1
0
mirror of https://github.com/postgres/postgres.git synced 2025-11-07 19:06:32 +03:00

Code review for protransform patches.

Fix loss of previous expression-simplification work when a transform
function fires: we must not simply revert to untransformed input tree.
Instead build a dummy FuncExpr node to pass to the transform function.
This has the additional advantage of providing a simpler, more uniform
API for transform functions.

Move documentation to a somewhat less buried spot, relocate some
poorly-placed code, be more wary of null constants and invalid typmod
values, add an opr_sanity check on protransform function signatures,
and some other minor cosmetic adjustments.

Note: although this patch touches pg_proc.h, no need for catversion
bump, because the changes are cosmetic and don't actually change the
intended catalog contents.
This commit is contained in:
Tom Lane
2012-03-23 17:29:57 -04:00
parent e08b4101e1
commit 0339047bc9
15 changed files with 205 additions and 172 deletions

View File

@@ -17,6 +17,7 @@
#include "catalog/pg_collation.h"
#include "catalog/pg_type.h"
#include "miscadmin.h"
#include "nodes/makefuncs.h"
#include "nodes/nodeFuncs.h"
#include "nodes/relation.h"
#include "utils/builtins.h"
@@ -547,6 +548,30 @@ exprIsLengthCoercion(const Node *expr, int32 *coercedTypmod)
return false;
}
/*
* relabel_to_typmod
* Add a RelabelType node that changes just the typmod of the expression.
*
* This is primarily intended to be used during planning. Therefore, it
* strips any existing RelabelType nodes to maintain the planner's invariant
* that there are not adjacent RelabelTypes, and it uses COERCE_DONTCARE
* which would typically be inappropriate earlier.
*/
Node *
relabel_to_typmod(Node *expr, int32 typmod)
{
Oid type = exprType(expr);
Oid coll = exprCollation(expr);
/* Strip any existing RelabelType node(s) */
while (expr && IsA(expr, RelabelType))
expr = (Node *) ((RelabelType *) expr)->arg;
/* Apply new typmod, preserving the previous exposed type and collation */
return (Node *) makeRelabelType((Expr *) expr, type, typmod, coll,
COERCE_DONTCARE);
}
/*
* expression_returns_set
* Test whether an expression returns a set result.
@@ -2694,7 +2719,9 @@ query_or_expression_tree_mutator(Node *node,
* that could appear under it, but not other statement types.
*/
bool
raw_expression_tree_walker(Node *node, bool (*walker) (), void *context)
raw_expression_tree_walker(Node *node,
bool (*walker) (),
void *context)
{
ListCell *temp;

