fix(snowflake): unload a table, not an inline query

The COPY INTO grammar places the source immediately before its copy options, so
an inlined query sits one parenthesis from being able to rewrite them. Guarding
that means matching Snowflake's tokenizer exactly, and three successive versions
of the guard were each defeated: // line comments, $$ dollar quoting, and a bare
carriage return, which the scanner did not treat as a line terminator but
Snowflake does. Each fix was a guess at a lexer the public docs do not specify.

Removes the inline-query source instead of guessing a fourth time. A table name
goes through qualifiedIdentifier, which is provably safe. Exporting a query
result now means materializing it first — a view, or CREATE TABLE AS SELECT via
Execute SQL — which the tool description, the block skill and the docs all say.

Also from the final audit:
- optionalBoolean accepts the string forms a direct tool call delivers, matching
  the other boolean readers on this block, and its TSDoc no longer states the
  serializer rule backwards
- the five JSON editors declare language: 'json', so invalid JSON is caught
  inline instead of at execution
- bound the RESULT_SCAN read in SQL, not only by rows_per_resultset
- pin every migration target to a live subblock id, for all blocks

Author: waleedlatif1

apps/docs/content/docs/en/integrations/snowflake-service-account.mdx

@@ -96,6 +96,14 @@ With a credential selected, these fields become pickers backed by metadata-only
 
 Each picker runs as the token's user under its **default** role — not the execution role set on the block — so an empty list is usually a privilege gap rather than an empty account. Switch any field to advanced mode to type a name directly or reference an upstream block's output instead.
 
+<Callout type="info">
+**Unload Data exports a table, not a query.** The COPY INTO grammar places the
+source immediately before its options, so an inline query would sit one
+parenthesis away from being able to rewrite them. To export a query result,
+materialize it first — a view, or `CREATE TABLE AS SELECT` via Execute SQL —
+then unload that object.
+</Callout>
+
 ## Rotating and Revoking
 
 A token's expiry is fixed at creation. To rotate, generate a new token on the same user and update the credential in Sim — the old one stays valid until you remove it. `ALTER USER ... REMOVE PROGRAMMATIC ACCESS TOKEN <name>` revokes immediately and cannot be undone.

apps/docs/content/docs/en/integrations/snowflake.mdx

@@ -373,7 +373,7 @@ Load files from an existing Snowflake stage with COPY INTO.
 
 ### Snowflake Unload Data
 
-Export a table or query result to files in a Snowflake stage with COPY INTO.
+Export a Snowflake table to files in a stage with COPY INTO.
 
 #### Input
 
@@ -386,8 +386,7 @@ Export a table or query result to files in a Snowflake stage with COPY INTO.
 | `database` | string | Yes | Database name |
 | `schema` | string | Yes | Schema name |
 | `stagePath` | string | Yes | Destination stage reference, for example @EXPORTS/daily |
-| `table` | string | No | Source table to unload; provide either this or a statement, not both |
-| `statement` | string | No | Source SELECT statement to unload; provide either this or a table, not both |
+| `table` | string | Yes | Source table to unload. To export a query result, materialize it first as a view or with CREATE TABLE AS SELECT, then unload that |
 | `fileFormat` | string | No | Named file format applied to the unloaded files |
 | `header` | boolean | No | Whether to write column headings into the unloaded files; supported for CSV and Parquet only |
 | `overwrite` | boolean | No | Whether to replace existing files with matching names in the stage |

apps/sim/app/workspace/[workspaceId]/w/[workflowId]/hooks/use-wand.ts

@@ -55,11 +55,15 @@ function buildWandContextInfo({
 
       case 'json-schema':
       case 'json-object':
+      case 'json-array':
       case 'table-schema':
         try {
           const parsed = JSON.parse(currentValue)
-          const keys = Object.keys(parsed)
-          contextInfo += `\n\nJSON analysis: Valid JSON with ${keys.length} top-level keys: ${keys.join(', ')}`
+          // Reporting "top-level keys" for an array would list numeric indices,
+          // which tells the model nothing about the shape it should produce.
+          contextInfo += Array.isArray(parsed)
+            ? `\n\nJSON analysis: Valid JSON array with ${parsed.length} items`
+            : `\n\nJSON analysis: Valid JSON with ${Object.keys(parsed).length} top-level keys: ${Object.keys(parsed).join(', ')}`
         } catch {
           contextInfo += `\n\nJSON analysis: Invalid JSON - needs fixing`
         }

apps/sim/blocks/blocks/snowflake.ts

@@ -157,7 +157,7 @@ const tableOperations = [
   'list_copy_history',
   'introspect_schema',
 ] as const
-const tableRequiredOperations = [...dataOperations, 'list_copy_history'] as const
+const tableRequiredOperations = [...dataOperations, 'unload_data', 'list_copy_history'] as const
 
