Stability of cid Column in PRAGMA table_xinfo and index_xinfo Results

---

Understanding the Role of cid in PRAGMA table_xinfo and index_xinfo

The cid column returned by SQLite’s PRAGMA table_xinfo and PRAGMA index_xinfo has been a subject of debate regarding its stability and reliability for correlating columns between tables and indexes. The core issue revolves around whether the cid values from these pragmas can be reliably used to match columns across different contexts, such as between a table and its associated indexes. Users have observed that while cid often appears to correspond to the ordinal position of a column in its table or index, this behavior is not explicitly documented or guaranteed by SQLite. This raises concerns about whether applications or tools relying on cid for schema introspection might break due to changes in SQLite’s internal implementation or schema modifications.

The PRAGMA table_xinfo command provides metadata about all columns in a table, including hidden and generated columns. Its output includes a cid column that historically reflects the order in which columns were declared in the table. Similarly, PRAGMA index_xinfo returns information about the columns (or expressions) in an index, including a cid column that, for indexed table columns, appears to match the cid from table_xinfo. However, SQLite’s documentation does not formally define the semantics of cid, leaving room for ambiguity. For example:

• Is cid stable across schema changes (e.g., ALTER TABLE operations)?
• Does cid have the same meaning in table_xinfo and index_xinfo?
• Are cid values guaranteed to align with the physical storage order of columns?

These questions are critical for developers building tools that introspect SQLite schemas programmatically. A mismatch in cid values could lead to incorrect assumptions about column relationships, especially when dealing with generated columns, hidden columns (e.g., rowid aliases), or virtual tables.

---

Undocumented Behavior and Schema Evolution Risks

The instability of cid stems from two primary factors: undocumented behavior and schema evolution risks.

1. Undocumented Behavior:
   SQLite’s documentation for PRAGMA table_xinfo and PRAGMA index_xinfo historically omitted details about the cid column, including its purpose and guarantees. While the cid was observed to represent the column’s ordinal position in the table or index, this was an implementation detail rather than a contractual guarantee. SQLite reserves the right to change such details in future releases, which could invalidate tools or workflows relying on cid stability. For instance, a future version of SQLite might renumber cid values after schema alterations or introduce new column types that disrupt the existing numbering.

2. Schema Evolution Risks:
   Even if cid were stable within a single SQLite version, schema changes can invalidate assumptions about cid alignment. Consider a table tbl with columns a, b, and c. The cid values for these columns might be 0, 1, and 2, respectively. If a generated column d is added, PRAGMA table_xinfo might assign it cid=3. However, if the table is recreated with a different column order (e.g., using ALTER TABLE ... RENAME TO workarounds), the cid values could shift unexpectedly. Similarly, hidden columns (e.g., rowid aliases) may occupy cid positions that are not contiguous with visible columns.

3. Index-Specific Ambiguities:
   Indexes in SQLite can include expressions or columns from multiple tables (in the case of partial indexes or indexes on virtual tables). The cid values in PRAGMA index_xinfo may not map cleanly to table columns, especially when indexes involve computed values. For example, an index on a+1 would not correspond to any cid in the base table, yet index_xinfo might still report a cid (e.g., -1 for rowid-based references).

---

Reliable Alternatives to cid-Based Column Matching

To avoid reliance on cid, developers should adopt strategies that use column names for correlating metadata between tables and indexes. Below is a structured approach to achieve robust schema introspection:

Step 1: Use Column Names for Cross-Referencing

Column names are stable, user-defined identifiers that are explicitly documented and guaranteed to be unique within their context (table or index). Instead of joining PRAGMA table_xinfo and PRAGMA index_xinfo results on cid, join them on the name column. For example:

-- Get indexed columns and their corresponding table columns
SELECT 
  idx.name AS index_name,
  idx_col.name AS index_column,
  tbl_col.name AS table_column
FROM 
  (SELECT * FROM pragma_index_list('my_table')) idx
JOIN 
  (SELECT * FROM pragma_index_xinfo(idx.name)) idx_col
  ON idx_col.name = tbl_col.name
JOIN 
  (SELECT * FROM pragma_table_xinfo('my_table')) tbl_col
  ON tbl_col.name = idx_col.name;

Step 2: Handle Generated and Hidden Columns Explicitly

Generated columns (virtual or stored) and hidden columns (e.g., rowid) may not appear in all pragma outputs. Use the hidden column in PRAGMA table_xinfo to filter or annotate these cases:

SELECT 
  name,
  type,
  hidden,
  CASE hidden
    WHEN 1 THEN 'Virtual Generated'
    WHEN 2 THEN 'Stored Generated'
    WHEN 3 THEN 'rowid Alias'
    ELSE 'Ordinary Column'
  END AS column_type
FROM pragma_table_xinfo('my_table');

Step 3: Validate Against SQLite’s Documentation Updates

Recent updates to SQLite’s documentation explicitly warn against relying on cid stability. Consult the latest documentation for PRAGMA table_xinfo and PRAGMA index_xinfo to verify guarantees about column names and behavior. For example:

• The name column is guaranteed to reflect the user-defined column name.
• The hidden column categorizes generated/hidden columns with documented values (0 for ordinary, 1–3 for specific hidden types).

Step 4: Test for cid Stability in Target Environments

If migrating an existing codebase away from cid is impractical, conduct thorough testing to identify cid-dependent assumptions. Use schema alteration scripts to simulate column reordering and validate whether cid values remain consistent:

-- Simulate column reordering by recreating a table
BEGIN TRANSACTION;
ALTER TABLE my_table RENAME TO my_table_old;
CREATE TABLE my_table (c, b, a);  -- Columns reversed
INSERT INTO my_table SELECT c, b, a FROM my_table_old;
DROP TABLE my_table_old;
COMMIT;

-- Check cid values after reordering
SELECT * FROM pragma_table_xinfo('my_table');

Step 5: Leverage SQLite’s Internal Schema Tables

For advanced use cases, query SQLite’s internal sqlite_master table and sqlite_schema pragmas to derive column order from the original CREATE TABLE/CREATE INDEX statements. This approach is brittle but can provide insights into the schema’s evolution:

-- Extract column order from the table definition
SELECT 
  sql
FROM 
  sqlite_master
WHERE 
  type = 'table' 
  AND name = 'my_table';

Step 6: Monitor for Future SQLite Changes

Subscribe to SQLite’s release notes and changelogs for updates affecting pragma behavior. For instance, if a future version formalizes cid semantics, adjust your codebase accordingly. Until then, treat cid as an undocumented implementation detail.

---

By prioritizing column names and adhering to documented guarantees, developers can future-proof their applications against SQLite internals changes. The cid column, while convenient for ad-hoc inspections, lacks the contractual stability required for programmatic use.