How to nest formattable values¤
When your type holds other formattable values, a spec aimed at those values
has to reach them. It does that only if you forward **kw.
Accept **kw and forward it¤
The engine passes every axis your type does not itself consume. Forward that
dictionary to each nested pparts call, and wrap each nested result in a
PGroup so its fragments stay grouped:
>>> import dataclasses
>>> from wadler_lindig_parts import PGroup, PPart, pparts, pspec
>>> from wadler_lindig_parts.examples import Term
>>> @dataclasses.dataclass
... class Interval:
... lo: Term
... hi: Term
>>> @pparts.dispatch
... def _(obj: Interval, /, *, markup="text", **kw):
... return (
... PPart("open", "[", "sep"),
... PGroup("lo", pparts(obj.lo, markup=markup, **kw)),
... PPart("comma", ", ", "sep"),
... PGroup("hi", pparts(obj.hi, markup=markup, **kw)),
... PPart("close", ")", "sep"),
... )
>>> iv = Interval(Term(0.0, "x"), Term(1.0, "x"))
>>> pspec(iv, "mul")
'[0.0 * x, 1.0 * x)'
Interval has never heard of the value axis, but a format spec still
reaches the coefficients inside both Terms, because **kw was forwarded:
>>> pspec(iv, ".2f")
'[0.00 x, 1.00 x)'
Markup reaches them the same way:
>>> pspec(iv, "html")
'[<span>0.0</span> <span>x</span>, <span>1.0</span> <span>x</span>)'
If you drop **kw¤
The nested values fall back to their defaults. Nothing raises, because the
fragments are still well formed — pspec(iv, ".2f") would simply print
[0.0 x, 1.0 x) and nobody would see a problem. Assert on a nested spec in
your own tests; a reviewer will not catch it.
If a nested field is not formattable¤
It degrades to str(obj) like any unregistered type. Wrap it in PPart with
a role of its own if you want a markup to be able to style it.
See also¤
- How to make a type formattable — the single-level case.
- Axes, layouts and markups — what is being forwarded.
- API —
PGroup,pparts.