MAINT: Add tests for quantecon/timings and deduplicate its accessors

[mmcky]

Problem

quantecon/timings/ is the only subpackage with no test file at all, yet it exposes public API — qe.timings.float_precision() is reachable from the top-level namespace. It is 66 lines with two public functions and a module-level mutable global. Three things to address:

1. No tests. float_precision() has a validation branch (ValueError for non-integer or negative input) and get/set dual behaviour, neither exercised. The module-level global means tests need to restore state — worth getting right in the first test rather than retrofitting.
2. Two names for one read. float_precision() called with no argument and get_default_precision() return the same value by the same mechanism. One can go, or the relationship should be documented — as it stands a reader cannot tell which is canonical.
3. Docstring examples are doctest-shaped but never executed (adjacent: Run docstring examples in CI (--doctest-modules) #866 on running doctests in CI), so they are unverified prose that reads as tested.

Acceptance criteria

• A test module covering get, set, validation errors, and state restoration between tests
• The duplicate accessor removed, or its relationship to float_precision() documented
• Doctest examples either run under a configured doctest collector or are reformatted so they don't claim to be tested

---

From the July 2026 technical-debt audit (AI-assisted; claims verified against 28d4b3b on 2026-07-25).

[lntutor]

I’m taking this. I’ll add isolated coverage for the get/set/error paths with state restoration, document the accessor relationship without breaking compatibility, and revise the uncollected doctest-style examples.

[quantecon-services]

Closing this with a correction, since the premise turned out to be only partly right.

Criterion 1 (no tests) was already satisfied when the issue was opened. quantecon/timings/ has no test directory, but the module is fully exercised by TestGlobalPrecision in quantecon/util/tests/test_timing.py: get, set, all three ValueError paths (negative int, float, string), and state restoration via setup_method/teardown_method. Running that file alone gives 100% line coverage of quantecon/timings/timings.py on main. The audit checked for a test file per directory rather than by coverage, which is how this slipped through.

Criterion 2 (accessor relationship undocumented) was real but small. Removing get_default_precision() is not an option since quantecon/util/timing.py calls it in three places as the read-only accessor. #959 documents the relationship in its docstring.

Criterion 3 (doctest examples never run) is being handled more broadly. The only thing that fails under strict doctest is the print(...) line with no expected output; #959 repairs that single line and keeps the >>> examples so the module is ready for the package-wide docstring sweep in #920.

Two PRs were opened against this issue and both authors put in careful work:

• TEST: add tests for quantecon/timings (#889) #915 by @bwcummings1 added a well-built test module. Because the module was already covered, it added no new coverage (Coveralls confirmed "remained the same"), so it is being closed with thanks.
• MAINT: Test and clarify timing precision accessors #914 by @lntutor tackled all three criteria. Its docstring half inspired DOC: clarify timing precision accessors and repair doctest example #959; the test half duplicated the existing coverage. Also closed with thanks.

Sorry to both of you for the wasted effort here. The issue description sent you after a problem that was already solved, and that is on us, not you.

---

Generated by Claude Code