Apr 16, 2026 · Accepted (cleanup)

Experiment 064: Drop redundant sqlite3_clear_bindings

Date: 2026-04-16

Status: Accepted (cleanup)

Problem

The bind_params function in resqlite.c calls sqlite3_clear_bindings before binding each parameter. The comment explained this as defensive code for "reusing a statement with fewer params than the previous call."

But examining the full control flow:

1. bind_params verifies sqlite3_bind_parameter_count(stmt) == param_count and returns SQLITE_RANGE if they differ.
2. The function only proceeds to the bind loop if all N slots will be rebound.
3. TEXT/BLOB bindings use SQLITE_STATIC, meaning SQLite doesn't copy or own the data — no cleanup needed on clear.
4. Cached statements are always reset before bind_params is called (via get_or_prepare_reader).

Given this, sqlite3_clear_bindings is always redundant — every slot will be overwritten by the subsequent bind loop, and there's no owned data to free.

Approach

Remove the sqlite3_clear_bindings(stmt) call. Move the param count check earlier so it guards the bind loop. Document why the clear is unnecessary.

 static int bind_params(sqlite3_stmt* stmt, const resqlite_param* params, int param_count) { int expected = sqlite3_bind_parameter_count(stmt); if (expected != param_count) { (void)sqlite3_bind_null(stmt, expected + 1); return SQLITE_RANGE; } // No clear_bindings needed: we verify param_count above and always // rebind every slot below. SQLITE_STATIC bindings have no owned copy // to free. Cached statements are always reset before this function. for (int i = 0; i < param_count; i++) { // bind slot i+1 with params[i] } } 

Results

No measurable benchmark signal. The savings (~50ns per bind) are below the noise floor of the benchmark suite.

Full suite run showed 1 apparent win (-11%) and 1 apparent regression (-16%) on unrelated metrics — both within run-to-run variance. Full test suite (126 tests) passes.

Decision

Accepted as cleanup. Not a performance win by benchmark measurement, but:

• The removed code was provably redundant given existing invariants
• Simplifies the bind path by one call
• Documents the invariants clearly for future maintainers

No correctness risk: every control-flow path that enters bind_params satisfies the preconditions that make the clear unnecessary.