View File

@@ -107,11 +107,11 @@ static List *simplify_and_arguments(List *args,
eval_const_expressions_context *context,
bool *haveNull, bool *forceFalse);
static Node *simplify_boolean_equality(Oid opno, List *args);
static Expr *simplify_function(Expr *oldexpr, Oid funcid,
Oid result_type, int32 result_typmod, Oid result_collid,
Oid input_collid, List **args,
static Expr *simplify_function(Oid funcid,
Oid result_type, int32 result_typmod,
Oid result_collid, Oid input_collid, List **args,
bool has_named_args,
bool allow_inline,
bool allow_non_const,
eval_const_expressions_context *context);
static List *reorder_function_arguments(List *args, Oid result_type,
HeapTuple func_tuple,
@@ -2332,8 +2332,7 @@ eval_const_expressions_mutator(Node *node,
* length coercion; we want to preserve the typmod in the
* eventual Const if so.
*/
simple = simplify_function((Expr *) expr,
expr->funcid,
simple = simplify_function(expr->funcid,
expr->funcresulttype,
exprTypmod(node),
expr->funccollid,
@@ -2389,8 +2388,7 @@ eval_const_expressions_mutator(Node *node,
* Code for op/func reduction is pretty bulky, so split it out
* as a separate function.
*/
simple = simplify_function((Expr *) expr,
expr->opfuncid,
simple = simplify_function(expr->opfuncid,
expr->opresulttype, -1,
expr->opcollid,
expr->inputcollid,
@@ -2491,8 +2489,7 @@ eval_const_expressions_mutator(Node *node,
* Code for op/func reduction is pretty bulky, so split it
* out as a separate function.
*/
simple = simplify_function((Expr *) expr,
expr->opfuncid,
simple = simplify_function(expr->opfuncid,
expr->opresulttype, -1,
expr->opcollid,
expr->inputcollid,
@@ -2698,8 +2695,7 @@ eval_const_expressions_mutator(Node *node,
getTypeInputInfo(expr->resulttype,
&infunc, &intypioparam);
simple = simplify_function(NULL,
outfunc,
simple = simplify_function(outfunc,
CSTRINGOID, -1,
InvalidOid,
InvalidOid,
@@ -2728,8 +2724,7 @@ eval_const_expressions_mutator(Node *node,
false,
true));
simple = simplify_function(NULL,
infunc,
simple = simplify_function(infunc,
expr->resulttype, -1,
expr->resultcollid,
InvalidOid,
@@ -3581,15 +3576,11 @@ simplify_boolean_equality(Oid opno, List *args)
* Subroutine for eval_const_expressions: try to simplify a function call
* (which might originally have been an operator; we don't care)
*
* Inputs are the original expression (can be NULL), function OID, actual
* result type OID (which is needed for polymorphic functions), result typmod,
* result collation, the input collation to use for the function, the
* pre-simplified argument list, and some flags; also the context data for
* eval_const_expressions. In common cases, several of the arguments could be
* derived from the original expression. Sending them separately avoids
* duplicating NodeTag-specific knowledge, and it's necessary for CoerceViaIO.
* A NULL original expression disables use of transform functions while
* retaining all other behaviors.
* Inputs are the function OID, actual result type OID (which is needed for
* polymorphic functions), result typmod, result collation,
* the input collation to use for the function,
* the pre-simplified argument list, and some flags;
* also the context data for eval_const_expressions.
*
* Returns a simplified expression if successful, or NULL if cannot
* simplify the function call.
@@ -3601,28 +3592,32 @@ simplify_boolean_equality(Oid opno, List *args)
* pass-by-reference, and it may get modified even if simplification fails.
*/
static Expr *
simplify_function(Expr *oldexpr, Oid funcid,
Oid result_type, int32 result_typmod, Oid result_collid,
Oid input_collid, List **args,
simplify_function(Oid funcid, Oid result_type, int32 result_typmod,
Oid result_collid, Oid input_collid, List **args,
bool has_named_args,
bool allow_inline,
bool allow_non_const,
eval_const_expressions_context *context)
{
HeapTuple func_tuple;
Form_pg_proc func_form;
Expr *newexpr;
Oid transform;
/*
* We have three strategies for simplification: execute the function to
* deliver a constant result, use a transform function to generate a
