Improve the performance of QueryContext by reusing the result channel

This commit improves the performance of QueryContext by changing it to
reuse the result channel instead of creating a new one for each query.
This is particularly impactful for queries that scan more than one row.

It also adds a test that actually exercises the sqlite3_interrupt logic
since the existing tests did not. Those tests cancelled the context
before scanning any of the rows and could be made to pass without ever
calling sqlite3_interrupt. The below version of SQLiteRows.Next passes
the previous tests:

```go
func (rc *SQLiteRows) Next(dest []driver.Value) error {
	rc.s.mu.Lock()
	defer rc.s.mu.Unlock()
	if rc.s.closed {
		return io.EOF
	}
	if err := rc.ctx.Err(); err != nil {
		return err
	}
	return rc.nextSyncLocked(dest)
}
```

Benchmark results:
```
goos: darwin
goarch: arm64
pkg: github.com/mattn/go-sqlite3
cpu: Apple M1 Max
                                          │    b.txt    │               n.txt                │
                                          │   sec/op    │   sec/op     vs base               │
Suite/BenchmarkQueryContext/Background-10   4.088µ ± 1%   4.154µ ± 3%  +1.60% (p=0.011 n=10)
Suite/BenchmarkQueryContext/WithCancel-10   12.84µ ± 3%   11.67µ ± 3%  -9.08% (p=0.000 n=10)
geomean                                     7.245µ        6.963µ       -3.89%

                                          │    b.txt     │                 n.txt                  │
                                          │     B/op     │     B/op      vs base                  │
Suite/BenchmarkQueryContext/Background-10     400.0 ± 0%     400.0 ± 0%        ~ (p=1.000 n=10) ¹
Suite/BenchmarkQueryContext/WithCancel-10   2.547Ki ± 0%   1.282Ki ± 0%  -49.67% (p=0.000 n=10)
geomean                                      1021.4          724.6       -29.06%
¹ all samples are equal

                                          │   b.txt    │                n.txt                 │
                                          │ allocs/op  │ allocs/op   vs base                  │
Suite/BenchmarkQueryContext/Background-10   12.00 ± 0%   12.00 ± 0%        ~ (p=1.000 n=10) ¹
Suite/BenchmarkQueryContext/WithCancel-10   49.00 ± 0%   28.00 ± 0%  -42.86% (p=0.000 n=10)
geomean                                     24.25        18.33       -24.41%
¹ all samples are equal
```

sqlite3.go

@@ -399,6 +399,7 @@ type SQLiteRows struct {
 	cls      bool
 	closed   bool
 	ctx      context.Context // no better alternative to pass context into Next() method
+	resultCh chan error
 }
 
 type functionInfo struct {
@@ -2172,24 +2173,29 @@ func (rc *SQLiteRows) Next(dest []driver.Value) error {
 		return io.EOF
 	}
 
-	if rc.ctx.Done() == nil {
+	done := rc.ctx.Done()
+	if done == nil {
 		return rc.nextSyncLocked(dest)
 	}
-	resultCh := make(chan error)
-	defer close(resultCh)
+	if err := rc.ctx.Err(); err != nil {
+		return err // Fast check if the channel is closed
+	}
+	if rc.resultCh == nil {
+		rc.resultCh = make(chan error)
+	}
 	go func() {
-		resultCh <- rc.nextSyncLocked(dest)
+		rc.resultCh <- rc.nextSyncLocked(dest)
 	}()
 	select {
-	case err := <-resultCh:
+	case err := <-rc.resultCh:
 		return err
-	case <-rc.ctx.Done():
+	case <-done:
 		select {
-		case <-resultCh: // no need to interrupt
+		case <-rc.resultCh: // no need to interrupt
 		default:
 			// this is still racy and can be no-op if executed between sqlite3_* calls in nextSyncLocked.
 			C.sqlite3_interrupt(rc.s.c.db)
-			<-resultCh // ensure goroutine completed
+			<-rc.resultCh // ensure goroutine completed
 		}
 		return rc.ctx.Err()
 	}

sqlite3_go18_test.go

@@ -11,10 +11,12 @@ package sqlite3
 import (
 	"context"
 	"database/sql"
+	"errors"
 	"fmt"
 	"io/ioutil"
 	"math/rand"
 	"os"
+	"strings"
 	"sync"
 	"testing"
 	"time"
@@ -268,6 +270,151 @@ func TestQueryRowContextCancelParallel(t *testing.T) {
 	}
 }
 
