From: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
To: tarantool-patches@freelists.org, Nikita Pettik <korablev@tarantool.org>
Subject: [tarantool-patches] Re: [PATCH] sql: fix row count calculation for DELETE optimization
Date: Mon, 19 Nov 2018 15:50:21 +0300 [thread overview]
Message-ID: <98a013cf-1c67-a5e9-0006-8384e57c4917@tarantool.org> (raw)
In-Reply-To: <20181119121256.8154-1-korablev@tarantool.org>
Thanks for the patch! LGTM.
On 19/11/2018 15:12, Nikita Pettik wrote:
> When SQL DELETE statement comes in most primitive from without WHERE
> clause and foreign key constraints, it is optimized and processed with
> one VDBE instruction (instead of several OP_Delete). However, it was
> forgotten to account affected tuples by row counter. Current patch fixes
> this obvious defect.
>
> Closes #3816
> ---
> Branch: https://github.com/tarantool/tarantool/tree/np/gh-3816-account-changes-for-simple-delete
> Issue: https://github.com/tarantool/tarantool/issues/3816
>
> src/box/sql.c | 5 ++++-
> src/box/sql/delete.c | 1 +
> src/box/sql/tarantoolInt.h | 2 +-
> src/box/sql/vdbe.c | 10 +++++++--
> test/sql/row-count.result | 54 +++++++++++++++++++++++++++++++++++++++++++++
> test/sql/row-count.test.lua | 20 +++++++++++++++++
> 6 files changed, 88 insertions(+), 4 deletions(-)
>
> diff --git a/src/box/sql.c b/src/box/sql.c
> index d7df84874..c3418008c 100644
> --- a/src/box/sql.c
> +++ b/src/box/sql.c
> @@ -579,8 +579,10 @@ int tarantoolSqlite3EphemeralClearTable(BtCursor *pCur)
> * Removes all instances from table.
> * Iterate through the space and delete one by one all tuples.
> */
> -int tarantoolSqlite3ClearTable(struct space *space)
> +int tarantoolSqlite3ClearTable(struct space *space, uint32_t *tuple_count)
> {
> + assert(tuple_count != NULL);
> + *tuple_count = 0;
> uint32_t key_size;
> box_tuple_t *tuple;
> int rc;
> @@ -602,6 +604,7 @@ int tarantoolSqlite3ClearTable(struct space *space)
> iterator_delete(iter);
> return SQL_TARANTOOL_DELETE_FAIL;
> }
> + (*tuple_count)++;
> }
> iterator_delete(iter);
>
> diff --git a/src/box/sql/delete.c b/src/box/sql/delete.c
> index 116832f90..f9c42fdec 100644
> --- a/src/box/sql/delete.c
> +++ b/src/box/sql/delete.c
> @@ -210,6 +210,7 @@ sql_table_delete_from(struct Parse *parse, struct SrcList *tab_list,
> assert(!is_view);
>
> sqlite3VdbeAddOp1(v, OP_Clear, space->def->id);
> + sqlite3VdbeChangeP5(v, OPFLAG_NCHANGE);
>
> /* Do not start Tarantool's transaction in case of
> * truncate optimization. This is workaround until
> diff --git a/src/box/sql/tarantoolInt.h b/src/box/sql/tarantoolInt.h
> index 1ba8476fc..3ff14d53a 100644
> --- a/src/box/sql/tarantoolInt.h
> +++ b/src/box/sql/tarantoolInt.h
> @@ -58,7 +58,7 @@ int
> sql_delete_by_key(struct space *space, uint32_t iid, char *key,
> uint32_t key_size);
>
> -int tarantoolSqlite3ClearTable(struct space *space);
> +int tarantoolSqlite3ClearTable(struct space *space, uint32_t *tuple_count);
>
> /**
> * Rename the table in _space. Update tuple with corresponding id
> diff --git a/src/box/sql/vdbe.c b/src/box/sql/vdbe.c
> index b6afe9184..ff8bf2afc 100644
> --- a/src/box/sql/vdbe.c
> +++ b/src/box/sql/vdbe.c
> @@ -4644,7 +4644,7 @@ case OP_IdxGE: { /* jump */
> break;
> }
>
> -/* Opcode: Clear P1 P2 * * *
> +/* Opcode: Clear P1 P2 * * P5
> * Synopsis: space id = P1
> * If P2 is not 0, use Truncate semantics.
> *
> @@ -4652,6 +4652,9 @@ case OP_IdxGE: { /* jump */
> * in P1 argument. It is worth mentioning, that clearing routine
> * doesn't involve truncating, since it features completely
> * different mechanism under hood.
> + *
> + * If the OPFLAG_NCHANGE flag is set, then the row change count
> + * is incremented by the number of deleted tuples.
