งานแบบเพิ่มทีละน้อยและการตรวจสอบความเป็นปัจจุบัน

Gradle is fast because it does not redo work. If a task ran before and nothing it depends on changed, Gradle marks it UP-TO-DATE and skips it. You enable this by declaring inputs and outputs.

Every task can declare what it reads (inputs) and what it produces (outputs). Gradle hashes them to decide whether to run.

Annotate or register input files so Gradle watches them for changes.

On each run Gradle compares current input/output hashes to the last run. If both match, the task is skipped; if any input changed, it re-runs.

Non-file inputs (a flag, a version string) are declared with inputs.property so changing them also invalidates the cache.

For reusable logic, write a task class with annotated properties: @InputFile, @OutputFile, @Input. Gradle reads the annotations automatically.

Annotations tell Gradle how to track each property without manual inputs/outputs wiring.

For tasks over many files, @Incremental inputs let the action process only the files that actually changed since last run.

Beyond up-to-date checks, the build cache can reuse outputs from a previous run on any machine, keyed by the same input hashes — huge for CI.

Forgetting to declare an input is the classic bug: Gradle skips a task that should have re-run, producing stale output. Always declare every input that affects the result.

To verify caching or work around a stale state, run with --rerun-tasks to ignore up-to-date checks for one build, or use outputs.upToDateWhen { false } to mark a task as never up to date.

Test your incremental build knowledge.

You learned to build fast, correct tasks:

• Gradle skips tasks whose inputs and outputs are unchanged
• Declare files with inputs.file/outputs.file and values with inputs.property
• Custom task classes use @InputFile/@OutputFile annotations
• The build cache reuses outputs across machines
• Undeclared inputs cause stale, incorrect skips