* substitute node tree, or expand in-line the body of the function
* definition (which only works for simple SQL-language functions, but
* that is a common case). Each needs access to the function's pg_proc
* tuple, so fetch it just once.
* that is a common case). Each case needs access to the function's
* pg_proc tuple, so fetch it just once.
*
* Note: the allow_non_const flag suppresses both the second and third
* strategies; so if !allow_non_const, simplify_function can only return
* a Const or NULL. Argument-list rewriting happens anyway, though.
*/
func_tuple = SearchSysCache1(PROCOID, ObjectIdGetDatum(funcid));
if (!HeapTupleIsValid(func_tuple))
elog(ERROR, "cache lookup failed for function %u", funcid);
func_form = (Form_pg_proc) GETSTRUCT(func_tuple);
/*
* While we have the tuple, reorder named arguments and add default
@@ -3631,48 +3626,38 @@ simplify_function(Expr *oldexpr, Oid funcid,
if (has_named_args)
*args = reorder_function_arguments(*args, result_type, func_tuple,
context);
else if (((Form_pg_proc) GETSTRUCT(func_tuple))->pronargs > list_length(*args))
else if (func_form->pronargs > list_length(*args))
*args = add_function_defaults(*args, result_type, func_tuple, context);
newexpr = evaluate_function(funcid, result_type, result_typmod,
result_collid, input_collid, *args,
func_tuple, context);
/*
* Some functions calls can be simplified at plan time based on properties
* specific to the function. For example, "varchar(s::varchar(4), 8,
* true)" simplifies to "s::varchar(4)", and "int4mul(n, 1)" could
* simplify to "n". To define such function-specific optimizations, write
* a "transform function" and store its OID in the pg_proc.protransform of
* the primary function. Give each transform function the signature
* "protransform(internal) RETURNS internal". The argument, internally an
* Expr *, is the node representing a call to the primary function. If
* the transform function's study of that node proves that a simplified
* Expr substitutes for all possible concrete calls represented thereby,
* return that simplified Expr. Otherwise, return the NULL pointer.
*
* Currently, the specific Expr nodetag can be FuncExpr, OpExpr or
* DistinctExpr. This list may change in the future. The function should
* check the nodetag and return the NULL pointer for unexpected inputs.
*
* We make no guarantee that PostgreSQL will never call the primary
* function in cases that the transform function would simplify. Ensure
* rigorous equivalence between the simplified expression and an actual
* call to the primary function.
*
* Currently, this facility is undocumented and not exposed to users at
* the SQL level. Core length coercion casts use it to avoid calls
* guaranteed to return their input unchanged. This in turn allows ALTER
* TABLE ALTER TYPE to avoid rewriting tables for some typmod changes. In
* the future, this facility may find other applications, like simplifying
* x*0, x*1, and x+0.
*/
transform = ((Form_pg_proc) GETSTRUCT(func_tuple))->protransform;
if (!newexpr && OidIsValid(transform) && oldexpr)
newexpr = (Expr *) DatumGetPointer(OidFunctionCall1(transform,
PointerGetDatum(oldexpr)));
if (!newexpr && allow_non_const && OidIsValid(func_form->protransform))
{
/*
* Build a dummy FuncExpr node containing the simplified arg list. We
* use this approach to present a uniform interface to the transform
* function regardless of how the function is actually being invoked.
*/
FuncExpr fexpr;
if (!newexpr && allow_inline)
fexpr.xpr.type = T_FuncExpr;
fexpr.funcid = funcid;
fexpr.funcresulttype = result_type;
fexpr.funcretset = func_form->proretset;
fexpr.funcformat = COERCE_DONTCARE;
fexpr.funccollid = result_collid;
fexpr.inputcollid = input_collid;
fexpr.args = *args;
fexpr.location = -1;
newexpr = (Expr *)
DatumGetPointer(OidFunctionCall1(func_form->protransform,
PointerGetDatum(&fexpr)));
}
if (!newexpr && allow_non_const)
newexpr = inline_function(funcid, result_type, result_collid,
input_collid, *args,
func_tuple, context);