> */
> case OP_Clear: {
> assert(pOp->p1 > 0);
> @@ -4663,7 +4666,10 @@ case OP_Clear: {
> if (box_truncate(space_id) != 0)
> rc = SQL_TARANTOOL_ERROR;
> } else {
> - rc = tarantoolSqlite3ClearTable(space);
> + uint32_t tuple_count;
> + rc = tarantoolSqlite3ClearTable(space, &tuple_count);
> + if (rc == 0 && (pOp->p5 & OPFLAG_NCHANGE) != 0)
> + p->nChange += tuple_count;
> }
> if (rc) goto abort_due_to_error;
> break;
> diff --git a/test/sql/row-count.result b/test/sql/row-count.result
> index 7577d2795..d4c86ac2b 100644
> --- a/test/sql/row-count.result
> +++ b/test/sql/row-count.result
> @@ -100,6 +100,60 @@ box.sql.execute("SELECT ROW_COUNT();")
> ---
> - - [3]
> ...
> +-- gh-3816: DELETE optimization returns valid number of
> +-- deleted tuples.
> +--
> +box.sql.execute("DELETE FROM t3 WHERE 0 = 0;")
> +---
> +...
> +box.sql.execute("SELECT ROW_COUNT();")
> +---
> +- - [3]
> +...
> +box.sql.execute("INSERT INTO t3 VALUES (1, 1), (2, 2), (3, 3);")
> +---
> +...
> +box.sql.execute("DELETE FROM t3")
> +---
> +...
> +box.sql.execute("SELECT ROW_COUNT();")
> +---
> +- - [3]
> +...
> +-- But triggers still should't be accounted.
> +--
> +box.sql.execute("CREATE TABLE tt1 (id INT PRIMARY KEY);")
> +---
> +...
> +box.sql.execute("CREATE TABLE tt2 (id INT PRIMARY KEY);")
> +---
> +...
> +box.sql.execute("CREATE TRIGGER tr1 AFTER DELETE ON tt1 BEGIN DELETE FROM tt2; END;")
> +---
> +...
> +box.sql.execute("INSERT INTO tt1 VALUES (1), (2), (3);")
> +---
> +...
> +box.sql.execute("INSERT INTO tt2 VALUES (1), (2), (3);")
> +---
> +...
> +box.sql.execute("DELETE FROM tt1 WHERE id = 2;")
> +---
> +...
> +box.sql.execute("SELECT ROW_COUNT();")
> +---
> +- - [1]
> +...
> +box.sql.execute("SELECT * FROM tt2;")
> +---
> +- []
> +...
> +box.sql.execute("DROP TABLE tt1;")
> +---
> +...
> +box.sql.execute("DROP TABLE tt2;")
> +---
> +...
> -- All statements which are not accounted as DML should
> -- return 0 (zero) as a row count.
> --
> diff --git a/test/sql/row-count.test.lua b/test/sql/row-count.test.lua
> index c29e1d051..45a39d19a 100644
> --- a/test/sql/row-count.test.lua
> +++ b/test/sql/row-count.test.lua
> @@ -31,6 +31,26 @@ box.sql.execute("SELECT ROW_COUNT();")
> box.sql.execute("INSERT INTO t3 VALUES (1, 1), (2, 2), (3, 3);")
> box.sql.execute("UPDATE t3 SET i2 = 666;")
> box.sql.execute("SELECT ROW_COUNT();")
> +-- gh-3816: DELETE optimization returns valid number of
> +-- deleted tuples.
> +--
> +box.sql.execute("DELETE FROM t3 WHERE 0 = 0;")
> +box.sql.execute("SELECT ROW_COUNT();")
> +box.sql.execute("INSERT INTO t3 VALUES (1, 1), (2, 2), (3, 3);")
> +box.sql.execute("DELETE FROM t3")
> +box.sql.execute("SELECT ROW_COUNT();")
> +-- But triggers still should't be accounted.
> +--
> +box.sql.execute("CREATE TABLE tt1 (id INT PRIMARY KEY);")
> +box.sql.execute("CREATE TABLE tt2 (id INT PRIMARY KEY);")
> +box.sql.execute("CREATE TRIGGER tr1 AFTER DELETE ON tt1 BEGIN DELETE FROM tt2; END;")
> +box.sql.execute("INSERT INTO tt1 VALUES (1), (2), (3);")
> +box.sql.execute("INSERT INTO tt2 VALUES (1), (2), (3);")
> +box.sql.execute("DELETE FROM tt1 WHERE id = 2;")
> +box.sql.execute("SELECT ROW_COUNT();")
> +box.sql.execute("SELECT * FROM tt2;")
> +box.sql.execute("DROP TABLE tt1;")
> +box.sql.execute("DROP TABLE tt2;")
>
> -- All statements which are not accounted as DML should
> -- return 0 (zero) as a row count.
>
next prev parent reply other threads:[~2018-11-19 12:50 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-11-19 12:12 [tarantool-patches] " Nikita Pettik
2018-11-19 12:50 ` Vladislav Shpilevoy [this message]
2018-11-23 5:51 ` [tarantool-patches] " Kirill Yukhin
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=98a013cf-1c67-a5e9-0006-8384e57c4917@tarantool.org \
--to=v.shpilevoy@tarantool.org \
--cc=korablev@tarantool.org \
--cc=tarantool-patches@freelists.org \
--subject='[tarantool-patches] Re: [PATCH] sql: fix row count calculation for DELETE optimization' \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox