SQLite .dump Failure with reverse_unordered_selects: Trigger/View Dependency Order

---

Understanding the Trigger/View Dependency Order in .dump Output

The core issue arises when using the SQLite .dump command after enabling the reverse_unordered_selects pragma, leading to an invalid dump file. The problem occurs because the .dump command outputs schema objects (specifically triggers and views) in an order that violates dependencies. For example, a trigger referencing a view may be dumped before the view itself, causing errors during re-import such as no such table: v. This behavior stems from a missing ORDER BY clause in the internal query responsible for schema extraction during .dump. The reverse_unordered_selects pragma reverses the default output order of SELECT statements lacking explicit ordering, which destabilizes implicit assumptions about schema object dependencies.

---

Root Causes of Dependency Violations in Schema Dumps

1. Missing Explicit Ordering in Schema Queries:
   The .dump command retrieves schema objects (views, triggers, indexes) using a SELECT query on sqlite_schema without an ORDER BY clause. By default, SQLite returns rows in an unspecified order when ORDER BY is omitted. However, the reverse_unordered_selects pragma inverts this implicit order, causing dependent objects (e.g., triggers) to appear before their prerequisites (e.g., views).

2. Dependency Awareness in Schema Serialization:
   SQLite’s .dump command does not perform dependency resolution when generating schema definitions. Instead, it relies on the order of entries in sqlite_schema. When reverse_unordered_selects is enabled, the absence of explicit ordering disrupts the natural insertion order of objects, leading to invalid sequences.

3. Pragma-Driven Query Behavior:
   The reverse_unordered_selects pragma modifies the query planner’s behavior for unordered SELECT statements. This pragma is designed for debugging and testing, but its interaction with internal schema queries (like those in .dump) was not fully accounted for, resulting in non-idiomatic schema output.

---

Resolving Invalid .dump Output via Schema Order Enforcement

Step 1: Diagnose the Presence of reverse_unordered_selects
Before executing .dump, check if reverse_unordered_selects is enabled:

PRAGMA reverse_unordered_selects;

If the result is 1, the pragma is active. Disable it temporarily for schema dumps:

PRAGMA reverse_unordered_selects = 0;

Note: This is a session-level setting. Closing and reopening the database connection resets it to the default (0).

Step 2: Patch the SQLite Shell (Advanced)
Modify the SQLite shell’s source code to enforce ordering in schema queries. The fix involves adding an ORDER BY clause to prioritize views over triggers. Apply this patch to src/shell.c.in:

@@ -8373,11 +8373,12 @@
   sqlite3_free(zSql);
   if( (p->shellFlgs & SHFLG_DumpDataOnly)==0 ){
    zSql = sqlite3_mprintf(
     "SELECT sql FROM sqlite_schema AS o "
     "WHERE (%s) AND sql NOT NULL"
-    " AND type IN ('index','trigger','view')",
+    " AND type IN ('index','trigger','view')"
+    " ORDER BY type='trigger'",
     zLike
    );
    run_table_dump_query(p, zSql);
    sqlite3_free(zSql);
   }

This change sorts schema objects such that triggers (type='trigger') are listed last. The expression type='trigger' evaluates to 1 for triggers and 0 for other types, ensuring non-trigger objects (views, indexes) are dumped first.

Step 3: Upgrade to SQLite 3.45.1 or Later
The issue is resolved in SQLite trunk (version 3.45.1 and newer). Upgrading ensures the .dump command includes the necessary ordering without manual patching. Verify your SQLite version:

sqlite3 --version

If using an affected version (e.g., 3.45.0), download the latest pre-release or build from source.

Step 4: Validate Dump Output
After applying fixes, test the .dump command:

sqlite3 test.db .dump > dump.sql
sqlite3 restored.db < dump.sql

Ensure no errors occur during restoration, particularly no such table or trigger cannot reference view messages.

Step 5: Avoid Reliance on Implicit Ordering
When scripting or automating schema operations, explicitly define dependencies using ORDER BY in queries against sqlite_schema. For example:

SELECT name, sql FROM sqlite_schema
WHERE type IN ('view', 'trigger')
ORDER BY
  CASE type
    WHEN 'view' THEN 1
    WHEN 'trigger' THEN 2
    ELSE 3
  END;

This ensures views are processed before triggers, regardless of pragma settings.

Step 6: Document Pragmas with Side Effects
If your workflow requires reverse_unordered_selects, document its implications for schema operations. Consider wrapping .dump commands in scripts that temporarily disable the pragma:

sqlite3 input.db "PRAGMA reverse_unordered_selects=0; .output dump.sql; .dump; .exit"

Conclusion: The .dump command’s dependency on implicit SELECT ordering makes it vulnerable to pragmas that alter default query behavior. Explicitly ordering schema objects by type resolves this, ensuring reliable database restoration.