MySQL UPDATE

How it works

UPDATE changes matching rows. A missing or incorrect WHERE clause can affect much more data than intended. Here the change identifies one part and also requires sufficient stock before subtracting two pieces. The predicate and arithmetic happen in one statement rather than a separate read followed by an unchecked write. The row count tells the application whether the conditional operation found an eligible record.

Run in a disposable MySQL lab

Use Oracle MySQL 8.4 with InnoDB, not Bhylu’s SQLite playground. The examples were reviewed on MySQL 8.4.8. XAMPP may provide MariaDB, which has different feature and collation behaviour. Check your actual server before running the script. These lessons do not require a change to Bhylu’s PHP, Apache or normal database service.

1. On a separate lab instance, create the empty bhylu_mysql_lab schema. Stop if it already exists; do not overwrite an existing database.
2. Select that schema and import the complete repair-parts seed once. It sets the documented session mode, utf8mb4 collation, UTC and English date-name locale.
3. Run this lesson in a fresh copy of the seed, not after another lesson’s writes. Use the command-line client or a compatible editor. DELIMITER lines are client instructions for stored-program examples.

Download this complete lesson script. Its setup is the linked seed; review the dataset before changing it. Administrative and destructive examples are restricted to the specifically named disposable objects. Every script repeats the session settings because each new connection has its own session. No MySQL browser runner is currently available.

Example

SET NAMES utf8mb4 COLLATE utf8mb4_0900_ai_ci;
SET SESSION sql_mode = 'STRICT_TRANS_TABLES,ONLY_FULL_GROUP_BY,ERROR_FOR_DIVISION_BY_ZERO,NO_ENGINE_SUBSTITUTION';
SET SESSION time_zone = '+00:00';
SET SESSION lc_time_names = 'en_US';
UPDATE parts SET quantity = quantity - 2
WHERE id = 1 AND quantity >= 2;
SELECT ROW_COUNT() AS changed;
SELECT quantity FROM parts WHERE id = 1;

Expected result

1
10

One row changes and ten pieces remain. A zero affected-row result must be interpreted under the chosen client semantics: a no-op update and no matching record are not always distinguishable from a generic row-count call. This subtraction necessarily changes a matching quantity, so zero means the protected operation did not happen. Multi-table business changes still need a transaction.