From: Mergen Imeev via Tarantool-patches <tarantool-patches@dev.tarantool.org>
To: Vladislav Shpilevoy <v.shpilevoy@tarantool.org>
Cc: tarantool-patches@dev.tarantool.org
Subject: Re: [Tarantool-patches] [PATCH v5 27/52] sql: introduce mem_set_str_*() functions
Date: Thu, 15 Apr 2021 04:25:54 +0300 [thread overview]
Message-ID: <20210415012554.GA416841@tarantool.org> (raw)
In-Reply-To: <f039216f-5ade-ebf1-5d66-606f67735728@tarantool.org>
Thank you for the review. My answer and two diffs below. First diff for this
patch, second - for similar patch with binary.
On Thu, Apr 15, 2021 at 01:49:46AM +0200, Vladislav Shpilevoy wrote:
> Thanks for working on this!
>
> >>> index ae55e4c29..171cb8946 100644
> >>> --- a/src/box/sql/vdbeaux.c
> >>> +++ b/src/box/sql/vdbeaux.c
> >>> @@ -1333,41 +1331,34 @@ sqlVdbeList(Vdbe * p)
> >>> mem_set_int(pMem, pOp->p3, pOp->p3 < 0);
> >>> pMem++;
> >>>
> >>> - if (sqlVdbeMemClearAndResize(pMem, 256)) {
> >>> - assert(p->db->mallocFailed);
> >>> + char *buf = sqlDbMallocRaw(sql_get(), 256);
> >>
> >> 2. I think you need some kind of mem_set_strlen(), or mem_grow()/mem_reserve(),
> >> or something else to reserve the memory. To extend zMalloc. Otherwise you
> >> can't reuse the memory which might already be in the mem object.
> > Wouldn't mem_copy_*() be enough? In general, allocated by MEM memory should not
> > be accessed for changing (except for MEM_Agg).
>
> Both your current solution and mem_copy suffer from unnecessary copying.
> In the old code the string was created in-place. This is what I am talking
> about.
>
> But nevermind, it is not a perf-critical place as I see. Can be tweaked
> later.
>
> > diff --git a/src/box/sql/vdbeapi.c b/src/box/sql/vdbeapi.c
> > index 6c08e772d..484c66b29 100644
> > --- a/src/box/sql/vdbeapi.c
> > +++ b/src/box/sql/vdbeapi.c
> > @@ -125,6 +125,12 @@ setResultStrOrError(sql_context * pCtx, /* Function context */
> > void (*xDel) (void *) /* Destructor function */
> > )
> > {
> > + if (xDel == SQL_STATIC)
> > + return mem_set_strl_static(pCtx->pOut, (char *)z, n);
> > + if (xDel == SQL_DYNAMIC)
> > + return mem_set_strl_allocated(pCtx->pOut, (char *)z, n);
> > + if (xDel != SQL_TRANSIENT)
> > + return mem_set_strl_dynamic(pCtx->pOut, (char *)z, n);
>
> I just can't help myself. It makes my eyes bleed - for dynamic we use
> allocated, and for non-transient we use dynamic! I see why, but I would
> want us to hide it as deep inside of mem.c as possible, so as we at
> least could change the names in a not very intrusive manner in the
> future. Or delete it in future entirely.
>
> Please, move this hunk and the one below to a separate functions
> like mem_set_strl(Mem, str, len_hint, destructor). Then it would be in
> one terrible place instead of 2 terrible places.
>
> The same for the bins.
>
Done.