 /** Operations that bound their result with a `limit` parameter. */
 const limitOperations = [
@@ -192,13 +192,22 @@ const maxRowsOperationSet: ReadonlySet<string> = new Set(maxRowsOperations)
 const limitOperationSet: ReadonlySet<string> = new Set(limitOperations)
 
 /**
- * A switch the user never touched serializes as `null`, and in advanced mode the
- * serializer emits every advanced sub-block regardless. Builders test optional
- * booleans with `!== undefined`, so an untouched switch would otherwise emit a
- * clause the user never asked for — `AUTO_RESUME = FALSE` being the damaging one.
+ * Normalizes an optional switch to a real boolean or `undefined`.
+ *
+ * With advanced mode on, the serializer evaluates each advanced sub-block's
+ * condition and emits an untouched switch as `null`. Builders test these with
+ * `!== undefined`, so that `null` would otherwise emit a clause the user never
+ * asked for — `AUTO_RESUME = FALSE`, which permanently disables auto-resume on
+ * the warehouse, being the damaging one.
+ *
+ * The string forms are accepted because a direct tool call delivers booleans
+ * that way, matching the other boolean readers on this block.
  */
 function optionalBoolean(value: unknown): boolean | undefined {
-  return typeof value === 'boolean' ? value : undefined
+  if (typeof value === 'boolean') return value
+  if (value === 'true') return true
+  if (value === 'false') return false
+  return undefined
 }
 
 function resolveCopyOnError(value: unknown, threshold: unknown): string | undefined {
@@ -294,12 +303,12 @@ export const SnowflakeBlock: BlockConfig<SnowflakeStatementResponse> = {
       title: 'SQL Statement',
       type: 'code',
       placeholder: 'SELECT * FROM ANALYTICS.PUBLIC.EVENTS LIMIT 100',
-      condition: { field: 'operation', value: ['execute_sql', 'unload_data'] },
+      condition: { field: 'operation', value: 'execute_sql' },
       required: { field: 'operation', value: 'execute_sql' },
       wandConfig: {
         enabled: true,
         prompt:
-          'Generate one Snowflake SQL statement for the described request. For Execute SQL, use positional ? placeholders for any values that will be bound; for Unload Data, write a complete SELECT with literal values and no placeholders. Return ONLY the SQL statement - no explanations, no extra text.',
+          'Generate one Snowflake SQL statement for the described request. Use positional ? placeholders for any values that will be bound. Return ONLY the SQL statement - no explanations, no extra text.',
         placeholder: 'Describe the query to run...',
         generationType: 'sql-query',
       },
@@ -308,6 +317,7 @@ export const SnowflakeBlock: BlockConfig<SnowflakeStatementResponse> = {
       id: 'bindings',
       title: 'Bindings',
       type: 'code',
+      language: 'json',
       placeholder: '{"1":{"type":"TEXT","value":"active"}}',
       condition: { field: 'operation', value: 'execute_sql' },
       mode: 'advanced',
@@ -423,6 +433,7 @@ export const SnowflakeBlock: BlockConfig<SnowflakeStatementResponse> = {
       id: 'rows',
       title: 'Rows',
       type: 'code',
+      language: 'json',
       placeholder: '[{"id":1,"status":"active"}]',
       condition: { field: 'operation', value: ['insert_rows', 'update_rows', 'upsert_rows'] },
       required: { field: 'operation', value: ['insert_rows', 'update_rows', 'upsert_rows'] },
@@ -438,6 +449,7 @@ export const SnowflakeBlock: BlockConfig<SnowflakeStatementResponse> = {
       id: 'matchColumns',
       title: 'Match Columns',
       type: 'code',
+      language: 'json',
       placeholder: '["id"]',
       condition: { field: 'operation', value: ['update_rows', 'upsert_rows'] },
       required: { field: 'operation', value: ['update_rows', 'upsert_rows'] },
@@ -453,6 +465,7 @@ export const SnowflakeBlock: BlockConfig<SnowflakeStatementResponse> = {
       id: 'filters',
       title: 'Match Filters',
       type: 'code',
+      language: 'json',
       placeholder: '{"status":"expired","tenant_id":42,"archived_at":null}',
       condition: { field: 'operation', value: 'delete_rows' },
       required: { field: 'operation', value: 'delete_rows' },
@@ -773,6 +786,7 @@ export const SnowflakeBlock: BlockConfig<SnowflakeStatementResponse> = {
       id: 'procedureArguments',
       title: 'Procedure Arguments',
       type: 'code',
+      language: 'json',
       placeholder: '[{"type":"TEXT","value":"daily"}]',
       condition: { field: 'operation', value: 'call_procedure' },
       mode: 'advanced',
@@ -1210,7 +1224,7 @@ export const SnowflakeBlockMeta = {
       name: 'export-snowflake-results',
       description: 'Unload a table or query result to files in a Snowflake stage.',
       content:
-        '# Export Snowflake Results\n\n## Steps\n1. Confirm the destination stage path and whether existing files may be overwritten.\n2. Choose exactly one source: a table, or a SELECT statement.\n3. Pick a named file format, and decide on column headings and whether one file or several.\n4. Run the unload and review the reported file count and row totals.\n\n## Output\nReturn the stage path, files written, and rows unloaded.',
+        '# Export Snowflake Results\n\n## Steps\n1. Confirm the destination table and the stage path, and whether existing files may be overwritten.\n2. To export a query result rather than a whole table, materialize it first — create a view or use CREATE TABLE AS SELECT via Execute SQL — then unload that object.\n3. Pick a named file format, and decide on column headings and whether one file or several.\n4. Run the unload and review the reported file count and row totals.\n\n## Output\nReturn the stage path, files written, and rows unloaded.',
     },
     {
       name: 'browse-snowflake-objects',