+// Test that we can successfully interrupt a long running query when
+// the context is canceled. The previous two QueryRowContext tests
+// only test that we handle a previously cancelled context and thus
+// do not call sqlite3_interrupt.
+func TestQueryRowContextCancelInterrupt(t *testing.T) {
+	db, err := sql.Open("sqlite3", ":memory:")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer db.Close()
+
+	// Test that we have the unixepoch function and if not skip the test.
+	if _, err := db.Exec(`SELECT unixepoch(datetime(100000, 'unixepoch', 'localtime'))`); err != nil {
+		libVersion, libVersionNumber, sourceID := Version()
+		if strings.Contains(err.Error(), "no such function: unixepoch") {
+			t.Skip("Skipping the 'unixepoch' function is not implemented in "+
+				"this version of sqlite3:", libVersion, libVersionNumber, sourceID)
+		}
+		t.Fatal(err)
+	}
+
+	const createTableStmt = `
+	CREATE TABLE timestamps (
+		ts TIMESTAMP NOT NULL
+	);`
+	if _, err := db.Exec(createTableStmt); err != nil {
+		t.Fatal(err)
+	}
+
+	stmt, err := db.Prepare(`INSERT INTO timestamps VALUES (?);`)
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer stmt.Close()
+
+	// Computationally expensive query that consumes many rows. This is needed
+	// to test cancellation because queries are not interrupted immediately.
+	// Instead, queries are only halted at certain checkpoints where the
+	// sqlite3.isInterrupted is checked and true.
+	queryStmt := `
+	SELECT
+		SUM(unixepoch(datetime(ts + 10, 'unixepoch', 'localtime'))) AS c1,
+		SUM(unixepoch(datetime(ts + 20, 'unixepoch', 'localtime'))) AS c2,
+		SUM(unixepoch(datetime(ts + 30, 'unixepoch', 'localtime'))) AS c3,
+		SUM(unixepoch(datetime(ts + 40, 'unixepoch', 'localtime'))) AS c4
+	FROM
+		timestamps
+	WHERE datetime(ts, 'unixepoch', 'localtime')
+	LIKE
+		?;`
+
+	query := func(t *testing.T, timeout time.Duration) (int, error) {
+		// Create a complicated pattern to match timestamps
+		const pattern = "%2%0%2%4%-%-%:%:%"
+		ctx, cancel := context.WithTimeout(context.Background(), timeout)
+		defer cancel()
+		rows, err := db.QueryContext(ctx, queryStmt, pattern)
+		if err != nil {
+			return 0, err
+		}
+		var count int
+		for rows.Next() {
+			var n int64
+			if err := rows.Scan(&n, &n, &n, &n); err != nil {
+				return count, err
+			}
+			count++
+		}
+		return count, rows.Err()
+	}
+
+	average := func(n int, fn func()) time.Duration {
+		start := time.Now()
+		for i := 0; i < n; i++ {
+			fn()
+		}
+		return time.Since(start) / time.Duration(n)
+	}
+
+	createRows := func(n int) {
+		t.Logf("Creating %d rows", n)
+		if _, err := db.Exec(`DELETE FROM timestamps; VACUUM;`); err != nil {
+			t.Fatal(err)
+		}
+		ts := time.Date(2024, 6, 6, 8, 9, 10, 12345, time.UTC).Unix()
+		rr := rand.New(rand.NewSource(1234))
+		for i := 0; i < n; i++ {
+			if _, err := stmt.Exec(ts + rr.Int63n(10_000) - 5_000); err != nil {
+				t.Fatal(err)
+			}
+		}
+	}
+
+	const TargetRuntime = 200 * time.Millisecond
+	const N = 5_000 // Number of rows to insert at a time
+
+	// Create enough rows that the query takes ~200ms to run.
+	start := time.Now()
+	createRows(N)
+	baseAvg := average(4, func() {
+		if _, err := query(t, time.Hour); err != nil {
+			t.Fatal(err)
+		}
+	})
+	t.Log("Base average:", baseAvg)
+	rowCount := N * (int(TargetRuntime/baseAvg) + 1)
+	createRows(rowCount)
+	t.Log("Table setup time:", time.Since(start))
+
+	// Set the timeout to 1/10 of the average query time.
+	avg := average(2, func() {
+		n, err := query(t, time.Hour)
+		if err != nil {
+			t.Fatal(err)
+		}
+		if n == 0 {
+			t.Fatal("scanned zero rows")
+		}
+	})
+	// Guard against the timeout being too short to reliably test.
+	if avg < TargetRuntime/2 {
+		t.Fatalf("Average query runtime should be around %s got: %s ",
+			TargetRuntime, avg)
+	}
+	timeout := (avg / 10).Round(100 * time.Microsecond)
+	t.Logf("Average: %s Timeout: %s", avg, timeout)
+
+	for i := 0; i < 10; i++ {
+		tt := time.Now()
+		n, err := query(t, timeout)
+		if !errors.Is(err, context.DeadlineExceeded) {
+			fn := t.Errorf
+			if err != nil {
+				fn = t.Fatalf
+			}
+			fn("expected error %v got %v", context.DeadlineExceeded, err)
+		}
+		d := time.Since(tt)
+		t.Logf("%d: rows: %d duration: %s", i, n, d)
+		if d > timeout*4 {
+			t.Errorf("query was cancelled after %s but did not abort until: %s", timeout, d)
+		}
+	}
+}
+
 func TestExecCancel(t *testing.T) {
 	db, err := sql.Open("sqlite3", ":memory:")
 	if err != nil {

