Repeatedly rewriting a mapped catalog table with VACUUM FULL or
CLUSTER could cause logical decoding to fail with:
ERROR, "could not map filenode \"%s\" to relation OID"
To trigger the problem the rewritten catalog had to have live tuples
with toasted columns.
The problem was triggered as during catalog table rewrites the
heap_insert() check that prevents logical decoding information to be
emitted for system catalogs, failed to treat the new heap's toast table
as a system catalog (because the new heap is not recognized as a
catalog table via RelationIsLogicallyLogged()). The relmapper, in
contrast to the normal catalog contents, does not contain historical
information. After a single rewrite of a mapped table the new relation
is known to the relmapper, but if the table is rewritten twice before
logical decoding occurs, the relfilenode cannot be mapped to a
relation anymore. Which then leads us to error out. This only
happens for toast tables, because the main table contents aren't
re-inserted with heap_insert().
The fix is simple, add a new heap_insert() flag that prevents logical
decoding information from being emitted, and accept during decoding
that there might not be tuple data for toast tables.
Unfortunately that does not fix pre-existing logical decoding
errors. Doing so would require not throwing an error when a filenode
cannot be mapped to a relation during decoding, and that seems too
likely to hide bugs. If it's crucial to fix decoding for an existing
slot, temporarily changing the ERROR in ReorderBufferCommit() to a
WARNING appears to be the best fix.
Author: Andres Freund
Discussion: https://postgr.es/m/20180914021046.oi7dm4ra3ot2g2kt@alap3.anarazel.de
Backpatch: 9.4-, where logical decoding was introduced
Without that the tests fail when forced to be run against a cluster
with synchronous_commit = off (as the WAL might not yet be flushed to
disk by the point logical decoding gets called, and thus the expected
output breaks). Most tests already do that, add it to a few newer tests.
Author: Andres Freund
It was previously a string setting that was converted into an enum by
custom code, but using the GUC enum facility seems much simpler and
doesn't change any functionality, except that
set transaction_isolation='default';
no longer works, but that was never documented and doesn't work with
any other transaction characteristics. (Note that this is not the
same as RESET or SET TO DEFAULT, which still work.)
Reviewed-by: Heikki Linnakangas <hlinnaka@iki.fi>
Discussion: https://www.postgresql.org/message-id/457db615-e84c-4838-310e-43841eb806e5@iki.fi
This code used elog where it really ought to use ereport, mainly so that
it can report a SQLSTATE different from ERRCODE_INTERNAL_ERROR. There
were some other random deviations from typical error report practice too.
In addition, we can make some cleanups that were impractical six months
ago:
* Use one variadic macro, instead of several with different numbers
of arguments, reducing the temptation to force-fit messages into
particular numbers of arguments;
* Use %m, even in the frontend case, simplifying the code.
Discussion: https://postgr.es/m/6025.1527351693@sss.pgh.pa.us
numerology_1.out and float8-small-is-zero_1.out differ from their
base files only in showing plain zero rather than minus zero for
some results. I believe that in the wake of commit 6eb3eb577,
we will print minus zero as such on all IEEE-float platforms
(and non-IEEE floats are going to cause many more regression diffs
than this, anyway). Hence we should be able to remove these and
eliminate a bit of maintenance pain. Let's see if the buildfarm
agrees.
Discussion: https://postgr.es/m/29037.1539021687@sss.pgh.pa.us
NetBSD-current generates a large number of warnings about "%m" not
being appropriate to use with *printf functions. While that's true
for their native printf, it's surely not true for snprintf.c, so I
think they have misunderstood gcc's definition of the "gnu_printf"
archetype. Nonetheless, choosing "__syslog__" instead silences the
warnings; so teach configure about that.
Since this is only a cosmetic warning issue (and anyway it depends
on previous hacking to be self-consistent), no back-patch.
Discussion: https://postgr.es/m/16785.1539046036@sss.pgh.pa.us
This function lists the contents of the WAL archive status directory,
and is intended to be used by monitoring tools. Unlike pg_ls_dir(),
access to it can be granted to non-superusers so that those monitoring
tools can observe the principle of least privilege. Access is also
given by default to members of pg_monitor.
