[warsaw]

Two fixes from another round of comparing the implementation with the PEP.

hug and Expand.ALWAYS. A hugged document was always wrapped in a group, so with ALWAYS a container nested in a hugging object stayed on one line: [list(range(8)), deque([[1, 2], [3]])] kept deque([[1, 2], [3]]) flat. The hugged document now drops its group below the top level, as other containers already do with ALWAYS, so every non-empty nested container expands as the PEP says.

compact in every mode. compact was used only with Expand.NEVER, and was silently ignored with FIT and ALWAYS, so Printer(compact=True) did nothing with the default expand. It now packs positional items in every mode, using PrettyPrinter's rule: an item whose one-line form fits shares lines with its packable neighbors, and any other item gets lines of its own.

[
    1, 2,
    [
        0, 1, 2, 3,
        4, 5, 6, 7,
        8, 9, 10,
        11,
    ],
    3, 4,
]

With ALWAYS, a non-empty container is never packed. Key/value and keyword items are not packed, as before.

This also fixes NEVER compact, where a broken container could start partway along a line ([1, 2, [0, 1, 2, 3,). Against PrettyPrinter(compact=True) on 3,000 random nested objects, all remaining differences are one boundary case: a packed line may now end at the last column of the width, where PrettyPrinter reserves room for a ', ' after the last item and breaks one item earlier. Our output never goes past the width where PrettyPrinter's doesn't.

Tests: test__pprint, test_pprint, test_collections, test_dataclasses, test_fstring, test_tstring and test_string pass.

🤖 Generated with Claude Code