> > if (sqlVdbeMemSetStr(pCtx->pOut, z, n, 1, xDel) != 0)
> > pCtx->is_aborted = true;
> > }
> > @@ -762,7 +768,13 @@ bindText(sql_stmt * pStmt, /* The statement to bind against */
> > if (zData == NULL)
> > return 0;
> > pVar = &p->aVar[i - 1];
> > - if (sqlVdbeMemSetStr(pVar, zData, nData, 1, xDel) != 0)
> > + if (xDel == SQL_STATIC)
> > + mem_set_strl_static(pVar, (char *)zData, nData);
> > + else if (xDel == SQL_DYNAMIC)
> > + mem_set_strl_allocated(pVar, (char *)zData, nData);
> > + else if (xDel != SQL_TRANSIENT)
> > + mem_set_strl_dynamic(pVar, (char *)zData, nData);
> > + else if (sqlVdbeMemSetStr(pVar, zData, nData, 1, xDel) != 0)
> > return -1;
> > return sql_bind_type(p, i, "text");
> > }
Diff for str:
diff --git a/src/box/sql/mem.h b/src/box/sql/mem.h
index f029ee0e5..258bb11e6 100644
--- a/src/box/sql/mem.h
+++ b/src/box/sql/mem.h
@@ -409,6 +409,18 @@ mem_set_strl_allocated(struct Mem *mem, char *value, int len_hint)
mem_set_str_allocated(mem, value, len_hint);
}
+static inline void
+mem_set_strl(struct Mem *mem, char *value, int len_hint,
+ void (*custom_free)(void *))
+{
+ if (custom_free == SQL_STATIC)
+ return mem_set_strl_static(mem, value, len_hint);
+ if (custom_free == SQL_DYNAMIC)
+ return mem_set_strl_allocated(mem, value, len_hint);
+ if (custom_free != SQL_TRANSIENT)
+ return mem_set_strl_dynamic(mem, value, len_hint);
+}
+
/**
* Copy content of MEM from one MEM to another. In case source MEM contains
* string or binary and allocation type is not STATIC, this value is copied to
diff --git a/src/box/sql/vdbeapi.c b/src/box/sql/vdbeapi.c
index 3cd60864f..ae65a53d8 100644
--- a/src/box/sql/vdbeapi.c
+++ b/src/box/sql/vdbeapi.c
@@ -125,12 +125,8 @@ setResultStrOrError(sql_context * pCtx, /* Function context */
void (*xDel) (void *) /* Destructor function */
)
{
- if (xDel == SQL_STATIC)
- return mem_set_strl_static(pCtx->pOut, (char *)z, n);
- if (xDel == SQL_DYNAMIC)
- return mem_set_strl_allocated(pCtx->pOut, (char *)z, n);
if (xDel != SQL_TRANSIENT)
- return mem_set_strl_dynamic(pCtx->pOut, (char *)z, n);
+ return mem_set_strl(pCtx->pOut, (char *)z, n, xDel);
if (sqlVdbeMemSetStr(pCtx->pOut, z, n, 1, xDel) != 0)
pCtx->is_aborted = true;
}
@@ -768,12 +764,8 @@ bindText(sql_stmt * pStmt, /* The statement to bind against */
if (zData == NULL)
return 0;
pVar = &p->aVar[i - 1];
- if (xDel == SQL_STATIC)
- mem_set_strl_static(pVar, (char *)zData, nData);
- else if (xDel == SQL_DYNAMIC)
- mem_set_strl_allocated(pVar, (char *)zData, nData);
- else if (xDel != SQL_TRANSIENT)
- mem_set_strl_dynamic(pVar, (char *)zData, nData);
+ if (xDel != SQL_TRANSIENT)
+ mem_set_strl(pVar, (char *)zData, nData, xDel);
else if (sqlVdbeMemSetStr(pVar, zData, nData, 1, xDel) != 0)
return -1;
return sql_bind_type(p, i, "text");
Diff for bin:
diff --git a/src/box/sql/mem.h b/src/box/sql/mem.h
index f728ac58a..5b9ac6a96 100644
--- a/src/box/sql/mem.h
+++ b/src/box/sql/mem.h
@@ -468,6 +468,18 @@ mem_set_bin_dynamic(struct Mem *mem, char *value, uint32_t size);
void
mem_set_bin_allocated(struct Mem *mem, char *value, uint32_t size);
+static inline void
+mem_set_binl(struct Mem *mem, char *value, uint32_t size,
+ void (*custom_free)(void *))
+{
+ if (custom_free == SQL_STATIC)
+ return mem_set_bin_static(mem, value, size);
+ if (custom_free == SQL_DYNAMIC)
+ return mem_set_bin_allocated(mem, value, size);
+ if (custom_free != SQL_TRANSIENT)
+ return mem_set_bin_dynamic(mem, value, size);
+}
+
/**
* Copy content of MEM from one MEM to another. In case source MEM contains
* string or binary and allocation type is not STATIC, this value is copied to