Author: Christoph Moench-Tegeder
Reviewed-by: Aya Iwata
Discussion: https://postgr.es/m/20180930205920.GA64534@elch.exwg.net
The idea here is that patches that add items to these lists will often
be easier to rebase over other additions to the same lists, because
they won't be trying to touch the very same line of configure.in.
There will still be merge conflicts in the configure script, but that
can be fixed just by re-running autoconf (or by leaving configure out
of the submitted patch to begin with ...)
Implementation note: use of m4_normalize() is necessary to get rid of
the newlines, else incorrect shell syntax will be emitted. But with
that hack, the generated configure script is identical to what it
was before.
Discussion: https://postgr.es/m/19344.1539050134@sss.pgh.pa.us
Originally committed as 15bc038f (plus some follow-ups), this was
reverted in 28e07270 due to a problem discovered in parallel
workers. This new version corrects that problem by sending the
list of uncommitted enum values to parallel workers.
Here follows the original commit message describing the change:
To prevent possibly breaking indexes on enum columns, we must keep
uncommitted enum values from getting stored in tables, unless we
can be sure that any such column is new in the current transaction.
Formerly, we enforced this by disallowing ALTER TYPE ... ADD VALUE
from being executed at all in a transaction block, unless the target
enum type had been created in the current transaction. This patch
removes that restriction, and instead insists that an uncommitted enum
value can't be referenced unless it belongs to an enum type created
in the same transaction as the value. Per discussion, this should be
a bit less onerous. It does require each function that could possibly
return a new enum value to SQL operations to check this restriction,
but there aren't so many of those that this seems unmaintainable.
Author: Andrew Dunstan and Tom Lane, with parallel query fix by Thomas Munro
Reviewed-by: Tom Lane
Discussion: https://postgr.es/m/CAEepm%3D0Ei7g6PaNTbcmAh9tCRahQrk%3Dr5ZWLD-jr7hXweYX3yg%40mail.gmail.com
Discussion: https://postgr.es/m/4075.1459088427%40sss.pgh.pa.us
A warning on a NetBSD box revealed to me that pg_waldump/compat.c
is using vprintf(), which snprintf.c did not provide coverage for.
This is not good if we want to have uniform *printf behavior, and
it's pretty silly to omit when it's a one-line function.
I also noted that snprintf.c has pg_vsprintf() but for some reason
it was not exposed to the outside world, creating another way in
which code might accidentally invoke the platform *printf family.
Let's just make sure that we replace all eight of the POSIX-standard
printf family.
Also, upgrade plperl.h and plpython.h to make sure that they do
their undefine/redefine rain dance for all eight, not some random
maybe-sufficient subset thereof.
Up to now, float4out/float8out handled NaN and Infinity cases explicitly,
and invoked psprintf only for ordinary float values. This was done because
platform implementations of snprintf produce varying representations of
these special cases. But now that we use snprintf.c always, it's better
to give it the responsibility to produce a uniform representation of
these cases, so that we have uniformity across the board not only in
float4out/float8out. Hence, move that work into fmtfloat().
Also, teach fmtfloat() to recognize IEEE minus zero and handle it
correctly. The previous coding worked only accidentally, and would
fail for e.g. "%+f" format (it'd print "+-0.00000"). Now that we're
using snprintf.c everywhere, it's not acceptable for it to do weird
things in corner cases. (This incidentally avoids a portability
problem we've seen on some really ancient platforms, that native
sprintf does the wrong thing with minus zero.)
Also, introduce a new entry point in snprintf.c to allow float[48]out
to bypass the work of interpreting a well-known format spec, as well
as bypassing the overhead of the psprintf layer. I modeled this API
loosely on strfromd(). In my testing, this brings float[48]out back
to approximately the same speed they had when using native snprintf,
fixing one of the main performance issues caused by using snprintf.c.
(There is some talk of more aggressive work to improve the speed of
floating-point output conversion, but these changes seem to provide
a better starting point for such work anyway.)
Getting rid of the previous ad-hoc hack for Infinity/NaN in fmtfloat()
allows removing <ctype.h> from snprintf.c's #includes. I also removed
a few other #includes that I think are historical, though the buildfarm
may expose that as wrong.
Discussion: https://postgr.es/m/13178.1538794717@sss.pgh.pa.us
If there are many ExecRowMark structs, we spent O(N^2) time in
ExecFindRowMark during executor startup. Once upon a time this was
not of great concern, but the addition of native partitioning has
squeezed out enough other costs that this can become the dominant
overhead in some use-cases for tables with many partitions.
To fix, simply replace that List data structure with an array.
This adds a little bit of cost to execCurrentOf(), but not much,
and anyway that code path is neither of large importance nor very
efficient now. If we ever decide it is a bottleneck, constructing a
hash table for lookup-by-tableoid would likely be the thing to do.
Per complaint from Amit Langote, though this is different from
his fix proposal.
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
gcc 6.3 does not whine about this mistake I made in 39808e8868 but
evidently lots of other compilers do, according to Michael Paquier,
Peter Eisentraut, Arthur Zakirov, Tomas Vondra.
Discussion: too many to list
Error messages for creating a foreign key on a partitioned table using
ONLY or NOT VALID were wrong in mentioning the objects they worked on.
This commit adds on the way some regression tests missing for those
cases.
Author: Laurenz Albe
Reviewed-by: Michael Paquier
Discussion: https://postgr.es/m/c11c05810a9ed65e9b2c817a9ef442275a32fe80.camel@cybertec.at
In the wake of commit f2343653f, we no longer need some fields that
were used before to control executor lock acquisitions:
* PlannedStmt.nonleafResultRelations can go away entirely.
* partitioned_rels can go away from Append, MergeAppend, and ModifyTable.
However, ModifyTable still needs to know the RT index of the partition
root table if any, which was formerly kept in the first entry of that
list. Add a new field "rootRelation" to remember that. rootRelation is
partly redundant with nominalRelation, in that if it's set it will have
the same value as nominalRelation. However, the latter field has a
different purpose so it seems best to keep them distinct.
Amit Langote, reviewed by David Rowley and Jesper Pedersen,
and whacked around a bit more by me
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
Commit 2fbdf1b38b changed the order in which we inserted catalog rows
when creating partitions, so that we could remove an unsightly hack
required for untimely relcache invalidations. However, that commit only
changed the ordering for CREATE TABLE PARTITION OF, and left ALTER TABLE
ATTACH PARTITION unchanged, so the latter can be affected when catalog
invalidations occur, for instance when the partition key involves an SQL
function.
Reported-by: Rajkumar Raghuwanshi
Author: Amit Langote
Reviewed-by: Michaël Paquier
Discussion: https://postgr.es/m/CAKcux6=nTz9KSfTr_6Z2mpzLJ_09JN-rK6=dWic6gGyTSWueyQ@mail.gmail.com
Index DDL cascading on partitioned tables introduced a way for ALTER
TABLE to be called reentrantly. This caused an an important deficiency
in event trigger support to be exposed: on exiting the reentrant call,
the alter table state object was clobbered, causing a crash when the
outer alter table tries to finalize its processing. Fix the crash by
creating a stack of event trigger state objects. There are still ways
to cause things to misbehave (and probably other crashers) with more
elaborate tricks, but at least it now doesn't crash in the obvious
scenario.
Backpatch to 9.5, where DDL deparsing of event triggers was introduced.
Reported-by: Marco Slot
Authors: Michaël Paquier, Álvaro Herrera
Discussion: https://postgr.es/m/CANNhMLCpi+HQ7M36uPfGbJZEQLyTy7XvX=5EFkpR-b1bo0uJew@mail.gmail.com
Commit 9a3cebeaa changed things so that parallel workers didn't obtain
any lock of their own on tables they access. That was clearly a bad
idea, but I'd mistakenly supposed that it was the intended end result
of the series of patches for simplifying the executor's lock management.
Undo that change in relation_open(), and adjust ExecOpenScanRelation()
so that it gets the correct lock if inside a parallel worker.
