insaf021 opened a new pull request, #355:
URL: https://github.com/apache/commons-math/pull/355

   On master, with custom bases:
   
       new HaltonSequenceGenerator(1, new int[] {1}, null).skipTo(1)    // 
never returns
       new HaltonSequenceGenerator(1, new int[] {0}, null).skipTo(1)    // 
ArithmeticException: / by zero
       new HaltonSequenceGenerator(1, new int[] {-2}, null).skipTo(1)   // 
[-0.5]
       new HaltonSequenceGenerator(3, new int[] {2, 3}, null)           // "3 
out of [1, 40] range"
   
   The constructor clones `bases` without looking at the values, and `get()` 
decomposes the index with `while (index > 0) { ...; index /= base[i]; }`. A 
base of 1 never reduces the index, so every point after the first spins 
forever. A base of 0 divides by zero on the second point instead of failing at 
construction, and a negative base gives coordinates outside the documented [0, 
1] interval. No base below 2 is meaningful for a Halton sequence, so the 
constructor now rejects it with `NumberIsTooSmallException`, after the existing 
argument checks.
   
   The dimension check in the same constructor reported `PRIMES.length` as the 
upper bound, although the javadoc documents `[1, len]` with len the length of 
`bases`. It now reports `bases.length`.
   
   `HaltonSequenceGeneratorTest` gains `testConstructorInvalidBase` and an 
upper-bound assertion in `testConstructor2`; both fail on master and pass with 
the patch. `mvn clean install apache-rat:check checkstyle:check spotbugs:check 
javadoc:javadoc -pl commons-math-legacy -am` is green on JDK 21 (3265 tests in 
legacy; the only reruns are the randomised SimplexOptimizerTest cases).
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to