Oracle Database n8n integrationAutomate Oracle Database with n8n.

This is the destructive end of the node, and Command decides how far it goes.

• Command: truncate empties the table and preserves its structure, delete removes the rows matching the conditions below, drop deletes data and structure permanently.
• Select Rows: the conditions that pick the rows, each built from a column, one of the 9 operators (Equal, Like, Is Null and the comparisons) and a value. Left empty with delete, every row goes.
• Combine Conditions: AND keeps rows matching all conditions, OR keeps rows matching at least one.

When the five row operations are too blunt, this one hands the wheel back: the statement is yours, and the rows it returns arrive as n8n items.

• Statement: required. The SQL to execute, with positional placeholders such as :1 and :2, or named ones such as :name.
• Bind Variable Placeholder Values: the values behind those placeholders. Binding them beats gluing them into the string, and keeps SQL injection out of reach.
• Statement Batching: Single Statement sends one statement for all incoming items, Independently runs one per item, Transaction groups them and rolls everything back if one fails.

Insert writes new rows and does nothing else: no lookup, no merge, one incoming item becomes one row.

• Columns: required. The mapping that says which value goes into which column of the table, either typed in or taken from the incoming item with an expression, {{ $json.field }}.
• Auto Commit: when it is on, the transaction on the current connection is committed at the end of statement execution.

Same write, one difference that changes everything: a row that already exists gets updated instead of being added a second time.

• Columns: required. The mapping between the incoming item and the table columns, values typed in or pulled with {{ $json.field }}.
• Statement Batching: Independently executes one statement per incoming item, so a single bad record does not take the others down; Transaction does the opposite and rolls the whole run back on the first failure.

Reading is where the node gives you the most knobs, and the first one decides the volume.

• Return All: on, every matching row comes back; off, Limit caps how many rows the node returns.
• Select Rows and Combine Conditions: the filter, one condition per column and operator, joined with AND or OR. Left empty, all rows are selected.
• Sort: one or more sort rules, each a column plus a Direction, ASC or DESC.
• Output Columns: restricts the returned columns, which keeps a wide table from flooding the rest of the workflow.

Update touches rows that already exist and leaves the rest of the table alone.

• Columns: required. The mapping that carries the new values, and the column used to find the row again, most often filled with {{ $json.field }} from the node before.
• Output Columns: the columns you want back after the write, handy when only a status matters downstream.
• Output Numbers As String: returns numbers as strings, on NUMBER, FLOAT and LONG columns only, which saves a long identifier from being rounded on its way out.