Feature or enhancement

Proposal:

Lib/configparser.py and Lib/test/test_configparser.py still carry a fair
amount of pre-f-string formatting (% formatting and str.format() calls) plus
a handful of constructs that have since acquired simpler spellings. Linting the
two modules with Ruff's current default rule set reports 72 violations:

35  UP032   f-string
17  UP031   printf-string-formatting
 3  UP030   format-literals
 3  RUF012  mutable-class-default
 2  SIM102  collapsible-if
 2  C414    unnecessary-double-cast-or-process
 1  C408    unnecessary-collection-call
 1  C405    unnecessary-literal-set
 1  C901    complex-structure
 1  F841    unused-variable
 1  I001    unsorted-imports
 1  ISC004  implicit-string-concatenation-in-collection-literal
 1  RUF019  unnecessary-key-check
 1  RUF022  unsorted-dunder-all
 1  RUF059  unused-unpacked-variable
 1  SIM117  multiple-with-statements

None of these are bugs, and CPython's own Ruff configuration only selects F401
and F811 for Lib/, so nothing here is failing today. The proposal is simply
to modernize the two modules: convert the %/.format() call sites to
f-strings, and apply the small set of simplifications above.

Two of the reports need more than a mechanical fix:

• C901 on ExtendedInterpolation._interpolate_some is a complexity metric,
  not a defect. Splitting that parser loop up would be a much larger and
  riskier change, so I propose an explicit # noqa: C901 # gh-159686. The
  issue reference rides along in the directive rather than in a comment above
  it, so the suppression carries its own trail without adding prose to the
  method; configparser: ExtendedInterpolation._interpolate_some exceeds common complexity limits #159686 records why the complexity is inherent to the
  interpolation grammar and what a refactoring might look like.

• RUF012 wants the mutable class attributes annotated ClassVar, and the
  obvious spelling of that would import typing at runtime, which this module
  deliberately avoids (cf. the gh-117703 note above the imports). Adding
  from __future__ import annotations and importing ClassVar only under
  TYPE_CHECKING keeps typing out of the import graph, as tomllib and
  importlib.metadata already do. -X importtime confirms typing is still
  not imported. A side benefit is that Iterable['ParsingError'] can lose its
  quotes (UP037).

  This does make TYPE_CHECKING a module-level name, so it needs adding to
  not_exported in the check__all__ test.

Motivation

Two reasons:

1. These are the usual readability wins. The error paths in particular read
   better as f-strings, e.g.

   -        Error.__init__(self, 'No section: %r' % (section,))
   +        Error.__init__(self, f'No section: {section!r}')

2. configparser is backported to PyPI as
   backports.configparser, whose
   module is merged from CPython periodically. That project lints with the
   default rule set and currently has to carry a per-file exemption for the
   CPython-derived files (jaraco/configparser#73). Landing the changes here
   lets the backport drop the exemption and keeps the two copies converging
   rather than diverging.

I have this working locally as four reviewable commits (the f-string
conversion, the simplifications, the handful of % sites Ruff will not
autofix, and the import/__all__/ClassVar changes above), and I am happy to
put up the PR. With all of it applied, both CPython's own Ruff configuration
and the default rule set report no findings for either file, and
test_configparser, test_logging, test_collections, test___all__ and
test_pydoc all pass.

Has this already been discussed elsewhere?

No response given

Links to previous discussion of this feature:

jaraco/configparser#73

Linked PRs

• gh-159684: Modernize string formatting and simplify constructs in configparser #159687