View File

@@ -2271,25 +2271,3 @@ transformFrameOffset(ParseState *pstate, int frameOptions, Node *clause)
return node;
}
/*
* relabel_to_typmod
* Add a RelabelType node that changes just the typmod, and remove all
* now-superfluous RelabelType nodes beneath it.
*/
Node *
relabel_to_typmod(Node *expr, int32 typmod)
{
Oid type = exprType(expr);
Oid coll = exprCollation(expr);
/*
* Strip any existing RelabelType, then add one. This is to preserve the
* invariant of no redundant RelabelTypes.
*/
while (IsA(expr, RelabelType))
expr = (Node *) ((RelabelType *) expr)->arg;
return (Node *) makeRelabelType((Expr *) expr, type, typmod, coll,
COERCE_DONTCARE);
}

View File

@@ -24,7 +24,6 @@
#include "funcapi.h"
#include "miscadmin.h"
#include "nodes/nodeFuncs.h"
#include "parser/parse_clause.h"
#include "utils/builtins.h"
#include "utils/date.h"
#include "utils/datetime.h"
@@ -4153,32 +4152,34 @@ CheckDateTokenTables(void)
}
/*
* Helper for temporal protransform functions. Types time, timetz, timestamp
* and timestamptz each have a range of allowed precisions. An unspecified
* precision is rigorously equivalent to the highest specifiable precision.
* Common code for temporal protransform functions. Types time, timetz,
* timestamp and timestamptz each have a range of allowed precisions. An
* unspecified precision is rigorously equivalent to the highest specifiable
* precision.
*
* Note: timestamp_scale throws an error when the typmod is out of range, but
* we can't get there from a cast: our typmodin will have caught it already.
*/
Node *
TemporalTransform(int32 max_precis, Node *node)
{
FuncExpr *expr = (FuncExpr *) node;
Node *typmod;
Node *ret = NULL;
Node *typmod;
if (!IsA(expr, FuncExpr))
return ret;
Assert(IsA(expr, FuncExpr));
Assert(list_length(expr->args) >= 2);
Assert(list_length(expr->args) == 2);
typmod = lsecond(expr->args);
typmod = (Node *) lsecond(expr->args);
if (IsA(typmod, Const))
if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
{
Node *source = linitial(expr->args);
Node *source = (Node *) linitial(expr->args);
int32 old_precis = exprTypmod(source);
int32 new_precis = DatumGetInt32(((Const *) typmod)->constvalue);
if (new_precis == -1 ||
new_precis == max_precis ||
(old_precis != -1 && new_precis >= old_precis))
if (new_precis < 0 || new_precis == max_precis ||
(old_precis >= 0 && new_precis >= old_precis))
ret = relabel_to_typmod(source, new_precis);
}

View File

@@ -31,7 +31,6 @@
#include "libpq/pqformat.h"
#include "miscadmin.h"
#include "nodes/nodeFuncs.h"
#include "parser/parse_clause.h"
#include "utils/array.h"
#include "utils/builtins.h"
#include "utils/int8.h"
@@ -717,7 +716,7 @@ numeric_send(PG_FUNCTION_ARGS)
/*
* numeric_transform() -
*
* Flatten calls to our length coercion function that solely represent
* Flatten calls to numeric's length coercion function that solely represent
* increases in allowable precision. Scale changes mutate every datum, so
* they are unoptimizable. Some values, e.g. 1E-1001, can only fit into an
* unconstrained numeric, so a change from an unconstrained numeric to any
@@ -727,18 +726,17 @@ Datum
numeric_transform(PG_FUNCTION_ARGS)
{
FuncExpr *expr = (FuncExpr *) PG_GETARG_POINTER(0);
Node *typmod;
Node *ret = NULL;
Node *typmod;
if (!IsA(expr, FuncExpr))
PG_RETURN_POINTER(ret);
Assert(IsA(expr, FuncExpr));
Assert(list_length(expr->args) >= 2);
Assert(list_length(expr->args) == 2);
typmod = lsecond(expr->args);
typmod = (Node *) lsecond(expr->args);
if (IsA(typmod, Const))
if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
{
Node *source = linitial(expr->args);
Node *source = (Node *) linitial(expr->args);
int32 old_typmod = exprTypmod(source);
int32 new_typmod = DatumGetInt32(((Const *) typmod)->constvalue);
int32 old_scale = (old_typmod - VARHDRSZ) & 0xffff;
@@ -748,11 +746,12 @@ numeric_transform(PG_FUNCTION_ARGS)
/*
* If new_typmod < VARHDRSZ, the destination is unconstrained; that's
* always OK. If old_typmod >= VARHDRSZ, the source is constrained.
* and we're OK if the scale is unchanged and the precison is not
* always OK. If old_typmod >= VARHDRSZ, the source is constrained,
* and we're OK if the scale is unchanged and the precision is not
* decreasing. See further notes in function header comment.
*/
if (new_typmod < VARHDRSZ || (old_typmod >= VARHDRSZ &&
if (new_typmod < (int32) VARHDRSZ ||
(old_typmod >= (int32) VARHDRSZ &&
new_scale == old_scale && new_precision >= old_precision))
ret = relabel_to_typmod(source, new_typmod);
}

