Files
java-interview-lab/binary-search

binary-search

Companion code for the ankurm.com article Binary Search in Java: Iterative, Recursive, Lower/Upper Bound and 10 Interview Problems (a full rewrite of the 2015 post "Implementing Binary Search in Java", same URL).

Iterative and recursive search, the overflow-safe midpoint, lowerBound/upperBound, rotated arrays, "search on the answer" and ten interview problems, each cross-checked against an obviously-correct brute-force version on seeded random inputs. Every figure and console line in the article is quoted from a file in output/, which the JUnit tests write while asserting the same numbers. There is no docs/ folder: the explanations live in the article itself.

Versions

Component Version Verified against
JDK Temurin 25.0.5.1+2 (LTS) java -version
JUnit 6.1.3 junit-bom maven-metadata.xml on Maven Central (latest GA)
Maven 3.9.x mvn -version

Quickstart

cd binary-search
export JAVA_HOME=/path/to/jdk-25 PATH="$JAVA_HOME/bin:$PATH"
./scripts/run-all.sh      # runs 14 tests and regenerates output/*.txt
java src/main/java/com/ankurm/interviewlab/binarysearch/InteractiveDemo.java   # the original console program

The first Maven run must be online so the plugins land in the local cache.

Source layout

binary-search/
├── pom.xml
├── scripts/run-all.sh
├── src/main/java/com/ankurm/interviewlab/binarysearch/
│   ├── BinarySearch.java        iterative, recursive, lowerBound, upperBound, first/last/count
│   ├── Overflow.java            the (lo + hi) / 2 bug at 2,000,000,000 elements, and two fixes
│   ├── RotatedArray.java        search, minimum, and the minimum with duplicates
│   ├── SearchOnAnswer.java      firstTrue template, integer sqrt, Koko bananas, ship packages
│   ├── InterviewProblems.java   range, peak element, 2D matrix; index of all ten problems
│   ├── BruteForce.java          slow, obviously-correct references (no shared helpers)
│   ├── DryRuns.java             instrumented copies that record every step
│   ├── Pitfalls.java            two plausible-looking wrong versions
│   └── InteractiveDemo.java     the original 2015 console program
├── src/test/java/.../           one test class per output file group
└── output/*.txt                 captured real output

Captured output (output/)

File What it shows
01-dry-run-iterative.txt the three searches from the original article, step by step
02-recursive-vs-iterative.txt identical answers on 2,003 keys; recursion depth 5 / 11 / 21
03-overflow-mid.txt (lo + hi) / 2 wrapping negative and throwing at n = 2,000,000,000
04-lower-upper-bound.txt bounds, first/last/count and Arrays.binarySearch on duplicates
05-rotated-array.txt rotated search and minimum traces; the duplicates degradation
06-search-on-answer.txt how many yes/no questions sqrt, Koko and ship packages ask
07-ten-problems.txt one example per problem beside the brute-force answer
08-random-agreement.txt 34,000 seeded comparisons, 0 disagreements
09-operation-counts.txt elements looked at: linear vs binary, n = 10 to 1,000,000,000
10-pitfalls.txt the classic mistakes, each reproduced
11-interactive-session.txt the original program, same sample session as the 2015 post
12-jdk-source.txt the JDK's own midpoint line, read from lib/src.zip
13-sixteen-elements.txt 26 among 16 values: four probes, against 13 for a linear scan
14-no-bound-methods.txt reflection check: no lowerBound/upperBound in Arrays, Collections or List