-
SQL operator precedence (
TaskRepository.java,search_tasks.sql,task_search_package.sql): Parenthesised the title/description OR soarchived = FALSEand the status filter apply to both branches. Addedid DESCtie-breaker in all three files. -
Thread.sleep / logging (
TaskController.java): RemovedThread.sleep(was delaying short queries up to 1 s). ReplacedSystem.out.printlnwith SLF4JLoggeratDEBUGlevel. -
Input validation (
TaskController.java): Invalidstatus→ HTTP 400 instead of 500.page < 1orpageSize <= 0→ HTTP 400.pageSizecapped at 100. Offset useslongto avoid int overflow. -
LIKE wildcard escaping (
TaskController.java, all SQL files):\,%,_in user input are escaped before the LIKE term is built; all queries useESCAPE '\'. -
useTasks.jsrace & stuck loading:AbortControllercancels stale in-flight requests.setError(null)clears the previous error at fetch start.setLoading(false)is in afinallyblock so it always fires (skipped on abort). Signal wired throughapi.js. -
Debounce + page reset (
App.jsx,api.js): 300 msuseDebounceon the search input. Page resets to 1 on query or status change. Removedconsole.logfromapi.js.
Paging is still done in Java via subList — see Biggest remaining risk. No auth added. H2 DB, schema/data SQL, folder structure, and startup commands are unchanged.
In-memory paging (subList). Every request loads all matching rows into heap before slicing. The fix is LIMIT/OFFSET in SQL, but that was left out to keep the diff small.
Status values are OPEN, IN_PROGRESS, DONE. Oracle PL/SQL is a reference artefact only. CORS is restricted to http://localhost:5173.
Antigravity AI (Google DeepMind) generated and applied all changes. Every change was reviewed by Yash Patil before committing.