In passing, clean up some more obsolete comments about when locks
are acquired.
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
We already have appropriate locks on every relation listed in the
query's rangetable before we reach the executor. Take the next step
in exploiting that knowledge by removing code that worries about
taking locks on non-leaf result relations in a partitioned table.
In particular, get rid of ExecLockNonLeafAppendTables and a stanza in
InitPlan that asserts we already have locks on certain such tables.
In passing, clean up some now-obsolete comments in InitPlan.
Amit Langote, reviewed by David Rowley and Jesper Pedersen,
and whacked around a bit more by me
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
Places that aren't testing for sign should not use the more expensive
function; it's just wasteful, not to mention being a cognitive load
for readers who may know what isinf() is but not is_infinite().
As things stand, we actually don't need is_infinite() anyplace except
float4out/float8out, which means it could potentially go away altogether
after the changes I proposed in <13178.1538794717@sss.pgh.pa.us>.
Previously, a worker process would establish values for these based on
its own start time. In v10 and up, this can trivially be shown to cause
misbehavior of transaction_timestamp(), timestamp_in(), and related
functions which are (perhaps unwisely?) marked parallel-safe. It seems
likely that other behaviors might diverge from what happens in the parent
as well.
It's not as trivial to demonstrate problems in 9.6 or 9.5, but I'm sure
it's still possible, so back-patch to all branches containing parallel
worker infrastructure.
In HEAD only, mark now() and statement_timestamp() as parallel-safe
(other affected functions already were). While in theory we could
still squeeze that change into v11, it doesn't seem important enough
to force a last-minute catversion bump.
Konstantin Knizhnik, whacked around a bit by me
Discussion: https://postgr.es/m/6406dbd2-5d37-4cb6-6eb2-9c44172c7e7c@postgrespro.ru
When computing statistical aggregates like variance, the common
schoolbook algorithm which computes the sum of the squares of the
values and subtracts the square of the mean can lead to a large loss
of precision when using floating point arithmetic, because the
difference between the two terms is often very small relative to the
terms themselves.
To avoid this, re-work these aggregates to use the Youngs-Cramer
algorithm, which is a proven, numerically stable algorithm that
directly aggregates the sum of the squares of the differences of the
values from the mean in a single pass over the data.
While at it, improve the test coverage to test the aggregate combine
functions used during parallel aggregation.
Per report and suggested algorithm from Erich Schubert.
Patch by me, reviewed by Madeleine Thompson.
Discussion: https://postgr.es/m/153313051300.1397.9594490737341194671@wrigleys.postgresql.org
This is for example used when attaching a partition to a partitioned
table which includes foreign keys, and in this case the constraint name
has been missing in the data cloned. This could lead to hard crashes,
as when validating the foreign key constraint, the constraint name is
always expected. Particularly, when using log_min_messages >= DEBUG1, a
log message would be generated with this unassigned constraint name,
leading to an assertion failure on HEAD.
While on it, rename a variable in ATExecAttachPartition which was
declared twice with the same name.
Author: Michael Paquier
Reviewed-by: Álvaro Herrera
Discussion: https://postgr.es/m/20181005042236.GG1629@paquier.xyz
Backpatch-through: 11
Historically we forbade datatype-specific comparison functions from
returning INT_MIN, so that it would be safe to invert the sort order
just by negating the comparison result. However, this was never
really safe for comparison functions that directly return the result
of memcmp(), strcmp(), etc, as POSIX doesn't place any such restriction
on those library functions. Buildfarm results show that at least on
recent Linux on s390x, memcmp() actually does return INT_MIN sometimes,
causing sort failures.
The agreed-on answer is to remove this restriction and fix relevant
call sites to not make such an assumption; code such as "res = -res"
should be replaced by "INVERT_COMPARE_RESULT(res)". The same is needed
in a few places that just directly negated the result of memcmp or
strcmp.
