From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from [87.239.111.99] (localhost [127.0.0.1]) by dev.tarantool.org (Postfix) with ESMTP id 4FAAB6EC5B; Tue, 30 Mar 2021 02:01:10 +0300 (MSK) DKIM-Filter: OpenDKIM Filter v2.11.0 dev.tarantool.org 4FAAB6EC5B DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=tarantool.org; s=dev; t=1617058870; bh=z6wnADq3+z458xAJvQYFUtR8MoIezxP37bP/5khCyYE=; h=To:Cc:References:Date:In-Reply-To:Subject:List-Id: List-Unsubscribe:List-Archive:List-Post:List-Help:List-Subscribe: From:Reply-To:From; b=oDNWe1Pm/+Nkrm84OYOk9a6HifN3ojt7XHVYUcjjJRlvh1uE3VqXb8ky/0mpUGZBj L8tR52AUuvlh/lGUd4bKqos5pO8g324QIyWZowUC4pRbNXcF2DxDAHLQ2S5e4E2x5k aQedNFc7q2/MMpjMPll0XZd4Zbvn4MoqP4GIKETc= Received: from smtpng3.m.smailru.net (smtpng3.m.smailru.net [94.100.177.149]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by dev.tarantool.org (Postfix) with ESMTPS id 0F82F6EC5B for ; Tue, 30 Mar 2021 02:01:09 +0300 (MSK) DKIM-Filter: OpenDKIM Filter v2.11.0 dev.tarantool.org 0F82F6EC5B Received: by smtpng3.m.smailru.net with esmtpa (envelope-from ) id 1lR0se-0007iH-6R; Tue, 30 Mar 2021 02:01:08 +0300 To: imeevma@tarantool.org, tsafin@tarantool.org Cc: tarantool-patches@dev.tarantool.org References: Message-ID: Date: Tue, 30 Mar 2021 01:01:07 +0200 User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:78.0) Gecko/20100101 Thunderbird/78.9.0 MIME-Version: 1.0 In-Reply-To: Content-Type: text/plain; charset=utf-8 Content-Language: en-US Content-Transfer-Encoding: 7bit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eAau8CL7WIMRKs4sN3D3tLDjz0dLbV79QFUyzQ2Ujvy7cMT6pYYqY16iZVKkSc3dCLJ7zSJH7+u4VD18S7Vl4ZUrpaVfd2+vE6kuoey4m4VkSEu530nj6fImhcD4MUrOEAnl0W826KZ9Q+tr5ycPtXkTV4k65bRjmOUUP8cvGozZ33TWg5HZplvhhXbhDGzqmQDTd6OAevLeAnq3Ra9uf7zvY2zzsIhlcp/Y7m53TZgf2aB4JOg4gkr2biojljIiQOC84rTk2nYgSSNSJw== X-Mailru-Sender: 689FA8AB762F73936BC43F508A063822B474D78C2D204418D490F57A6F4649843841015FED1DE5223CC9A89AB576DD93FB559BB5D741EB963CF37A108A312F5C27E8A8C3839CE0E267EA787935ED9F1B X-Mras: Ok Subject: Re: [Tarantool-patches] [PATCH v4 11/53] sql: introduce mem_is_*() functions() X-BeenThere: tarantool-patches@dev.tarantool.org X-Mailman-Version: 2.1.34 Precedence: list List-Id: Tarantool development patches List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , From: Vladislav Shpilevoy via Tarantool-patches Reply-To: Vladislav Shpilevoy Errors-To: tarantool-patches-bounces@dev.tarantool.org Sender: "Tarantool-patches" Thanks for working on this! I have a general comment affecting this entire naming schema. The names seem too long. Besides, we already had some mem_is_*() and mem_set_*() functions before this patch, which used short names for the types. What I mean is 'integer' -> 'int', 'boolean' -> 'bool', 'string' -> 'str', 'binary' -> 'bin', and so on. For the integers we have several functions because we split unsigned, signed, and always negative integers. So we would need more int-like names. For instance, mem_set_uint(uint64_t) - for MEM_UInt. mem_set_nint(int64_t) - for MEM_Int. mem_set_int(int64_t) - for