[DATABRICKS-4] PLU-600: add databricks create row action

[pregnantboy]

Add create-row action for Databricks integration

Add create-row action with zod validation and parameterized SQL insert.
Supports dynamic table/column selection via dropdowns with inline
create-new-column and modal create-new-table options.

The action validates table and column names to only allow lowercase letters, numbers, and underscores. It constructs INSERT statements with named parameters for safe SQL execution and includes comprehensive error handling with structured logging.

[pregnantboy]

Warning

This pull request is not mergeable via GitHub because a downstack PR is open. Once all requirements are satisfied, merge this PR as a stack on Graphite.
Learn more

• [DATABRICKS-5] PLU-600: add frontend support for databricks dynamic actions #1530
• [DATABRICKS-4] PLU-600: add databricks create row action #1529 👈 (View in Graphite)
• [DATABRICKS-3] PLU-600: add dynamic data queries, dynamic actions, and GraphQL mutation #1528
• [DATABRICKS-2] PLU-600: add databricks app skeleton, auth, and client setup #1527
• [DATABRICKS-1] PLU-600: add databricks config, env vars, and package dependency #1526
• trunk/databricks

This stack of pull requests is managed by Graphite. Learn more about stacking.

[linear]

PLU-600 Databricks integration - Road to production

[datadog-opengovsg]

✅ Tests

🎉 All green!

❄️ No new flaky tests detected
🧪 All tests passed

_{This comment will be updated automatically if new data arrives.
🔗 Commit SHA: 775b175 | Docs | Was this helpful? Give us feedback!}

[graphite-app]

The rowData array schema has no minimum length validation, allowing empty arrays. This will generate invalid SQL like INSERT INTO table () VALUES () and cause execution to fail.

Fix by adding minimum length validation:

rowData: z.array(
  z.object({
    columnName: z
      .string()
      .min(1, { message: 'Column name is required' })
      .regex(/^[a-z0-9_]+$/, {
        message:
          'Column name can only contain lowercase letters, numbers and underscores',
      }),
    columnValue: z.string(),
  }),
).min(1, { message: 'At least one column is required' }),

Suggested change

	rowData: z.array(
	z.object({
	columnName: z
	.string()
	.min(1, { message: 'Column name is required' })
	.regex(/^[a-z0-9_]+$/, {
	message:
	'Column name can only contain lowercase letters, numbers and underscores',
	}),
	columnValue: z.string(),
	}),
	),
	rowData: z.array(
	z.object({
	columnName: z
	.string()
	.min(1, { message: 'Column name is required' })
	.regex(/^[a-z0-9_]+$/, {
	message:
	'Column name can only contain lowercase letters, numbers and underscores',
	}),
	columnValue: z.string(),
	}),
	).min(1, { message: 'At least one column is required' }),

Spotted by Graphite



Is this helpful? React 👍 or 👎 to let us know.

[kevinkim-ogp]

unlikely to happen but good to have

[graphite-app]

Resource leak: if an error occurs during statement execution or fetchAll (lines 140-143), the session created on line 126 is never closed, as endSession() on line 144 is skipped when the catch block is entered. This will leak Databricks sessions in production.

Fix by ensuring cleanup:

const { session, endSession } = await createSession($)
try {
  const { tableName, rowData } = parametersParseResult.data
  // ... SQL execution ...
  await operation.fetchAll()
} finally {
  await endSession()
}

Suggested change

	const { session, endSession } = await createSession($)
	const { tableName, rowData } = parametersParseResult.data
	const columnNames = rowData.map((row) => row.columnName)
	const columnValues = rowData.map((row) => row.columnValue)
	const statement = `INSERT INTO \`${tableName}\` (${columnNames
	.map((col) => `\`${col}\``)
	.join(',')}) VALUES (${columnValues
	.map((_val, index) => `:val${index}`)
	.join(',')})`
	const namedParameters = {} as Record<string, string>
	for (let i = 0; i < columnValues.length; i++) {
	namedParameters[`val${i}`] = columnValues[i]
	}
	const operation = await session.executeStatement(statement, {
	namedParameters,
	})
	await operation.fetchAll()
	await endSession()
	const { session, endSession } = await createSession($)
	try {
	const { tableName, rowData } = parametersParseResult.data
	const columnNames = rowData.map((row) => row.columnName)
	const columnValues = rowData.map((row) => row.columnValue)
	const statement = `INSERT INTO \`${tableName}\` (${columnNames
	.map((col) => `\`${col}\``)
	.join(',')}) VALUES (${columnValues
	.map((_val, index) => `:val${index}`)
	.join(',')})`
	const namedParameters = {} as Record<string, string>
	for (let i = 0; i < columnValues.length; i++) {
	namedParameters[`val${i}`] = columnValues[i]
	}
	const operation = await session.executeStatement(statement, {
	namedParameters,
	})
	await operation.fetchAll()
	} finally {
	await endSession()
	}

Spotted by Graphite



Is this helpful? React 👍 or 👎 to let us know.

[kevinkim-ogp]

seems valid. if somehow createSession fails, it will still be caught and thrown anw

[kevinkim-ogp]

lgtm, can create row successfully via the action!

probably can address the two comments from Graphite, both seem valid.

sidenote: users can manually upload datasets to databricks with invalid column names i.e., with spaces, and our action will fail. doesn't seem like we have much control over this, but probably good to include somewhere in our guide

[kevinkim-ogp]

unlikely to happen but good to have

[kevinkim-ogp]

seems valid. if somehow createSession fails, it will still be caught and thrown anw