diff options
| author | Andres Freund <andres@anarazel.de> | 2017-03-14 15:45:36 -0700 |
|---|---|---|
| committer | Andres Freund <andres@anarazel.de> | 2017-03-25 14:52:06 -0700 |
| commit | b8d7f053c5c2bf2a7e8734fe3327f6a8bc711755 (patch) | |
| tree | 6fd5db4d05a3dec9bed6b8cc4c98ca9d3f80425e /src/backend/utils/adt | |
| parent | 7d3957e53ebf26fc8d72dee1dacc2c827cc07caa (diff) | |
| download | postgresql-b8d7f053c5c2bf2a7e8734fe3327f6a8bc711755.tar.gz | |
Faster expression evaluation and targetlist projection.
This replaces the old, recursive tree-walk based evaluation, with
non-recursive, opcode dispatch based, expression evaluation.
Projection is now implemented as part of expression evaluation.
This both leads to significant performance improvements, and makes
future just-in-time compilation of expressions easier.
The speed gains primarily come from:
- non-recursive implementation reduces stack usage / overhead
- simple sub-expressions are implemented with a single jump, without
function calls
- sharing some state between different sub-expressions
- reduced amount of indirect/hard to predict memory accesses by laying
out operation metadata sequentially; including the avoidance of
nearly all of the previously used linked lists
- more code has been moved to expression initialization, avoiding
constant re-checks at evaluation time
Future just-in-time compilation (JIT) has become easier, as
demonstrated by released patches intended to be merged in a later
release, for primarily two reasons: Firstly, due to a stricter split
between expression initialization and evaluation, less code has to be
handled by the JIT. Secondly, due to the non-recursive nature of the
generated "instructions", less performance-critical code-paths can
easily be shared between interpreted and compiled evaluation.
The new framework allows for significant future optimizations. E.g.:
- basic infrastructure for to later reduce the per executor-startup
overhead of expression evaluation, by caching state in prepared
statements. That'd be helpful in OLTPish scenarios where
initialization overhead is measurable.
- optimizing the generated "code". A number of proposals for potential
work has already been made.
- optimizing the interpreter. Similarly a number of proposals have
been made here too.
The move of logic into the expression initialization step leads to some
backward-incompatible changes:
- Function permission checks are now done during expression
initialization, whereas previously they were done during
execution. In edge cases this can lead to errors being raised that
previously wouldn't have been, e.g. a NULL array being coerced to a
different array type previously didn't perform checks.
- The set of domain constraints to be checked, is now evaluated once
during expression initialization, previously it was re-built
every time a domain check was evaluated. For normal queries this
doesn't change much, but e.g. for plpgsql functions, which caches
ExprStates, the old set could stick around longer. The behavior
around might still change.
Author: Andres Freund, with significant changes by Tom Lane,
changes by Heikki Linnakangas
Reviewed-By: Tom Lane, Heikki Linnakangas
Discussion: https://postgr.es/m/20161206034955.bh33paeralxbtluv@alap3.anarazel.de
Diffstat (limited to 'src/backend/utils/adt')
| -rw-r--r-- | src/backend/utils/adt/domains.c | 29 | ||||
| -rw-r--r-- | src/backend/utils/adt/ruleutils.c | 4 | ||||
| -rw-r--r-- | src/backend/utils/adt/xml.c | 45 |
3 files changed, 36 insertions, 42 deletions
diff --git a/src/backend/utils/adt/domains.c b/src/backend/utils/adt/domains.c index c2ad440013..73deaa7e1c 100644 --- a/src/backend/utils/adt/domains.c +++ b/src/backend/utils/adt/domains.c @@ -107,7 +107,7 @@ domain_state_setup(Oid domainType, bool binary, MemoryContext mcxt) fmgr_info_cxt(my_extra->typiofunc, &my_extra->proc, mcxt); /* Look up constraints for domain */ - InitDomainConstraintRef(domainType, &my_extra->constraint_ref, mcxt); + InitDomainConstraintRef(domainType, &my_extra->constraint_ref, mcxt, true); /* We don't make an ExprContext until needed */ my_extra->econtext = NULL; @@ -122,7 +122,9 @@ domain_state_setup(Oid domainType, bool binary, MemoryContext mcxt) /* * domain_check_input - apply the cached checks. * - * This is extremely similar to ExecEvalCoerceToDomain in execQual.c. + * This is roughly similar to the handling of CoerceToDomain nodes in + * execExpr*.c, but we execute each constraint separately, rather than + * compiling them in-line within a larger expression. */ static void domain_check_input(Datum value, bool isnull, DomainIOData *my_extra) @@ -149,9 +151,6 @@ domain_check_input(Datum value, bool isnull, DomainIOData *my_extra) break; case DOM_CONSTRAINT_CHECK: { - Datum conResult; - bool conIsNull; - /* Make the econtext if we didn't already */ if (econtext == NULL) { @@ -165,24 +164,20 @@ domain_check_input(Datum value, bool isnull, DomainIOData *my_extra) /* * Set up value to be returned by CoerceToDomainValue - * nodes. Unlike ExecEvalCoerceToDomain, this econtext - * couldn't be shared with anything else, so no need to - * save and restore fields. But we do need to protect the - * passed-in value against being changed by called - * functions. (It couldn't be a R/W expanded object for - * most uses, but that seems possible for domain_check().) + * nodes. Unlike in the generic expression case, this + * econtext couldn't be shared with anything else, so no + * need to save and restore fields. But we do need to + * protect the passed-in value against being changed by + * called functions. (It couldn't be a R/W expanded + * object for most uses, but that seems possible for + * domain_check().) */ econtext->domainValue_datum = MakeExpandedObjectReadOnly(value, isnull, my_extra->constraint_ref.tcache->typlen); econtext->domainValue_isNull = isnull; - conResult = ExecEvalExprSwitchContext(con->check_expr, - econtext, - &conIsNull); - - if (!conIsNull && - !DatumGetBool(conResult)) + if (!ExecCheck(con->check_exprstate, econtext)) ereport(ERROR, (errcode(ERRCODE_CHECK_VIOLATION), errmsg("value for domain %s violates check constraint \"%s\"", diff --git a/src/backend/utils/adt/ruleutils.c b/src/backend/utils/adt/ruleutils.c index 81c91039e4..d57d5568b2 100644 --- a/src/backend/utils/adt/ruleutils.c +++ b/src/backend/utils/adt/ruleutils.c @@ -7000,7 +7000,7 @@ find_param_referent(Param *param, deparse_context *context, foreach(lc2, ps->subPlan) { SubPlanState *sstate = (SubPlanState *) lfirst(lc2); - SubPlan *subplan = (SubPlan *) sstate->xprstate.expr; + SubPlan *subplan = sstate->subplan; ListCell *lc3; ListCell *lc4; @@ -7041,7 +7041,7 @@ find_param_referent(Param *param, deparse_context *context, continue; /* No parameters to be had here. */ - Assert(((SubPlan *) sstate->xprstate.expr)->parParam == NIL); + Assert(sstate->subplan->parParam == NIL); /* Keep looking, but we are emerging from an initplan. */ in_same_plan_level = false; diff --git a/src/backend/utils/adt/xml.c b/src/backend/utils/adt/xml.c index 1908b13db5..2f87151bec 100644 --- a/src/backend/utils/adt/xml.c +++ b/src/backend/utils/adt/xml.c @@ -72,7 +72,6 @@ #include "catalog/pg_class.h" #include "catalog/pg_type.h" #include "commands/dbcommands.h" -#include "executor/executor.h" #include "executor/spi.h" #include "executor/tablefunc.h" #include "fmgr.h" @@ -620,10 +619,11 @@ xmltotext_with_xmloption(xmltype *data, XmlOptionType xmloption_arg) xmltype * -xmlelement(XmlExprState *xmlExpr, ExprContext *econtext) +xmlelement(XmlExpr *xexpr, + Datum *named_argvalue, bool *named_argnull, + Datum *argvalue, bool *argnull) { #ifdef USE_LIBXML - XmlExpr *xexpr = (XmlExpr *) xmlExpr->xprstate.expr; xmltype *result; List *named_arg_strings; List *arg_strings; @@ -635,48 +635,47 @@ xmlelement(XmlExprState *xmlExpr, ExprContext *econtext) volatile xmlTextWriterPtr writer = NULL; /* - * We first evaluate all the arguments, then start up libxml and create - * the result. This avoids issues if one of the arguments involves a call - * to some other function or subsystem that wants to use libxml on its own - * terms. + * All arguments are already evaluated, and their values are passed in the + * named_argvalue/named_argnull or argvalue/argnull arrays. This avoids + * issues if one of the arguments involves a call to some other function + * or subsystem that wants to use libxml on its own terms. We examine the + * original XmlExpr to identify the numbers and types of the arguments. */ named_arg_strings = NIL; i = 0; - foreach(arg, xmlExpr->named_args) + foreach(arg, xexpr->named_args) { - ExprState *e = (ExprState *) lfirst(arg); - Datum value; - bool isnull; + Expr *e = (Expr *) lfirst(arg); char *str; - value = ExecEvalExpr(e, econtext, &isnull); - if (isnull) + if (named_argnull[i]) str = NULL; else - str = map_sql_value_to_xml_value(value, exprType((Node *) e->expr), false); + str = map_sql_value_to_xml_value(named_argvalue[i], + exprType((Node *) e), + false); named_arg_strings = lappend(named_arg_strings, str); i++; } arg_strings = NIL; - foreach(arg, xmlExpr->args) + i = 0; + foreach(arg, xexpr->args) { - ExprState *e = (ExprState *) lfirst(arg); - Datum value; - bool isnull; + Expr *e = (Expr *) lfirst(arg); char *str; - value = ExecEvalExpr(e, econtext, &isnull); /* here we can just forget NULL elements immediately */ - if (!isnull) + if (!argnull[i]) { - str = map_sql_value_to_xml_value(value, - exprType((Node *) e->expr), true); + str = map_sql_value_to_xml_value(argvalue[i], + exprType((Node *) e), + true); arg_strings = lappend(arg_strings, str); } + i++; } - /* now safe to run libxml */ xmlerrcxt = pg_xml_init(PG_XML_STRICTNESS_ALL); PG_TRY(); |
