From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from localhost (localhost [127.0.0.1]) by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTP id 2F5B52B4A9 for ; Fri, 23 Mar 2018 12:01:24 -0400 (EDT) Received: from turing.freelists.org ([127.0.0.1]) by localhost (turing.freelists.org [127.0.0.1]) (amavisd-new, port 10024) with ESMTP id 5dQJz2uprltM for ; Fri, 23 Mar 2018 12:01:24 -0400 (EDT) Received: from smtp35.i.mail.ru (smtp35.i.mail.ru [94.100.177.95]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by turing.freelists.org (Avenir Technologies Mail Multiplex) with ESMTPS id 679BA26400 for ; Fri, 23 Mar 2018 12:01:22 -0400 (EDT) Content-Type: text/plain; charset=utf-8 Mime-Version: 1.0 (Mac OS X Mail 10.3 \(3273\)) Subject: [tarantool-patches] Re: [PATCH 1/4] sql: pass space pointer to OP_OpenRead/OpenWrite From: "n.pettik" In-Reply-To: Date: Fri, 23 Mar 2018 19:01:18 +0300 Content-Transfer-Encoding: quoted-printable Message-Id: References: <46d0750257b6b5a256210b063515b1b05f4d7d37.1521583434.git.korablev@tarantool.org> <20180321131450.otxkluudlhf3yh2i@tarantool.org> <7F0DB390-F7BA-4E1F-AB93-9DB8498BCFD5@tarantool.org> Sender: tarantool-patches-bounce@freelists.org Errors-to: tarantool-patches-bounce@freelists.org Reply-To: tarantool-patches@freelists.org List-help: List-unsubscribe: List-software: Ecartis version 1.0.0 List-Id: tarantool-patches List-subscribe: List-owner: List-post: List-archive: To: Vladislav Shpilevoy Cc: tarantool-patches@freelists.org > On 22 Mar 2018, at 14:04, v.shpilevoy@tarantool.org wrote: >=20 > See below 1 comment, and 1 off topic comment. >=20 > And look at travis - it fails on the branch. It was tricky bug which I didn=E2=80=99t manage to reproduce on macOS. Travis failed on box.cfg{} when initialising SQL subsystem. (Somehow I used sqlite3_free() on memory allocated by ordinary malloc.) >=20 >>=20 >> Originally in SQLite, to open table (i.e. btree) it was required to = pass >> number of page root to OP_OpenRead or OP_OpenWrite opcodes as an >> argument. However, now there are only Tarantool spaces and nothing >> prevents from operating directly on pointers to them. Thus, to pass >> pointers from compile time to runtime, opcode OP_LoadPtr has been >> introduced. It fetches pointer from P4 and stores to the register >> specified by P2. >> It is worth mentioning that, pointers are able to expire after schema >> changes (i.e. after DML routine). For this reason, schema version is >> saved to VDBE at compile time and checked each time during cursor >> opening. >>=20 >> Part of #3252 >> --- >> src/box/sql/analyze.c | 17 +++++- >> src/box/sql/build.c | 7 ++- >> src/box/sql/expr.c | 14 ++++- >> src/box/sql/fkey.c | 11 +++- >> src/box/sql/insert.c | 36 ++++++++++-- >> src/box/sql/opcodes.c | 137 = +++++++++++++++++++++---------------------- >> src/box/sql/opcodes.h | 157 = +++++++++++++++++++++++++------------------------- >> src/box/sql/select.c | 11 +++- >> src/box/sql/vdbe.c | 13 +++++ >> src/box/sql/vdbe.h | 2 + >> src/box/sql/vdbeInt.h | 3 + >> src/box/sql/vdbeaux.c | 11 ++++ >> src/box/sql/where.c | 12 +++- >> 13 files changed, 272 insertions(+), 159 deletions(-) >>=20 >> diff --git a/src/box/sql/analyze.c b/src/box/sql/analyze.c >> index db06d0182..d121dd2b9 100644 >> --- a/src/box/sql/analyze.c >> +++ b/src/box/sql/analyze.c >> @@ -174,10 +174,16 @@ openStatTable(Parse * pParse, /* Parsing = context */ >>=20 >> /* Open the sql_stat[134] tables for writing. */ >> for (i =3D 0; aTable[i]; i++) { >> + struct space *space =3D >> + space_by_id(SQLITE_PAGENO_TO_SPACEID(aRoot[i])); >> + assert(space !=3D NULL); >> + int space_ptr_reg =3D ++pParse->nMem; >> + sqlite3VdbeAddOp4Ptr(v, OP_LoadPtr, 0, space_ptr_reg, 0, >> + (void *) space); >> int addr; >> addr =3D >> sqlite3VdbeAddOp3(v, OP_OpenWrite, iStatCur + i, = aRoot[i], >> - 0); >> + space_ptr_reg); >=20 > 1. Why do you pass here p3, if it is not used in OP_OpenWrite? Why can = not you just use sqlite3VdbeAddOp2 here? Same about OpenRead in all the = diff below.=20 In this patch I don=E2=80=99t update OP_OpenWrite, since it will be = still incomplete. Only after introducing new DDL it will be possible to refactor = OP_OpenWrite. > And how about to wrap this pair of calls sqlite3VdbeAddOp4Ptr(space) + = sqlite3VdbeAddOp3(open_read/write) into a separate function? This code = is repeated many times, as I can see. Done (updated on branch): @@ -1160,6 +1160,30 @@ index_collation_name(Index *idx, uint32_t column) return index->def->key_def->parts[column].coll->name; } +/** + * Create cursor which will be positioned to the space/index. + * It makes space lookup and loads pointer to it into register, + * which is passes to OP_OpenWrite as an argument. + * + * @param parse_context Parse context. + * @param cursor Number of cursor to be created. + * @param entity_id Encoded space and index ids. + * @retval address of last opcode. + */ +int +emit_open_cursor(Parse *parse_context, int cursor, int entity_id) +{ + assert(entity_id > 0); + struct space *space =3D = space_by_id(SQLITE_PAGENO_TO_SPACEID(entity_id)); + assert(space !=3D NULL); + Vdbe *vdbe =3D parse_context->pVdbe; + int space_ptr_reg =3D ++parse_context->nMem; + sqlite3VdbeAddOp4Ptr(vdbe, OP_LoadPtr, 0, space_ptr_reg, 0, + (void *) space); + return sqlite3VdbeAddOp3(vdbe, OP_OpenWrite, cursor, entity_id, + space_ptr_reg); +} + >=20 >> v->aOp[addr].p4.pKeyInfo =3D 0; >> v->aOp[addr].p4type =3D P4_KEYINFO; >> sqlite3VdbeChangeP5(v, aCreateTbl[i]); >> @@ -814,6 +820,7 @@ analyzeOneTable(Parse * pParse, /* Parser = context */ >> int iTabCur; /* Table cursor */ >> Vdbe *v; /* The virtual machine being built up */ >> int i; /* Loop counter */ >> + int space_ptr_reg =3D iMem++; >=20 > Off topic: How about create a mempool of Mems? I see, that this object = is created really often and in big amount (see = src/lib/small/small/mempool.h). (It is not a part of the patch - just = think about it, and possibly open a ticket, if it worth). Memory for =E2=80=99struct Mem=E2=80=99 objects is allocated at once (at = sqlite3VdbeMakeReady()), so I don=E2=80=99t think that there will be any = benefit from mempool.