apps/sim/lib/integrations/integrations.json

@@ -18286,7 +18286,7 @@
         },
         {
           "name": "Unload Data",
-          "description": "Export a table or query result to files in a Snowflake stage with COPY INTO."
+          "description": "Export a Snowflake table to files in a stage with COPY INTO."
         },
         {
           "name": "List Databases",

apps/sim/lib/workflows/migrations/subblock-migrations.test.ts

@@ -2,13 +2,18 @@
  * @vitest-environment node
  */
 import { afterAll, describe, expect, it, vi } from 'vitest'
+import { getAllBlocks } from '@/blocks/registry'
 import type { BlockState } from '@/stores/workflows/workflow/types'
 
 vi.unmock('@/blocks/registry')
 
 import * as blocksBarrel from '@/blocks'
 import { getBlock as getRealBlock } from '@/blocks/registry'
-import { backfillCanonicalModes, migrateSubblockIds } from './subblock-migrations'
+import {
+  backfillCanonicalModes,
+  migrateSubblockIds,
+  SUBBLOCK_ID_MIGRATIONS,
+} from './subblock-migrations'
 
 /**
  * Under `isolate: false` the module under test may already be cached from an
@@ -35,6 +40,47 @@ function makeBlock(overrides: Partial<BlockState> & { type: string }): BlockStat
   } as BlockState
 }
 
