Undetermined pipelines and thunks

Placeholders turn an expression into a function. The same idea, applied to the seed (first stage) of a pipeline, turns the whole pipeline into a function.

Undetermined pipelines

Following magrittr, if any placeholders appear in the first step of a pipeline, the pipeline as a whole represents a function – an undetermined pipeline:

>>> second_largest_value = $ |> sorted($, reverse=True) |> $[1]
>>> [3, 8, 6, 5, 1] |> second_largest_value
6

Since each $ in the seed is a fresh argument, the first step can introduce more than one, yielding a multi-argument function. Use named placeholders when a value must be reused within the seed:

>>> hypotenuse = $a * $a + $b * $b |> $ ** 0.5
>>> hypotenuse(3, 4)
5.0

Note that once past the seed, $ has its usual per-stage meaning; it is only the placeholders in the seed that become the pipeline function’s parameters.

Thunks: the leading |>

Where $ |> ... defines a one-argument function, a leading |> defines a zero-argument one – a thunk that defers a pipeline until you call it:

>>> fetch = |> load() |> transform |> save
>>> fetch()   # runs load() |> transform |> save now

The pipeline starts from its own seed (the first stage), so no input is needed. A leading |> is simply sugar for lambda: – it works for any body, including a single stage (thunk = |> expensive()), and it re-runs on every call (it is deferral, not memoization). Inside the body, $ keeps its usual meaning in every stage after the seed:

>>> describe = |> 5 |> $ + 1 |> str
>>> describe()
'6'

Related

• Use placeholders ($) – how $ induces functions in the first place.

• Placeholder scope – which $ belongs to which function.