both, checks the sign inside. mem_set_sint(int64_t, bool) - for both, takes the sign flag in the second argument This can be discussed. The main point - shorter is better IMO. See 14 comments below. On 23.03.2021 10:35, Mergen Imeev via Tarantool-patches wrote: > This patch introduces mem_is_*() functions that allows to check current > MEM state. > > Part of #5818 > --- > diff --git a/src/box/sql/func.c b/src/box/sql/func.c > index e600a9800..81b537d9b 100644 > --- a/src/box/sql/func.c > +++ b/src/box/sql/func.c > @@ -736,8 +736,8 @@ substrFunc(sql_context * context, int argc, sql_value ** argv) > context->is_aborted = true; > return; > } > - if (sql_value_is_null(argv[1]) > - || (argc == 3 && sql_value_is_null(argv[2])) > + if (mem_is_null(argv[1]) > + || (argc == 3 && mem_is_null(argv[2])) 1. This is not movement of huge code blocks, it is rather new code now. And in the new code better use our code style. Such as - || goes in the end of line, not in the beginning of a next line; - Unary operators don't have a whitespace after them; - Comparison with NULL should be explicit, no implicit boot casts. I mark the places below which I was able to find after a swift look. > ) { > return; > } > @@ -1578,13 +1576,13 @@ replaceFunc(sql_context * context, int argc, sql_value ** argv) > assert(zStr == sql_value_text(argv[0])); /* No encoding change */ > zPattern = sql_value_text(argv[1]); > if (zPattern == 0) { > - assert(sql_value_is_null(argv[1]) > + assert(mem_is_null(argv[1]) > || sql_context_db_handle(context)->mallocFailed); > return; > } > nPattern = sql_value_bytes(argv[1]); > if (nPattern == 0) { > - assert(! sql_value_is_null(argv[1])); > + assert(! mem_is_null(argv[1])); 2. Whitespace after unary operator. > sql_result_value(context, argv[0]); > return; > } > @@ -2039,7 +2035,7 @@ countStep(sql_context * context, int argc, sql_value ** argv) > return; > } > p = sql_aggregate_context(context, sizeof(*p)); > - if ((argc == 0 || ! sql_value_is_null(argv[0])) && p) { > + if ((argc == 0 || ! mem_is_null(argv[0])) && p) { 3. Ditto. > p->n++; > } > } > diff --git a/src/box/sql/mem.c b/src/box/sql/mem.c > index ec6aaab64..abc9291ef 100644 > --- a/src/box/sql/mem.c > +++ b/src/box/sql/mem.c > @@ -37,6 +37,142 @@ > #include "box/tuple.h" > #include "mpstream/mpstream.h" > > +bool > +mem_is_null(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Null) != 0; > +} 4. Maybe better move them all to mem.h. These one-liners easily can be inlined (the ones which are <= 3 lines long could be moved). > + > +bool > +mem_is_unsigned(const struct Mem *mem) > +{ > + return (mem->flags & MEM_UInt) != 0; > +} > + > +bool > +mem_is_string(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Str) != 0; > +} > + > +bool > +mem_is_number(const struct Mem *mem) > +{ > + return (mem->flags & (MEM_Real | MEM_Int |MEM_UInt)) != 0; 5. Missed whitespace after the last '|'. > +} > + > +bool > +mem_is_double(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Real) != 0; > +} > + > +bool > +mem_is_integer(const