Limitations of Zend_Db_Table::save on missing primary key

Answer

When a primary‑key value is supplied but no row with that key exists, Zend_Db_Table::save attempts an UPDATE. The UPDATE affects zero rows, so the method returns false (or 0) and does not throw an exception. It does not automatically fall back to an INSERT.

Confirmed facts

• The decision is based solely on the presence of a non‑null primary‑key value; the adapter’s update() is called.
• An UPDATE on a non‑existent key returns zero affected rows on MySQL, PostgreSQL, SQLite, etc.
• If the primary key is null or omitted, save falls back to insert(), which creates a new row (and a new auto‑increment value if applicable).
• Retrying the same save call with the same missing key repeats the UPDATE, again affecting zero rows – no duplicate row is created.

Steps to verify the behavior

1. Create a table with a primary key column (e.g., id INT PRIMARY KEY).
2. Instantiate the table class and call save on a row object where id is set to a value that does not exist in the table.
3. Check the return value; it should be false/0 and no exception.
4. Repeat the call; the return value remains false/0 and a second row is not inserted.
5. Now set the primary key to null (or omit it) and call save twice; each call will perform an INSERT and generate two distinct rows.

Missing diagnostic detail

To confirm whether duplicates could appear in your specific setup, please tell us whether the primary key column is defined as AUTO_INCREMENT (or a sequence) and is supplied by the application, or whether it is a user‑assigned key that you expect to exist. This determines if a retry with a null key could cause duplicate inserts.