Re: [HACKERS] tuple-routing and constraint violation error message, revisited

[Robert Haas]

On Mon, Apr 10, 2017 at 8:14 AM, Ashutosh Bapat
 wrote:
> On Fri, Mar 31, 2017 at 7:43 AM, Amit Langote
>  wrote:
>> Summary is: We decided in f1b4c771ea7 [1] that passing the original slot
>> (one containing the tuple formatted per root partitioned table's tupdesc)
>> to ExecConstraints(), but that breaks certain cases.  Imagine what would
>> happen if a BR insert trigger changed the tuple - the original slot would
>> not contain those changes. So, it seems better to convert (if necessary)
>> the tuple formatted per partition tupdesc after tuple-routing back to the
>> root table's format and use the converted tuple to make val_desc shown in
>> the message if an error occurs.
>>
>> Attached rebased version of the patch that I had originally proposed
>> (summary above is the commit message).  Robert thought it would be fine to
>> show the row formatted per partition rowtype, but would look better if we
>> could show the column names as well (remember that we're trying to account
>> for possible differences in the ordering of columns between the root table
>> and leaf partitions to which tuples are routed.)
>>
>> Added this to PostgreSQL 10 open items list.
>
> The changes look good to me. Now, ExecConstraint() has three blocks
> which are almost similar, differing only in the constraints checked
> and the error message. It was manageable without partitioning and may
> be it's still manageable, but it's certainly being pushed to the
> limits. May be we should refactor error reporting code into a separate
> function and call it in those three places.

Yeah, possibly.

Thanks for the review.  I have committed the patch.

-- 
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Re: [HACKERS] tuple-routing and constraint violation error message, revisited

[Ashutosh Bapat]

On Fri, Mar 31, 2017 at 7:43 AM, Amit Langote
 wrote:
> Hi,
>
> Last message regarding this was by Robert on the original partitioning thread:
>
> https://www.postgresql.org/message-id/CA%2BTgmoZjGzSM5WwnyapFaw3GxnDLWh7pm8Xiz8_QWQnUQy%3DSCA%40mail.gmail.com
>
> Summary is: We decided in f1b4c771ea7 [1] that passing the original slot
> (one containing the tuple formatted per root partitioned table's tupdesc)
> to ExecConstraints(), but that breaks certain cases.  Imagine what would
> happen if a BR insert trigger changed the tuple - the original slot would
> not contain those changes. So, it seems better to convert (if necessary)
> the tuple formatted per partition tupdesc after tuple-routing back to the
> root table's format and use the converted tuple to make val_desc shown in
> the message if an error occurs.
>
> Attached rebased version of the patch that I had originally proposed
> (summary above is the commit message).  Robert thought it would be fine to
> show the row formatted per partition rowtype, but would look better if we
> could show the column names as well (remember that we're trying to account
> for possible differences in the ordering of columns between the root table
> and leaf partitions to which tuples are routed.)
>
> Added this to PostgreSQL 10 open items list.

The changes look good to me. Now, ExecConstraint() has three blocks
which are almost similar, differing only in the constraints checked
and the error message. It was manageable without partitioning and may
be it's still manageable, but it's certainly being pushed to the
limits. May be we should refactor error reporting code into a separate
function and call it in those three places.

-- 
Best Wishes,
Ashutosh Bapat
EnterpriseDB Corporation
The Postgres Database Company


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

Re: [HACKERS] tuple-routing and constraint violation error message, revisited

[Noah Misch]

On Fri, Mar 31, 2017 at 11:13:03AM +0900, Amit Langote wrote:
> Last message regarding this was by Robert on the original partitioning thread:
> 
> https://www.postgresql.org/message-id/CA%2BTgmoZjGzSM5WwnyapFaw3GxnDLWh7pm8Xiz8_QWQnUQy%3DSCA%40mail.gmail.com
> 
> Summary is: We decided in f1b4c771ea7 [1] that passing the original slot
> (one containing the tuple formatted per root partitioned table's tupdesc)
> to ExecConstraints(), but that breaks certain cases.  Imagine what would
> happen if a BR insert trigger changed the tuple - the original slot would
> not contain those changes. So, it seems better to convert (if necessary)
> the tuple formatted per partition tupdesc after tuple-routing back to the
> root table's format and use the converted tuple to make val_desc shown in
> the message if an error occurs.
> 
> Attached rebased version of the patch that I had originally proposed
> (summary above is the commit message).  Robert thought it would be fine to
> show the row formatted per partition rowtype, but would look better if we
> could show the column names as well (remember that we're trying to account
> for possible differences in the ordering of columns between the root table
> and leaf partitions to which tuples are routed.)
> 
> Added this to PostgreSQL 10 open items list.

[Action required within three days.  This is a generic notification.]

The above-described topic is currently a PostgreSQL 10 open item.  Robert,
since you committed the patch believed to have created it, you own this open
item.  If some other commit is more relevant or if this does not belong as a
v10 open item, please let us know.  Otherwise, please observe the policy on
open item ownership[1] and send a status update within three calendar days of
this message.  Include a date for your subsequent status update.  Testers may
discover new open items at any time, and I want to plan to get them all fixed
well in advance of shipping v10.  Consequently, I will appreciate your efforts
toward speedy resolution.  Thanks.