View File

@@ -28,7 +28,6 @@
#include "libpq/pqformat.h"
#include "miscadmin.h"
#include "nodes/nodeFuncs.h"
#include "parser/parse_clause.h"
#include "parser/scansup.h"
#include "utils/array.h"
#include "utils/builtins.h"
@@ -316,10 +315,6 @@ timestamptypmodout(PG_FUNCTION_ARGS)
Datum
timestamp_transform(PG_FUNCTION_ARGS)
{
/*
* timestamp_scale throws an error when the typmod is out of range, but we
* can't get there from a cast: our typmodin will have caught it already.
*/
PG_RETURN_POINTER(TemporalTransform(MAX_TIMESTAMP_PRECISION,
(Node *) PG_GETARG_POINTER(0)));
}
@@ -937,18 +932,17 @@ Datum
interval_transform(PG_FUNCTION_ARGS)
{
FuncExpr *expr = (FuncExpr *) PG_GETARG_POINTER(0);
Node *typmod;
Node *ret = NULL;
Node *typmod;
if (!IsA(expr, FuncExpr))
PG_RETURN_POINTER(ret);
Assert(IsA(expr, FuncExpr));
Assert(list_length(expr->args) >= 2);
Assert(list_length(expr->args) == 2);
typmod = lsecond(expr->args);
typmod = (Node *) lsecond(expr->args);
if (IsA(typmod, Const))
if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
{
Node *source = linitial(expr->args);
Node *source = (Node *) linitial(expr->args);
int32 old_typmod = exprTypmod(source);
int32 new_typmod = DatumGetInt32(((Const *) typmod)->constvalue);
int old_range;
@@ -958,7 +952,7 @@ interval_transform(PG_FUNCTION_ARGS)
int new_range_fls;
int old_range_fls;
if (old_typmod == -1)
if (old_typmod < 0)
{
old_range = INTERVAL_FULL_RANGE;
old_precis = INTERVAL_FULL_PRECISION;
@@ -978,11 +972,9 @@ interval_transform(PG_FUNCTION_ARGS)
*/
new_range_fls = fls(new_range);
old_range_fls = fls(old_range);
if (new_typmod == -1 ||
((new_range_fls >= SECOND ||
new_range_fls >= old_range_fls) &&
(old_range_fls < SECOND ||
new_precis >= MAX_INTERVAL_PRECISION ||
if (new_typmod < 0 ||
((new_range_fls >= SECOND || new_range_fls >= old_range_fls) &&
(old_range_fls < SECOND || new_precis >= MAX_INTERVAL_PRECISION ||
new_precis >= old_precis)))
ret = relabel_to_typmod(source, new_typmod);
}
@@ -1068,7 +1060,7 @@ AdjustIntervalForTypmod(Interval *interval, int32 typmod)
* can't do it consistently. (We cannot enforce a range limit on the
* highest expected field, since we do not have any equivalent of
* SQL's <interval leading field precision>.) If we ever decide to
* revisit this, interval_transform will likely requite adjusting.
* revisit this, interval_transform will likely require adjusting.
*
* Note: before PG 8.4 we interpreted a limited set of fields as
* actually causing a "modulo" operation on a given value, potentially

