Tutorial: untangling nested calls

pipescript was written alongside Advent of Code 2025, where the input-parsing step is often a tangle of nested calls. This tutorial takes one such tangle and rewrites it as a readable pipeline. It assumes you have read Your first pipeline and Use placeholders ($).

The starting point

Consider computing, across a collection of arrays, the maximum finite absolute value (with a floor of 1.0):

import numpy as np

result = max(
    np.max(np.abs(array[np.isfinite(array)]), initial=1.0)
    for array in arrays
)

It is hard to read: you have to work inside-out, and it is genuinely ambiguous at a glance which call the initial=1.0 keyword belongs to.

Rewriting as a pipeline

The same computation reads top-to-bottom in pipescript. We map each array through a small inner pipeline, then take the max:

result = arrays |> map[$
  |> $array[np.isfinite($array)]
  |> np.abs
  |> np.max($, initial=1.0)
] |> max

Read it as a recipe:

arrays |> map[...] – run the bracketed pipeline once per array.
$array[np.isfinite($array)] – keep only the finite entries. The named placeholder $array is used twice, so both refer to the same array (an unnamed $ would introduce two separate arguments – see Placeholder scope).
|> np.abs – take absolute values.
|> np.max($, initial=1.0) – reduce to a maximum, with the floor made unambiguous by the explicit $.
|> max – the maximum across all arrays.

Why this is nicer in a notebook

Beyond readability, the pipeline form lets you verify each step interactively: run arrays |> map[$ |> $array[np.isfinite($array)]] |> list on its own to confirm the filtering, then append the next stage, and so on – the incremental workflow from Your first pipeline.

More examples

The author’s full set of AoC 2025 solutions – many built around pipescript input parsing – lives at https://github.com/smacke/aoc2025. The day 6 solution pushes the syntax to its limits (optimized for pipescript, not readability). See also Example gallery.