diff --git a/src/box/sql/vdbeapi.c b/src/box/sql/vdbeapi.c
index 1ba2f4260..66908e2ec 100644
--- a/src/box/sql/vdbeapi.c
+++ b/src/box/sql/vdbeapi.c
@@ -158,12 +158,8 @@ sql_result_blob(sql_context * pCtx,
)
{
assert(n >= 0);
- if (xDel == SQL_STATIC)
- mem_set_bin_static(pCtx->pOut, (char *)z, n);
- else if (xDel == SQL_DYNAMIC)
- mem_set_bin_allocated(pCtx->pOut, (char *)z, n);
- else if (xDel != SQL_TRANSIENT)
- mem_set_bin_dynamic(pCtx->pOut, (char *)z, n);
+ if (xDel != SQL_TRANSIENT)
+ mem_set_binl(pCtx->pOut, (char *)z, n, xDel);
else if (sqlVdbeMemSetStr(pCtx->pOut, z, n, 0, xDel) != 0)
pCtx->is_aborted = true;
}
@@ -793,12 +789,8 @@ sql_bind_blob(sql_stmt * pStmt,
if (zData == NULL)
return 0;
struct Mem *var = &p->aVar[i - 1];
- if (xDel == SQL_STATIC)
- mem_set_bin_static(var, (char *)zData, nData);
- else if (xDel == SQL_DYNAMIC)
- mem_set_bin_allocated(var, (char *)zData, nData);
- else if (xDel != SQL_TRANSIENT)
- mem_set_bin_dynamic(var, (char *)zData, nData);
+ if (xDel != SQL_TRANSIENT)
+ mem_set_binl(var, (char *)zData, nData, xDel);
else if (sqlVdbeMemSetStr(var, zData, nData, 0, xDel) != 0)
return -1;
return sql_bind_type(p, i, "varbinary");
next prev parent reply other threads:[~2021-04-15 1:25 UTC|newest]
Thread overview: 107+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-04-09 16:51 [Tarantool-patches] [PATCH v5 00/52] Move mem-related functions to mem.c/mem.h Mergen Imeev via Tarantool-patches
2021-04-09 16:51 ` [Tarantool-patches] [PATCH v5 01/52] sql: enhance vdbe_decode_msgpack_into_mem() Mergen Imeev via Tarantool-patches
2021-04-11 17:42 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 12:01 ` Mergen Imeev via Tarantool-patches
2021-04-13 12:12 ` Mergen Imeev via Tarantool-patches
2021-04-13 23:22 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 23:34 ` Mergen Imeev via Tarantool-patches
2021-04-09 16:51 ` [Tarantool-patches] [PATCH v5 02/52] sql: disable unused code in sql/analyze.c Mergen Imeev via Tarantool-patches
2021-04-09 16:51 ` [Tarantool-patches] [PATCH v5 03/52] sql: disable unused code in sql/legacy.c Mergen Imeev via Tarantool-patches
2021-04-09 16:51 ` [Tarantool-patches] [PATCH v5 04/52] sql: remove NULL-termination in OP_ResultRow Mergen Imeev via Tarantool-patches
2021-04-14 22:23 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 22:37 ` Mergen Imeev via Tarantool-patches
2021-04-09 16:51 ` [Tarantool-patches] [PATCH v5 05/52] sql: move MEM-related functions to mem.c/mem.h Mergen Imeev via Tarantool-patches
2021-04-09 16:59 ` [Tarantool-patches] [PATCH v5 06/52] sql: refactor port_vdbemem_*() functions Mergen Imeev via Tarantool-patches
2021-04-09 16:59 ` [Tarantool-patches] [PATCH v5 07/52] sql: remove unused MEM-related functions Mergen Imeev via Tarantool-patches
2021-04-09 16:59 ` [Tarantool-patches] [PATCH v5 08/52] sql: disable unused code in sql/vdbemem.c Mergen Imeev via Tarantool-patches
2021-04-09 16:59 ` [Tarantool-patches] [PATCH v5 09/52] sql: introduce mem_str() Mergen Imeev via Tarantool-patches
2021-04-11 17:44 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 12:36 ` Mergen Imeev via Tarantool-patches
2021-04-14 22:23 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 22:42 ` Mergen Imeev via Tarantool-patches
2021-04-09 16:59 ` [Tarantool-patches] [PATCH v5 10/52] sql: introduce mem_create() Mergen Imeev via Tarantool-patches
2021-04-09 17:36 ` [Tarantool-patches] [PATCH v5 11/52] sql: introduce mem_destroy() Mergen Imeev via Tarantool-patches