[1] 
https://www.postgresql.org/message-id/20170404140717.GA2675809%40tornado.leadboat.com


-- 
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers

[HACKERS] tuple-routing and constraint violation error message, revisited

[Amit Langote]

Hi,

Last message regarding this was by Robert on the original partitioning thread:

https://www.postgresql.org/message-id/CA%2BTgmoZjGzSM5WwnyapFaw3GxnDLWh7pm8Xiz8_QWQnUQy%3DSCA%40mail.gmail.com

Summary is: We decided in f1b4c771ea7 [1] that passing the original slot
(one containing the tuple formatted per root partitioned table's tupdesc)
to ExecConstraints(), but that breaks certain cases.  Imagine what would
happen if a BR insert trigger changed the tuple - the original slot would
not contain those changes. So, it seems better to convert (if necessary)
the tuple formatted per partition tupdesc after tuple-routing back to the
root table's format and use the converted tuple to make val_desc shown in
the message if an error occurs.

Attached rebased version of the patch that I had originally proposed
(summary above is the commit message).  Robert thought it would be fine to
show the row formatted per partition rowtype, but would look better if we
could show the column names as well (remember that we're trying to account
for possible differences in the ordering of columns between the root table
and leaf partitions to which tuples are routed.)

Added this to PostgreSQL 10 open items list.

Thanks,
Amit

[1] https://git.postgresql.org/gitweb/?p=postgresql.git=commit=f1b4c77
>From fb3e65de8018b867f755fe145fe4759be5a0fb54 Mon Sep 17 00:00:00 2001
From: amit 
Date: Fri, 31 Mar 2017 11:00:43 +0900
Subject: [PATCH] Fix reporting of violation in ExecConstraints, again

We decided in f1b4c771ea74f42447dccaed42ffcdcccf3aa694 that passing
the original slot (one containing the tuple formatted per root
partitioned table's tupdesc) to ExecConstraints(), but that breaks
certain cases.  Imagine what would happen if a BR trigger changed the
tuple - the original slot would not contain those changes.
So, it seems better to convert (if necessary) the tuple formatted
per partition tupdesc after tuple-routing back to the root table's
format and use the converted tuple to make val_desc shown in the
message if an error occurs.
---
 src/backend/commands/copy.c|  6 ++--
 src/backend/executor/execMain.c| 53 +-
 src/backend/executor/execReplication.c |  4 +--
 src/backend/executor/nodeModifyTable.c |  7 ++---
 src/include/executor/executor.h|  3 +-
 src/test/regress/expected/insert.out   | 21 --
 src/test/regress/sql/insert.sql| 21 --
 7 files changed, 90 insertions(+), 25 deletions(-)

diff --git a/src/backend/commands/copy.c b/src/backend/commands/copy.c
index 0158eda591..b537f84278 100644
--- a/src/backend/commands/copy.c
+++ b/src/backend/commands/copy.c
@@ -2505,8 +2505,7 @@ CopyFrom(CopyState cstate)
 
 	for (;;)
 	{
-		TupleTableSlot *slot,
-   *oldslot;
+		TupleTableSlot *slot;
 		bool		skip_tuple;
 		Oid			loaded_oid = InvalidOid;
 
@@ -2548,7 +2547,6 @@ CopyFrom(CopyState cstate)
 		ExecStoreTuple(tuple, slot, InvalidBuffer, false);
 
 		/* Determine the partition to heap_insert the tuple into */
-		oldslot = slot;
 		if (cstate->partition_dispatch_info)
 		{
 			int			leaf_part_index;
@@ -2650,7 +2648,7 @@ CopyFrom(CopyState cstate)
 /* Check the constraints of the tuple */
 if (cstate->rel->rd_att->constr ||
 	resultRelInfo->ri_PartitionCheck)
-	ExecConstraints(resultRelInfo, slot, oldslot, estate);
+	ExecConstraints(resultRelInfo, slot, estate);
 
 if (useHeapMultiInsert)
 {
diff --git a/src/backend/executor/execMain.c b/src/backend/executor/execMain.c
index f2995f2e7b..0f92bd49db 100644
--- a/src/backend/executor/execMain.c
+++ b/src/backend/executor/execMain.c
@@ -1825,8 +1825,7 @@ ExecPartitionCheck(ResultRelInfo *resultRelInfo, TupleTableSlot *slot,
  */
 void
 ExecConstraints(ResultRelInfo *resultRelInfo,
-TupleTableSlot *slot, TupleTableSlot *orig_slot,
-EState *estate)
+TupleTableSlot *slot, EState *estate)
 {
 	Relation	rel = resultRelInfo->ri_RelationDesc;
 	TupleDesc	tupdesc = RelationGetDescr(rel);
@@ -1849,23 +1848,37 @@ ExecConstraints(ResultRelInfo *resultRelInfo,
 			{
 char	   *val_desc;
 Relation	orig_rel = rel;
-TupleDesc	orig_tupdesc = tupdesc;
+TupleDesc	orig_tupdesc = RelationGetDescr(rel);
 
 /*
- * choose the correct relation to build val_desc from the
- * tuple contained in orig_slot
+ * If the tuple has been routed, it's been converted to the
+ * partition's rowtype, which might differ from the root
+ * table's.  We must convert it back to the root table's
+ * rowtype so that val_desc shown error message matches the
+ * input tuple.
  */
 if (resultRelInfo->ri_PartitionRoot)
 {
+	HeapTuple	tuple = ExecFetchSlotTuple(slot);
+	TupleConversionMap	*map;
+
 	rel = resultRelInfo->ri_PartitionRoot;
 	tupdesc = RelationGetDescr(rel);
+	/* a reverse map */
+	map = convert_tuples_by_name(orig_tupdesc, tupdesc,
+gettext_noop("could not convert row type"));