distsql: don't use VALUE encodings in HashJoiner

It looks like the VALUE encodings are not unique (they contain a column ID) so
they shouldn't be used for equality testing in hash joins.

pkg/sql/distsql/hashjoiner.go

@@ -240,7 +240,12 @@ func (h *hashJoiner) encode(
 		if row[colIdx].IsNull() {
 			return nil, true, nil
 		}
-		appendTo, err = row[colIdx].Encode(&h.datumAlloc, sqlbase.DatumEncoding_VALUE, appendTo)
+		// Note: we cannot compare VALUE encodings because they contain column IDs
+		// which can vary.
+		// TODO(radu): we should figure out what encoding is readily available and
+		// use that (though it needs to be consistent across all rows). We could add
+		// functionality to compare VALUE encodings ignoring the column ID.
+		appendTo, err = row[colIdx].Encode(&h.datumAlloc, sqlbase.DatumEncoding_ASCENDING_KEY, appendTo)
 		if err != nil {
 			return appendTo, false, err
 		}

pkg/sql/sqlbase/encoded_datum.go

@@ -197,6 +197,8 @@ func (ed *EncDatum) Encoding() (DatumEncoding, bool) {
 
 // Encode appends the encoded datum to the given slice using the requested
 // encoding.
+// Note: DatumEncoding_VALUE encodings are not unique because they can contain
+// a column ID so they should not be used to test for equality.
 func (ed *EncDatum) Encode(a *DatumAlloc, enc DatumEncoding, appendTo []byte) ([]byte, error) {
 	if ed.encoded != nil && enc == ed.encoding {
 		// We already have an encoding that matches