fix: log(0.0::float8) should error, not return -inf

[shinzoxD]

Which issue does this PR close?

• Closes PostgreSQL compatibility: log(0.0::float8) should error, not return -inf #22261.

Rationale for this change

SELECT log(0.0::float8) currently returns -inf. PostgreSQL raises ERROR: cannot take logarithm of zero.

This is the same class of domain error already accepted for sqrt of a negative number (#22260 / #22308), power(0, negative) (#22272), and factorial of a negative number (#22270). IEEE 754 would yield -inf; PostgreSQL treats it as undefined.

A previous attempt (#22564) grew into a log-function refactor (simplification rewrites, log(1, 64), etc.) and was closed. This PR stays scoped to the issue: reject a zero value at evaluation time.

What changes are included in this PR?

• log now returns cannot take logarithm of zero when the number being logged is zero (including -0.0, float32/float64, and decimal zeros).
• A zero base (log(0, 64)) and log(1, 64) are unchanged. Those are separate compatibility cases.
• Existing log simplifications (log(a, 1) => 0, log(a, a) => 1, log(a, power(a, b)) => b) are unchanged.
• User-guide note updated: it still claimed log(0) returns -inf and sqrt(-1) returns NaN, both of which are now domain errors.

Are these changes tested?

Yes.

• Unit tests in datafusion/functions/src/math/log.rs cover unary/binary float64, float32, -0.0, array input, decimal128, and decimal256.
• math.slt adds the issue query plus two-arg, negative-zero, column, and decimal cases, matching the style of the power(0, -1) tests.
• scalar.slt updates the previous log(0) => -Infinity expectation to the domain error, and keeps log(1, 64) => Infinity.

Ran:

• cargo test -p datafusion-functions --lib math::log
• cargo clippy -p datafusion-functions --lib -- -D warnings
• cargo test -p datafusion-sqllogictest --test sqllogictests -- math.slt
• cargo test -p datafusion-sqllogictest --test sqllogictests -- scalar.slt (log cases pass; later failures are missing testing submodule testdata for aggregate_test_100, unrelated)

Are there any user-facing changes?

Yes. log(0), log(0.0::float8), and other zero values now fail the query with cannot take logarithm of zero instead of returning -inf. Queries that relied on the IEEE result need to filter zeros first.

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[kosiew]

hi @shinzoxD
Can you fix the CI errors?

[kosiew]

@shinzoxD,

Thanks for working on this. The new PostgreSQL-compatible domain error handling and the additional coverage look useful overall.

I found one simplification path that can still bypass the new zero-value invariant, so I think that needs to be addressed before merging. I also left a small testing suggestion for the additional physical types touched by this change.

[kosiew]

I think this simplification can bypass the new invariant. log(a, a) is simplified directly to 1, so the logged value is never evaluated. For example, SELECT log(0.0::float8, 0.0::float8) returns 1.0 on this commit instead of the expected domain error.

The similar log(a, power(a, b)) rewrite around lines 405-412 can also bypass the check when the power result is zero.

Could we guard these rewrites so they only apply when the domain-error preconditions are known to hold? It would also be good to add planner-level regression cases for these zero-valued forms.

[kosiew]

Nice to see the new coverage for Float32/64 and Decimal128/256. Since this change also touches the separate Float16, Decimal32, and Decimal64 branches, could we add a small table-driven unit test that exercises every changed physical type?
That should help catch branch-specific regressions in the future.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 97.82609% with 3 lines in your changes missing coverage. Please review.
✅ Project coverage is 81.19%. Comparing base (e7e037d) to head (3e6d538).
⚠️ Report is 18 commits behind head on main.

Files with missing lines	Patch %	Lines
datafusion/functions/src/math/log.rs	97.82%	2 Missing and 1 partial ⚠️

Additional details and impacted files

@@           Coverage Diff            @@
##             main   #24399    +/-   ##
========================================
  Coverage   81.18%   81.19%            
========================================
  Files        1110     1110            
  Lines      388906   389029   +123     
  Branches   388906   389029   +123     
========================================
+ Hits       315733   315854   +121     
- Misses      54576    54577     +1     
- Partials    18597    18598     +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.