Merge #32805 #32866

32805: parser: batch some allocations in the scanner r=RaduBerinde a=RaduBerinde

The scanner allocates copies of all uppercase identifiers; preallocate
a buffer (proportional to the size of the query) to save some of these
allocations.

Benchmark:
```
name   old time/op    new time/op    delta
Parse    33.8µs ± 1%    33.7µs ± 1%     ~     (p=0.699 n=6+6)

name   old alloc/op   new alloc/op   delta
Parse    5.54kB ± 0%    5.68kB ± 0%   +2.60%  (p=0.002 n=6+6)

name   old allocs/op  new allocs/op  delta
Parse      92.0 ± 0%      81.0 ± 0%  -11.96%  (p=0.002 n=6+6)
```

Release note: None

32866: update cloud storage deps r=RaduBerinde a=RaduBerinde

Informs #30774.

Release note: None

googleapis/google-cloud-go@0fd7230...0ebda48
Azure/azure-pipeline-go@7571e8e...b8e3409
aws/aws-sdk-go@ee1f179...ddc06f9

I ran all the ccl tests with `customenv.mk`. The Azure keys are no longer valid, all other tests passed.

Co-authored-by: Radu Berinde <[email protected]>

pkg/ccl/storageccl/export_storage.go

@@ -782,7 +782,7 @@ func (g *gcsStorage) Size(ctx context.Context, basename string) (int64, error) {
 	if err != nil {
 		return 0, err
 	}
-	sz := r.Size()
+	sz := r.Attrs.Size
 	_ = r.Close()
 	return sz, nil
 }

pkg/sql/parser/parse_test.go

@@ -2785,21 +2785,38 @@ func TestUnimplementedSyntax(t *testing.T) {
 }
 
 func BenchmarkParse(b *testing.B) {
-	for i := 0; i < b.N; i++ {
-		st, err := parser.Parse(`
-			BEGIN;
-			UPDATE pgbench_accounts SET abalance = abalance + 77 WHERE aid = 5;
-			SELECT abalance FROM pgbench_accounts WHERE aid = 5;
-			INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (1, 2, 5, 77, CURRENT_TIMESTAMP);
-			END`)
-		if err != nil {
-			b.Fatal(err)
-		}
-		if len(st) != 5 {
-			b.Fatal("parsed wrong number of statements: ", len(st))
-		}
-		if _, ok := st[1].(*tree.Update); !ok {
-			b.Fatalf("unexpected statement type: %T", st[1])
-		}
+	testCases := []struct {
+		name, query string
+	}{
+		{
+			"simple",
+			`SELECT a FROM t WHERE a = 1`,
+		},
+		{
+			"string",
+			`SELECT a FROM t WHERE a = 'some-string' AND b = 'some-other-string'`,
+		},
+		{
+			"tpcc-delivery",
+			`SELECT no_o_id FROM new_order WHERE no_w_id = $1 AND no_d_id = $2 ORDER BY no_o_id ASC LIMIT 1`,
+		},
+		{
+			"account",
+			`BEGIN;
+			 UPDATE pgbench_accounts SET abalance = abalance + 77 WHERE aid = 5;
+			 SELECT abalance FROM pgbench_accounts WHERE aid = 5;
+			 INSERT INTO pgbench_history (tid, bid, aid, delta, mtime) VALUES (1, 2, 5, 77, CURRENT_TIMESTAMP);
+			 END`,
+		},
+	}
+	for _, tc := range testCases {
+		b.Run(tc.name, func(b *testing.B) {
+			for i := 0; i < b.N; i++ {
+				_, err := parser.Parse(tc.query)
+				if err != nil {
+					b.Fatal(err)
+				}
+			}
+		})
 	}
 }

pkg/sql/parser/scan.go

@@ -48,7 +48,8 @@ type Scanner struct {
 	// stmts contains the list of statements at the end of parsing.
 	stmts []tree.Statement
 
-	initialized bool
+	initialized   bool
+	bytesPrealloc []byte
 }
 
 // scanErr holds error state for a Scanner.
@@ -72,6 +73,41 @@ func (s *Scanner) init(str string) {
 	}
 	s.initialized = true
 	s.in = str
+	// Preallocate some buffer space for identifiers etc.
+	s.bytesPrealloc = make([]byte, len(str))
+}
+
+func (s *Scanner) allocBytes(length int) []byte {
+	if len(s.bytesPrealloc) >= length {
+		res := s.bytesPrealloc[:length:length]
+		s.bytesPrealloc = s.bytesPrealloc[length:]
+		return res
+	}
+	return make([]byte, length)
+}
+
+// buffer returns an empty []byte buffer that can be appended to. Any unused
+// portion can be returned later using returnBuffer.
+func (s *Scanner) buffer() []byte {
+	buf := s.bytesPrealloc[:0]
+	s.bytesPrealloc = nil
+	return buf
+}
+
+// returnBuffer returns the unused portion of buf to the scanner, to be used for
+// future allocBytes() or buffer() calls. The caller must not use buf again.
+func (s *Scanner) returnBuffer(buf []byte) {
+	if len(buf) < cap(buf) {
+		s.bytesPrealloc = buf[len(buf):]
+	}
+}
+
+// finishString casts the given buffer to a string and returns the unused
+// portion of the buffer. The caller must not use buf again.
+func (s *Scanner) finishString(buf []byte) string {
+	str := *(*string)(unsafe.Pointer(&buf))
+	s.returnBuffer(buf)
+	return str
 }
 
 // Tokens calls f on all tokens of the input until an EOF is encountered.
@@ -671,7 +707,8 @@ func (s *Scanner) scanIdent(lval *sqlSymType) {
 	} else if isASCII {
 		// We know that the identifier we've seen so far is ASCII, so we don't need
 		// to unicode normalize. Instead, just lowercase as normal.
-		b := make([]byte, s.pos-start)
+		b := s.allocBytes(s.pos - start)
+		_ = b[s.pos-start-1] // For bounds check elimination.
 		for i, c := range s.in[start:s.pos] {
 			if c >= 'A' && c <= 'Z' {
 				c += 'a' - 'A'
@@ -810,7 +847,8 @@ func (s *Scanner) scanPlaceholder(lval *sqlSymType) {
 
 // scanHexString scans the content inside x'....'.
 func (s *Scanner) scanHexString(lval *sqlSymType, ch int) bool {
-	var buf []byte
+	buf := s.buffer()
+
 	var curbyte byte
 	bytep := 0
 	const errInvalidBytesLiteral = "invalid hexadecimal bytes literal"
@@ -860,13 +898,13 @@ outer:
 	}
 
 	lval.id = BCONST
-	lval.str = *(*string)(unsafe.Pointer(&buf))
+	lval.str = s.finishString(buf)
 	return true
 }
 
 // scanBitString scans the content inside B'....'.
 func (s *Scanner) scanBitString(lval *sqlSymType, ch int) bool {
-	var buf []byte
+	buf := s.buffer()
 outer:
 	for {
 		b := s.next()
@@ -896,15 +934,15 @@ outer:
 	}
 
 	lval.id = BITCONST
-	lval.str = *(*string)(unsafe.Pointer(&buf))
+	lval.str = s.finishString(buf)
 	return true
 }
 
 // scanString scans the content inside '...'. This is used for simple
 // string literals '...' but also e'....' and b'...'. For x'...', see
 // scanHexString().
 func (s *Scanner) scanString(lval *sqlSymType, ch int, allowEscapes, requireUTF8 bool) bool {
-	var buf []byte
+	buf := s.buffer()
 	var runeTmp [utf8.UTFMax]byte
 	start := s.pos
 
@@ -995,6 +1033,6 @@ outer:
 		return false
 	}
 
-	lval.str = *(*string)(unsafe.Pointer(&buf))
+	lval.str = s.finishString(buf)
 	return true
 }