2021-04-11 17:46 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 12:42 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:36 ` [Tarantool-patches] [PATCH v5 12/52] sql: introduce mem_is_*() functions() Mergen Imeev via Tarantool-patches
2021-04-11 17:59 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 16:09 ` Mergen Imeev via Tarantool-patches
2021-04-14 22:48 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 23:07 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:36 ` [Tarantool-patches] [PATCH v5 13/52] sql: introduce mem_copy() Mergen Imeev via Tarantool-patches
2021-04-11 18:06 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 16:18 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:36 ` [Tarantool-patches] [PATCH v5 14/52] sql: introduce mem_copy_as_ephemeral() Mergen Imeev via Tarantool-patches
2021-04-11 18:10 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 16:31 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:37 ` [Tarantool-patches] [PATCH v5 15/52] sql: rework mem_move() Mergen Imeev via Tarantool-patches
2021-04-11 18:10 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 16:38 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:57 ` [Tarantool-patches] [PATCH v5 16/52] sql: rework vdbe_decode_msgpack_into_mem() Mergen Imeev via Tarantool-patches
2021-04-09 17:57 ` [Tarantool-patches] [PATCH v5 17/52] sql: remove sql_column_to_messagepack() Mergen Imeev via Tarantool-patches
2021-04-14 22:58 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 23:14 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:57 ` [Tarantool-patches] [PATCH v5 18/52] sql: introduce mem_concat() Mergen Imeev via Tarantool-patches
2021-04-11 18:11 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 16:57 ` Mergen Imeev via Tarantool-patches
2021-04-14 23:04 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 23:22 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:57 ` [Tarantool-patches] [PATCH v5 19/52] sql: introduce arithmetic operations for MEM Mergen Imeev via Tarantool-patches
2021-04-11 18:13 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 17:06 ` Mergen Imeev via Tarantool-patches
2021-04-14 23:10 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 23:33 ` Mergen Imeev via Tarantool-patches
2021-04-09 17:57 ` [Tarantool-patches] [PATCH v5 20/52] sql: introduce mem_compare() Mergen Imeev via Tarantool-patches
2021-04-11 18:16 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 18:33 ` Mergen Imeev via Tarantool-patches
2021-04-14 23:20 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-14 23:40 ` Mergen Imeev via Tarantool-patches
2021-04-09 18:11 ` [Tarantool-patches] [PATCH v5 21/52] sql: introduce bitwise operations for MEM Mergen Imeev via Tarantool-patches
2021-04-12 23:31 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 20:49 ` Mergen Imeev via Tarantool-patches
2021-04-09 18:11 ` [Tarantool-patches] [PATCH v5 22/52] sql: Initialize MEM in sqlVdbeAllocUnpackedRecord() Mergen Imeev via Tarantool-patches
2021-04-09 18:11 ` [Tarantool-patches] [PATCH v5 23/52] sql: introduce mem_set_null() Mergen Imeev via Tarantool-patches
2021-04-09 18:11 ` [Tarantool-patches] [PATCH v5 24/52] sql: introduce mem_set_int() Mergen Imeev via Tarantool-patches
2021-04-12 23:32 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 20:56 ` Mergen Imeev via Tarantool-patches
2021-04-09 18:11 ` [Tarantool-patches] [PATCH v5 25/52] sql: introduce mem_set_uint() Mergen Imeev via Tarantool-patches
2021-04-09 19:45 ` [Tarantool-patches] [PATCH v5 26/52] sql: move mem_set_bool() and mem_set_double() Mergen Imeev via Tarantool-patches