+/**
+ * `dropParkedSubblocks` deletes any subblock whose id starts with `_removed_`,
+ * on the assumption that no live block declares one. Nothing enforces that
+ * naming rule at the block level, so pin it here — a block that adopted the
+ * prefix for a real field would have its value silently deleted on every load.
+ */
+describe('_removed_ prefix invariant', () => {
+  it('is never used as a live subblock id', () => {
+    const offenders = getAllBlocks().flatMap((block) =>
+      (block.subBlocks ?? [])
+        .filter((subBlock) => subBlock.id.startsWith('_removed_'))
+        .map((subBlock) => `${block.type}.${subBlock.id}`)
+    )
+    expect(offenders).toEqual([])
+  })
+})
+
+/**
+ * A migration target that names no live subblock silently drops the value: the
+ * rename writes a key nothing reads, and the sweep or the serializer discards
+ * it. Nothing else checks the right-hand side of the map.
+ */
+describe('migration targets', () => {
+  it('every rename points at a subblock that still exists', () => {
+    const offenders: string[] = []
+    for (const [blockType, renames] of Object.entries(SUBBLOCK_ID_MIGRATIONS)) {
+      const config = getAllBlocks().find((block) => block.type === blockType)
+      if (!config) {
+        offenders.push(`${blockType} (block not registered)`)
+        continue
+      }
+      const liveIds = new Set((config.subBlocks ?? []).map((subBlock) => subBlock.id))
+      for (const [legacyId, currentId] of Object.entries(renames)) {
+        if (currentId.startsWith('_removed_')) continue
+        if (!liveIds.has(currentId)) offenders.push(`${blockType}.${legacyId} -> ${currentId}`)
+      }
+    }
+    expect(offenders).toEqual([])
+  })
+})
+
 describe('migrateSubblockIds', () => {
   it('should preserve Instagram insight metrics after the subblock rename', () => {
     const input: Record<string, BlockState> = {
@@ -68,9 +114,14 @@ describe('migrateSubblockIds', () => {
           type: 'snowflake',
           subBlocks: {
             operation: { id: 'operation', type: 'dropdown', value: 'insert_rows' },
-            database: { id: 'database', type: 'short-input', value: 'ANALYTICS' },
-            schema: { id: 'schema', type: 'short-input', value: 'PUBLIC' },
-            table: { id: 'table', type: 'short-input', value: 'EVENTS' },
+            // Every legacy id in the map, so a rename added later without a
+            // matching assertion still fails here.
+            ...Object.fromEntries(
+              Object.keys(SUBBLOCK_ID_MIGRATIONS.snowflake).map((legacyId) => [
+                legacyId,
+                { id: legacyId, type: 'short-input', value: `value-${legacyId}` },
+              ])
+            ),
           },
         }),
       }
@@ -80,10 +131,13 @@ describe('migrateSubblockIds', () => {
       expect(migrated).toBe(true)
       // The advanced text members, not the pickers: a migrated block has no
       // credential yet, so a picker could not hydrate the stored name.
-      expect(blocks.b1.subBlocks.databaseName?.value).toBe('ANALYTICS')
-      expect(blocks.b1.subBlocks.schemaName?.value).toBe('PUBLIC')
-      expect(blocks.b1.subBlocks.tableName?.value).toBe('EVENTS')
-      expect(blocks.b1.subBlocks.database).toBeUndefined()
+      for (const [legacyId, currentId] of Object.entries(SUBBLOCK_ID_MIGRATIONS.snowflake)) {
+        if (currentId.startsWith('_removed_')) continue
+        expect(blocks.b1.subBlocks[currentId]?.value, `${legacyId} -> ${currentId}`).toBe(
+          `value-${legacyId}`
+        )
+        expect(blocks.b1.subBlocks[legacyId], legacyId).toBeUndefined()
+      }
     })
 
     /**

apps/sim/tools/generated/tool-metadata.ts

@@ -482,7 +482,14 @@ describe('Snowflake SQL builders', () => {
     )
   })
 