sqlite3_test.go

@@ -10,6 +10,7 @@ package sqlite3
 
 import (
 	"bytes"
+	"context"
 	"database/sql"
 	"database/sql/driver"
 	"errors"
@@ -2030,7 +2031,7 @@ func BenchmarkCustomFunctions(b *testing.B) {
 }
 
 func TestSuite(t *testing.T) {
-	initializeTestDB(t)
+	initializeTestDB(t, false)
 	defer freeTestDB()
 
 	for _, test := range tests {
@@ -2039,7 +2040,7 @@ func TestSuite(t *testing.T) {
 }
 
 func BenchmarkSuite(b *testing.B) {
-	initializeTestDB(b)
+	initializeTestDB(b, true)
 	defer freeTestDB()
 
 	for _, benchmark := range benchmarks {
@@ -2068,8 +2069,13 @@ type TestDB struct {
 
 var db *TestDB
 
-func initializeTestDB(t testing.TB) {
-	tempFilename := TempFilename(t)
+func initializeTestDB(t testing.TB, memory bool) {
+	var tempFilename string
+	if memory {
+		tempFilename = ":memory:"
+	} else {
+		tempFilename = TempFilename(t)
+	}
 	d, err := sql.Open("sqlite3", tempFilename+"?_busy_timeout=99999")
 	if err != nil {
 		os.Remove(tempFilename)
@@ -2084,9 +2090,11 @@ func freeTestDB() {
 	if err != nil {
 		panic(err)
 	}
-	err = os.Remove(db.tempFilename)
-	if err != nil {
-		panic(err)
+	if db.tempFilename != "" && db.tempFilename != ":memory:" {
+		err := os.Remove(db.tempFilename)
+		if err != nil {
+			panic(err)
+		}
 	}
 }
 
@@ -2107,6 +2115,7 @@ var tests = []testing.InternalTest{
 var benchmarks = []testing.InternalBenchmark{
 	{Name: "BenchmarkExec", F: benchmarkExec},
 	{Name: "BenchmarkQuery", F: benchmarkQuery},
+	{Name: "BenchmarkQueryContext", F: benchmarkQueryContext},
 	{Name: "BenchmarkParams", F: benchmarkParams},
 	{Name: "BenchmarkStmt", F: benchmarkStmt},
 	{Name: "BenchmarkRows", F: benchmarkRows},
@@ -2479,6 +2488,65 @@ func benchmarkQuery(b *testing.B) {
 	}
 }
 
+// benchmarkQueryContext is benchmark for QueryContext
+func benchmarkQueryContext(b *testing.B) {
+	const createTableStmt = `
+	CREATE TABLE IF NOT EXISTS query_context(
+		id INTEGER PRIMARY KEY
+	);
+	DELETE FROM query_context;
+	VACUUM;`
+	test := func(ctx context.Context, b *testing.B) {
+		if _, err := db.Exec(createTableStmt); err != nil {
+			b.Fatal(err)
+		}
+		for i := 0; i < 10; i++ {
+			_, err := db.Exec("INSERT INTO query_context VALUES (?);", int64(i))
+			if err != nil {
+				db.Fatal(err)
+			}
+		}
+		stmt, err := db.PrepareContext(ctx, `SELECT id FROM query_context;`)
+		if err != nil {
+			b.Fatal(err)
+		}
+		b.Cleanup(func() { stmt.Close() })
+
+		var n int
+		for i := 0; i < b.N; i++ {
+			rows, err := stmt.QueryContext(ctx)
+			if err != nil {
+				b.Fatal(err)
+			}
+			for rows.Next() {
+				if err := rows.Scan(&n); err != nil {
+					b.Fatal(err)
+				}
+			}
+			if err := rows.Err(); err != nil {
+				b.Fatal(err)
+			}
+		}
+	}
+
+	// When the context does not have a Done channel we should use
+	// the fast path that directly handles the query instead of
+	// handling it in a goroutine. This benchmark also serves to
+	// highlight the performance impact of using a cancelable
+	// context.
+	b.Run("Background", func(b *testing.B) {
+		test(context.Background(), b)
+	})
+
+	// Benchmark a query with a context that can be canceled. This
+	// requires using a goroutine and is thus much slower.
+	b.Run("WithCancel", func(b *testing.B) {
+		ctx, cancel := context.WithCancel(context.Background())
+		defer cancel()
+		test(ctx, b)
+	})
+}
+
 // benchmarkParams is benchmark for params
 func benchmarkParams(b *testing.B) {
 	for i := 0; i < b.N; i++ {