[tarantool-patches] Re: [PATCH v1 2/4] sql: disallow use of TYPEOF in Check

[Konstantin Osipov]

* Kirill Shcherbatov <kshcherbatov at tarantool.org> [19/01/10 23:12]:
> Due to the fact that we are going to perform CHECKs validations
> on the server side, checks are performed on the fields with
> affinity already applied that may differ with type of original
> data.
> After the introduction of static types, the need for type checks
> based on the CHECKs disappeared.
> 
> Need for #3691
> ---
>  src/box/sql/build.c         | 23 ++++++++++
>  test/sql-tap/check.test.lua | 89 ++-----------------------------------
>  2 files changed, 26 insertions(+), 86 deletions(-)
> 
> diff --git a/src/box/sql/build.c b/src/box/sql/build.c
> index 7e5bcc518..9c6b1b49a 100644
> --- a/src/box/sql/build.c
> +++ b/src/box/sql/build.c
> @@ -756,12 +756,35 @@ primary_key_exit:
>  	return;
>  }
>  
> +static int
> +sql_check_expr_test_callback(struct Walker *walker, struct Expr *expr)
> +{
> +	if (expr->op == TK_FUNCTION && strcmp(expr->u.zToken, "TYPEOF") == 0) {
> +		diag_set(ClientError, ER_SQL, "TYPEOF function is forbidden "
> +			 "for usage in Check constraint");
> +		walker->eCode = 1;
> +		return WRC_Abort;
> +	}
> +	return WRC_Continue;
> +}

This callback needs a comment. All functions need a comment, even
static ones. A commonly accepted abbreviation for callbacks is _cb
ending. 

A better option for function name would be ck_constraint_walk_expr_cb.

Please reply to my questions in previous emails.

>  void
>  sql_add_check_constraint(struct Parse *parser, struct ExprSpan *span)
>  {
>  	struct Expr *expr = span->pExpr;
>  	struct Table *table = parser->pNewTable;
>  	if (table != NULL) {
> +		/* Test if Check expression valid. */

is valid. valid is a general term. What problems specifically are
you looking for? What is forbidden in a check expression? please
add a better comment.

> +		struct Walker w;
> +		memset(&w, 0, sizeof(w));
> +		w.xExprCallback = sql_check_expr_test_callback;
> +		sqlite3WalkExpr(&w, expr);
> +		if (w.eCode != 0) {
> +			parser->rc = SQL_TARANTOOL_ERROR;
> +			parser->nErr++;
> +			goto release_expr;
> +		}
> +

I don't understand the premise for this code. Why do yo uneed a
separate walker, aren't these expressions checked already
someplace else in the code?

-- 
Konstantin Osipov, Moscow, Russia, +7 903 626 22 32
http://tarantool.io - www.twitter.com/kostja_osipov