View File

@@ -19,7 +19,6 @@
#include "access/htup.h"
#include "libpq/pqformat.h"
#include "nodes/nodeFuncs.h"
#include "parser/parse_clause.h"
#include "utils/array.h"
#include "utils/varbit.h"
@@ -649,31 +648,31 @@ varbit_send(PG_FUNCTION_ARGS)
/*
* varbit_transform()
* Flatten calls to our length coercion function that leave the new maximum
* length >= the previous maximum length. We ignore the isExplicit argument,
* which only affects truncation.
* Flatten calls to varbit's length coercion function that set the new maximum
* length >= the previous maximum length. We can ignore the isExplicit
* argument, since that only affects truncation cases.
*/
Datum
varbit_transform(PG_FUNCTION_ARGS)
{
FuncExpr *expr = (FuncExpr *) PG_GETARG_POINTER(0);
Node *typmod;
Node *ret = NULL;
Node *typmod;
if (!IsA(expr, FuncExpr))
PG_RETURN_POINTER(ret);
Assert(IsA(expr, FuncExpr));
Assert(list_length(expr->args) >= 2);
Assert(list_length(expr->args) == 3);
typmod = lsecond(expr->args);
typmod = (Node *) lsecond(expr->args);
if (IsA(typmod, Const))
if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
{
Node *source = linitial(expr->args);
Node *source = (Node *) linitial(expr->args);
int32 new_typmod = DatumGetInt32(((Const *) typmod)->constvalue);
int32 old_max = exprTypmod(source);
int32 new_max = new_typmod;
if (new_max <= 0 || (old_max >= 0 && old_max <= new_max))
/* Note: varbit() treats typmod 0 as invalid, so we do too */
if (new_max <= 0 || (old_max > 0 && old_max <= new_max))
ret = relabel_to_typmod(source, new_typmod);
}

View File

@@ -19,7 +19,6 @@
#include "access/tuptoaster.h"
#include "libpq/pqformat.h"
#include "nodes/nodeFuncs.h"
#include "parser/parse_clause.h"
#include "utils/array.h"
#include "utils/builtins.h"
#include "mb/pg_wchar.h"
@@ -551,31 +550,32 @@ varcharsend(PG_FUNCTION_ARGS)
/*
* Flatten calls to our length coercion function that leave the new maximum
* length >= the previous maximum length. We ignore the isExplicit argument,
* which only affects truncation.
* varchar_transform()
* Flatten calls to varchar's length coercion function that set the new maximum
* length >= the previous maximum length. We can ignore the isExplicit
* argument, since that only affects truncation cases.
*/
Datum
varchar_transform(PG_FUNCTION_ARGS)
{
FuncExpr *expr = (FuncExpr *) PG_GETARG_POINTER(0);
Node *typmod;
Node *ret = NULL;
Node *typmod;
if (!IsA(expr, FuncExpr))
PG_RETURN_POINTER(ret);
Assert(IsA(expr, FuncExpr));
Assert(list_length(expr->args) >= 2);
Assert(list_length(expr->args) == 3);
typmod = lsecond(expr->args);
typmod = (Node *) lsecond(expr->args);
if (IsA(typmod, Const))
if (IsA(typmod, Const) && !((Const *) typmod)->constisnull)
{
Node *source = linitial(expr->args);
Node *source = (Node *) linitial(expr->args);
int32 old_typmod = exprTypmod(source);
int32 new_typmod = DatumGetInt32(((Const *) typmod)->constvalue);
int32 old_max = exprTypmod(source) - VARHDRSZ;
int32 old_max = old_typmod - VARHDRSZ;
int32 new_max = new_typmod - VARHDRSZ;
if (new_max < 0 || (old_max >= 0 && old_max <= new_max))
if (new_typmod < 0 || (old_typmod >= 0 && old_max <= new_max))
ret = relabel_to_typmod(source, new_typmod);
}