struct Mem *mem) > +{ > + return (mem->flags & (MEM_Int | MEM_UInt)) != 0; > +} > + > +bool > +mem_is_boolean(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Bool) != 0; > +} > + > +bool > +mem_is_binary(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Blob) != 0; > +} > + > +bool > +mem_is_map(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Blob) != 0 && > + (mem->flags & MEM_Subtype) != 0 && > + mem->subtype == SQL_SUBTYPE_MSGPACK && > + mp_typeof(*mem->z) == MP_MAP; > +} > + > +bool > +mem_is_array(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Blob) != 0 && > + (mem->flags & MEM_Subtype) != 0 && > + mem->subtype == SQL_SUBTYPE_MSGPACK && > + mp_typeof(*mem->z) == MP_ARRAY; > +} > + > +bool > +mem_is_aggregate(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Agg) != 0; > +} > + > +bool > +mem_is_varstring(const struct Mem *mem) > +{ > + return (mem->flags & (MEM_Blob | MEM_Str)) != 0; 6. It does not look right to call it varstring if it includes binary. A string is always binary, but not each binary object is a string. Maybe mem_is_bytes()? mem_is_bytearray()? > +} > + > +bool > +mem_is_frame(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Frame) != 0; > +} > + > +bool > +mem_is_undefined(const struct Mem *mem) > +{ > + return (mem->flags & MEM_Undefined) != 0; > +} > + > +bool > +mem_is_static(const struct Mem *mem) > +{ > + return (mem->flags & (MEM_Str | MEM_Blob)) != 0 && > + (mem->flags & MEM_Static) != 0; > +} > + > +bool > +mem_is_ephemeral(const struct Mem *mem) > +{ > + return (mem->flags & (MEM_Str | MEM_Blob)) != 0 && > + (mem->flags & MEM_Ephem) != 0; 7. How can it be that MEM_Ephem is set, but Str/Blob are not? > +} > + > +bool > +mem_is_dynamic(const struct Mem *mem) > +{ > + return (mem->flags & (MEM_Str | MEM_Blob)) != 0 && > + (mem->flags & MEM_Dyn) != 0; > +} > diff --git a/src/box/sql/vdbe.c b/src/box/sql/vdbe.c > index 12712efb4..05e0f78c1 100644 > --- a/src/box/sql/vdbe.c > +++ b/src/box/sql/vdbe.c > @@ -1088,10 +1086,10 @@ case OP_Concat: { /* same as TK_CONCAT, in1, in2, out3 */ > * Concatenation operation can be applied only to > * strings and blobs. > */ > - uint32_t str_type_p1 = pIn1->flags & (MEM_Blob | MEM_Str); > - uint32_t str_type_p2 = pIn2->flags & (MEM_Blob | MEM_Str); > - if (str_type_p1 == 0 || str_type_p2 == 0) { > - char *inconsistent_type = str_type_p1 == 0 ? > + bool str_type_p1 = mem_is_varstring(pIn1); > + bool str_type_p2 = mem_is_varstring(pIn2); 8. They are not types now. Only flags. Should be renamed to something more appropriate. > + if (!str_type_p1 || !str_type_p2) { > + char *inconsistent_type = !str_type_p1 ? > mem_type_to_str(pIn1) : > mem_type_to_str(pIn2); > diag_set(ClientError, ER_INCONSISTENT_TYPES, > @@ -1100,7 +1098,7 @@ case OP_Concat: { /* same as TK_CONCAT, in1, in2, out3 */ > } > > /* Moreover, both operands must be of the same type. */ > - if (str_type_p1 != str_type_p2) { > + if (mem_is_string(pIn1) != mem_is_string(pIn2)) { 9. I would recommend mem_is_same_type(). Up to you. > diag_set(ClientError, ER_INCONSISTENT_TYPES, > mem_type_to_str(pIn2), mem_type_to_str(pIn1)); > goto abort_due_to_error; > @@ -1186,14 +1183,16 @@ case OP_Remainder: { /* same as TK_REM, in1, in2, out3 */ > pIn2 = &aMem[pOp->p2]; > type2 = numericType(pIn2); > pOut = vdbe_prepare_null_out(p, pOp->p3); > - flags = pIn1->flags | pIn2->flags; > - if ((flags & MEM_Null)!