-  it('unloads from exactly one source and orders the COPY INTO clauses', () => {
+  /**
+   * The source is a table name, never an inline query. An inlined query sits
+   * directly before the copy-option slot, so anything escaping its parentheses
+   * becomes a copy clause — a guard for that has to match Snowflake's tokenizer
+   * exactly, and three versions of one were each defeated. `qualifiedIdentifier`
+   * removes the class instead of re-guarding it.
+   */
+  it('unloads a table and orders the COPY INTO clauses', () => {
     const base = {
       ...context,
       database: 'ANALYTICS',
@@ -500,68 +507,19 @@ describe('Snowflake SQL builders', () => {
         maxFileSizeBytes: 16_777_216,
       }).statement
     ).toBe(
+      // OVERWRITE/SINGLE/MAX_FILE_SIZE are all copyOptions members, so their
+      // order among themselves is free; HEADER must stay last.
       "COPY INTO @EXPORTS/daily FROM ANALYTICS.PUBLIC.EVENTS FILE_FORMAT = (FORMAT_NAME = 'ANALYTICS.PUBLIC.CSV_FORMAT') OVERWRITE = FALSE SINGLE = TRUE MAX_FILE_SIZE = 16777216 HEADER = TRUE"
     )
-    // A trailing semicolon on the source query would close the COPY statement.
-    // The newline before the closing paren keeps a trailing line comment in the
-    // source query from swallowing it.
-    expect(buildUnloadData({ ...base, statement: 'SELECT 1 -- daily' }).statement).toBe(
-      'COPY INTO @EXPORTS/daily FROM (SELECT 1 -- daily\n) OVERWRITE = FALSE'
-    )
-    expect(buildUnloadData({ ...base, statement: 'SELECT 1;' }).statement).toBe(
-      'COPY INTO @EXPORTS/daily FROM (SELECT 1\n) OVERWRITE = FALSE'
-    )
-    expect(() => buildUnloadData({ ...base, table: 'EVENTS', statement: 'SELECT 1' })).toThrow(
-      /exactly one of table or statement/
-    )
-    // Each of these ends the derived table early and supplies its own copy
-    // options, including OVERWRITE = TRUE, by hiding a paren in a construct the
-    // paren counter must skip. All are valid Snowflake syntax.
-    const breakouts: Array<[string, string, RegExp]> = [
-      [
-        'bare paren',
-        'SELECT 1) OVERWRITE = TRUE FILE_FORMAT = (TYPE = CSV',
-        /unbalanced parentheses/,
-      ],
-      [
-        'double-slash comment',
-        'SELECT 1 // (\n) OVERWRITE = TRUE FILE_FORMAT = (TYPE = CSV // )',
-        /unbalanced parentheses/,
-      ],
-      [
-        'dollar quoting',
-        'SELECT $$($$ ) OVERWRITE = TRUE FILE_FORMAT = (TYPE = CSV, RECORD_DELIMITER = $$)$$',
-        /unbalanced parentheses/,
-      ],
-      [
-        'nested block comment',
-        'SELECT 1 /* /* */ ( */ ) OVERWRITE = TRUE /* /* */ ) */',
-        /nested block comment/,
-      ],
-    ]
-    for (const [name, statement, expected] of breakouts) {
-      expect(() => buildUnloadData({ ...base, statement }), name).toThrow(expected)
-    }
-
-    // A paren legitimately inside a dollar-quoted string is not a breakout.
-    expect(buildUnloadData({ ...base, statement: 'SELECT $$a)b$$ AS x' }).statement).toContain(
-      'FROM (SELECT $$a)b$$ AS x\n)'
-    )
-    expect(() => buildUnloadData({ ...base, statement: 'SELECT $$unterminated' })).toThrow(
-      /unterminated dollar-quoted string/
-    )
-
-    // OVERWRITE is always emitted, so an injected duplicate collides instead of
-    // silently replacing staged files.
-    expect(buildUnloadData({ ...base, table: 'EVENTS' }).statement).toContain('OVERWRITE = FALSE')
-    expect(() => buildUnloadData({ ...base, statement: 'DROP TABLE EVENTS' })).toThrow(
-      /must be a SELECT or WITH query/
+    // OVERWRITE is always emitted so the option can never be left to a default.
+    expect(buildUnloadData({ ...base, table: 'EVENTS' }).statement).toBe(
+      'COPY INTO @EXPORTS/daily FROM ANALYTICS.PUBLIC.EVENTS OVERWRITE = FALSE'
     )
-    // Parens and quotes inside string literals and comments are not miscounted.
-    expect(
-      buildUnloadData({ ...base, statement: "SELECT ')' AS a -- )\nFROM T" }).statement
-    ).toContain("FROM (SELECT ')' AS a -- )\nFROM T\n)")
-    expect(() => buildUnloadData(base)).toThrow(/exactly one of table or statement/)
+    expect(() => buildUnloadData({ ...base, table: '' })).toThrow(/table is required/)
+    // No SQL text can reach the statement, so no breakout is expressible.
+    expect(() =>
+      buildUnloadData({ ...base, table: 'EVENTS) OVERWRITE = TRUE FILE_FORMAT = (TYPE = CSV' })
+    ).toThrow(/Invalid Snowflake identifier/)
     expect(() =>
       buildUnloadData({ ...base, table: 'EVENTS', maxFileSizeBytes: 5_368_709_121 })
     ).toThrow(/maxFileSizeBytes/)