model-writtenrun 8933 - may be wrong, links go to the source messages. Commit-by-commit LLM analysis is in the Commits dialog.
- Pavel Stehule revived an earlier numeric trimming proposal with prototypes for minscale(), rtrim(), and btrim() #1; Tom Lane objected to confusion with string functions and identical rtrim()/btrim() semantics #2, and Pavel accepted different naming while arguing for a convenient trimming function #3.
- Pavel's first C patch implemented minscale() and trim_scale() using a shared minimum-scale calculation, with documentation and regression tests #4.
- Karl O. Pinc reviewed the implementation and requested that trim_scale(NaN) preserve NaN, along with clearer descriptions, explicit tests, and grouping of scale-related code #5; Pavel incorporated the behavioral fix and several presentation changes in a rebased revision #6.
- Tom requested consistent function naming #7, Karl proposed min_scale() #8 and supplied broader regression cases #10, and Pavel adopted these alongside comment and formatting corrections #13.
- Karl's remaining requests concerned code placement and local variable naming #14; Pavel implemented them while retaining the broader tests #15, after which Karl reported passing tests and successfully built documentation and marked the patch ready for a committer #16.
- Tom committed the final submission with modifications as 20d6225d16, making get_min_scale() handle trailing zero NumericDigits internally instead of relying on caller stripping, and also refining documentation, catalog entries, and tests; he explained that the helper change prevents a potential infinite loop if future callers supply unstripped values #18.
- Both min_scale() and trim_scale() were landed by the end of the thread, and Pavel closed by thanking Tom and suggesting that the feature might correspond to a TODO item #19.