2021-04-09 19:45 ` [Tarantool-patches] [PATCH v5 27/52] sql: introduce mem_set_str_*() functions Mergen Imeev via Tarantool-patches
2021-04-12 23:34 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 21:36 ` Mergen Imeev via Tarantool-patches
2021-04-14 23:49 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-15 1:25 ` Mergen Imeev via Tarantool-patches [this message]
2021-04-09 19:45 ` [Tarantool-patches] [PATCH v5 28/52] sql: introduce mem_copy_str() and mem_copy_str0() Mergen Imeev via Tarantool-patches
2021-04-12 23:35 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:00 ` Mergen Imeev via Tarantool-patches
2021-04-14 23:54 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-15 0:30 ` Mergen Imeev via Tarantool-patches
2021-04-09 19:45 ` [Tarantool-patches] [PATCH v5 29/52] sql: introduce mem_set_bin_*() functions Mergen Imeev via Tarantool-patches
2021-04-09 19:45 ` [Tarantool-patches] [PATCH v5 30/52] sql: introduce mem_copy_bin() Mergen Imeev via Tarantool-patches
2021-04-12 23:36 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:06 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:05 ` [Tarantool-patches] [PATCH v5 31/52] sql: introduce mem_set_zerobin() Mergen Imeev via Tarantool-patches
2021-04-09 20:05 ` [Tarantool-patches] [PATCH v5 32/52] sql: introduce mem_set_*() for map and array Mergen Imeev via Tarantool-patches
2021-04-12 23:36 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:08 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:05 ` [Tarantool-patches] [PATCH v5 33/52] sql: introduce mem_set_invalid() Mergen Imeev via Tarantool-patches
2021-04-09 20:05 ` [Tarantool-patches] [PATCH v5 34/52] sql: refactor mem_set_ptr() Mergen Imeev via Tarantool-patches
2021-04-09 20:05 ` [Tarantool-patches] [PATCH v5 35/52] sql: introduce mem_set_frame() Mergen Imeev via Tarantool-patches
2021-04-12 23:37 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:19 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:25 ` [Tarantool-patches] [PATCH v5 36/52] sql: introduce mem_set_agg() Mergen Imeev via Tarantool-patches
2021-04-12 23:37 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:46 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:25 ` [Tarantool-patches] [PATCH v5 37/52] sql: introduce mem_set_null_clear() Mergen Imeev via Tarantool-patches
2021-04-12 23:38 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:50 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:25 ` [Tarantool-patches] [PATCH v5 38/52] sql: move MEM flags to mem.c Mergen Imeev via Tarantool-patches
2021-04-13 20:42 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:25 ` [Tarantool-patches] [PATCH v5 39/52] sql: introduce mem_to_int*() functions Mergen Imeev via Tarantool-patches
2021-04-12 23:39 ` Vladislav Shpilevoy via Tarantool-patches
2021-04-13 22:58 ` Mergen Imeev via Tarantool-patches
2021-04-13 23:10 ` Mergen Imeev via Tarantool-patches
2021-04-09 20:26 ` [Tarantool-patches] [PATCH v5 40/52] sql: introduce mem_to_double() Mergen Imeev via Tarantool-patches
2021-04-13 23:21 ` Mergen Imeev via Tarantool-patches
2021-04-15 0:39 ` [Tarantool-patches] [PATCH v5 00/52] Move mem-related functions to mem.c/mem.h Vladislav Shpilevoy via Tarantool-patches
2021-04-15 6:49 ` Kirill Yukhin via Tarantool-patches
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=20210415012554.GA416841@tarantool.org \
--to=tarantool-patches@dev.tarantool.org \
--cc=imeevma@tarantool.org \
--cc=v.shpilevoy@tarantool.org \
--subject='Re: [Tarantool-patches] [PATCH v5 27/52] sql: introduce mem_set_str_*() functions' \
/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