=0) goto arithmetic_result_is_null; > + if (mem_is_null(pIn1) || mem_is_null(pIn2)) > + goto arithmetic_result_is_null; > if ((type1 & (MEM_Int | MEM_UInt)) != 0 && > (type2 & (MEM_Int | MEM_UInt)) != 0) { > iA = pIn1->u.i; > iB = pIn2->u.i; > - bool is_lhs_neg = pIn1->flags & MEM_Int; > - bool is_rhs_neg = pIn2->flags & MEM_Int; > + bool is_lhs_neg = mem_is_integer(pIn1) && > + !mem_is_unsigned(pIn1); > + bool is_rhs_neg = mem_is_integer(pIn2) && > + !mem_is_unsigned(pIn2); 10. The checks look overcomplicated. Worth adding mem_is_nint()? > bool is_res_neg; > switch( pOp->opcode) { > case OP_Add: { > @@ -1509,7 +1508,7 @@ case OP_ShiftRight: { /* same as TK_RSHIFT, in1, in2, out3 */ > pIn1 = &aMem[pOp->p1]; > pIn2 = &aMem[pOp->p2]; > pOut = vdbe_prepare_null_out(p, pOp->p3); > - if ((pIn1->flags | pIn2->flags) & MEM_Null) { > + if (mem_is_null(pIn1) || mem_is_null(pIn2)) { 11. This is at least third time I see the check of kind "one of them is null". And I see more below. Probably worth adding a function which would do it more efficient, in one check: mem_is_any_null(mem1, mem2) or something. Up to you. > @@ -1757,11 +1756,10 @@ case OP_Ge: { /* same as TK_GE, jump, in1, in3 */ > * or not both operands are null. > */ > assert(pOp->opcode==OP_Eq || pOp->opcode==OP_Ne); > - assert((flags1 & MEM_Cleared)==0); > + assert(!mem_is_cleared(pIn1)); > assert((pOp->p5 & SQL_JUMPIFNULL)==0); > - if ((flags1&flags3&MEM_Null)!=0 > - && (flags3&MEM_Cleared)==0 > - ) { > + if (mem_is_null(pIn1) && mem_is_null(pIn3) && 12. You already know pIn1 or pIn3 is NULL from the 'if' above. So it would be just a bit faster and easier to do mem_is_same_type(). Only one branch. > + !mem_is_cleared(pIn3)) { > res = 0; /* Operands are equal */ > } else { > res = 1; /* Operands are not equal */ > @@ -2982,18 +2965,18 @@ case OP_SeekGT: { /* jump, in3 */ > * the seek, so convert it. > */ > pIn3 = &aMem[int_field]; > - if ((pIn3->flags & MEM_Null) != 0) > + if (mem_is_null(pIn3)) > goto skip_truncate; > - if ((pIn3->flags & MEM_Str) != 0) > + if (mem_is_string(pIn3)) > mem_apply_numeric_type(pIn3); > int64_t i; > - if ((pIn3->flags & MEM_Int) == MEM_Int) { > + if (mem_is_integer(pIn3) && !mem_is_unsigned(pIn3)) { 13. Better be mem_is_nint(), mentioned in one of the previous comments. > @@ -3352,11 +3336,11 @@ case OP_FCopy: { /* out2 */ > pIn1 = &aMem[pOp->p1]; > } > > - if ((pOp->p3 & OPFLAG_NOOP_IF_NULL) && (pIn1->flags & MEM_Null)) { > + if ((pOp->p3 & OPFLAG_NOOP_IF_NULL) && mem_is_null(pIn1)) { 14. Should be explicit != 0. > pOut = vdbe_prepare_null_out(p, pOp->p2); > } else { > assert(memIsValid(pIn1)); > - assert((pIn1->flags & (MEM_Int | MEM_UInt)) != 0); > + assert(mem_is_integer(pIn1));