To help find places having this problem, I've also added a compile option
to nbtcompare.c that causes some of the commonly used comparators to
return INT_MIN/INT_MAX instead of their usual -1/+1. It'd likely be
a good idea to have at least one buildfarm member running with
"-DSTRESS_SORT_INT_MIN". That's far from a complete test of course,
but it should help to prevent fresh introductions of such bugs.
This is a longstanding portability hazard, so back-patch to all supported
branches.
Discussion: https://postgr.es/m/20180928185215.ffoq2xrq5d3pafna@alap3.anarazel.de
This lists the contents of a temporary directory associated to a given
tablespace, useful to get information about on-disk consumption caused
by temporary files used by a session query. By default, pg_default is
scanned, and a tablespace can be specified as argument.
This function is intended to be used by monitoring tools, and, unlike
pg_ls_dir(), access to them can be granted to non-superusers so that
those monitoring tools can observe the principle of least privilege.
Access is also given by default to members of pg_monitor.
Author: Nathan Bossart
Reviewed-by: Laurenz Albe
Discussion: https://postgr.es/m/92F458A2-6459-44B8-A7F2-2ADD3225046A@amazon.com
Instead of doing a lot of list_nth() accesses to es_range_table,
create a flattened pointer array during executor startup and index
into that to get at individual RangeTblEntrys.
This eliminates one source of O(N^2) behavior with lots of partitions.
(I'm not exactly convinced that it's the most important source, but
it's an easy one to fix.)
Amit Langote and David Rowley
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
Create an array estate->es_relations[] paralleling the es_range_table,
and store references to Relations (relcache entries) there, so that any
given RT entry is opened and closed just once per executor run. Scan
nodes typically still call ExecOpenScanRelation, but ExecCloseScanRelation
is no more; relation closing is now done centrally in ExecEndPlan.
This is slightly more complex than one would expect because of the
interactions with relcache references held in ResultRelInfo nodes.
The general convention is now that ResultRelInfo->ri_RelationDesc does
not represent a separate relcache reference and so does not need to be
explicitly closed; but there is an exception for ResultRelInfos in the
es_trig_target_relations list, which are manufactured by
ExecGetTriggerResultRel and have to be cleaned up by
ExecCleanUpTriggerState. (That much was true all along, but these
ResultRelInfos are now more different from others than they used to be.)
To allow the partition pruning logic to make use of es_relations[] rather
than having its own relcache references, adjust PartitionedRelPruneInfo
to store an RT index rather than a relation OID.
Amit Langote, reviewed by David Rowley and Jesper Pedersen,
some mods by me
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
This moves the system administration functions for signalling backends
from backend/utils/adt/misc.c into a separate file dedicated to backend
signalling. No new functionality is introduced in this commit.
Author: Daniel Gustafsson
Reviewed-by: Michael Paquier, Álvaro Herrera
Discussion: https://postgr.es/m/C2C7C3EC-CC5F-44B6-9C78-637C88BD7D14@yesql.se
When specified, this option allows VACUUM to skip the work on a relation
if there is a conflicting lock on it when trying to open it at the
beginning of its processing.
Similarly to autovacuum, this comes with a couple of limitations while
the relation is processed which can cause the process to still block:
- when opening the relation indexes.
- when acquiring row samples for table inheritance trees, partition trees
or certain types of foreign tables, and that a lock is taken on some
leaves of such trees.
Author: Nathan Bossart
Reviewed-by: Michael Paquier, Andres Freund, Masahiko Sawada
Discussion: https://postgr.es/m/9EF7EBE4-720D-4CF1-9D0E-4403D7E92990@amazon.com
Discussion: https://postgr.es/m/20171201160907.27110.74730@wrigleys.postgresql.org
This isn't actually a live bug, as the output happens to be the
same. But it upsets tools like UBSan, which makes it worthwhile to
fix.
As it's an issue without practical consequences, don't backpatch.
Author: Andres Freund
Discussion: https://postgr.es/m/20180928001121.hhx5n6dsygqxr5wu@alap3.anarazel.de
Instead of locking tables during executor startup, just Assert that
suitable locks were obtained already during the parse/plan pipeline
(or re-obtained by the plan cache). This must be so, else we have a
hazard that concurrent DDL has invalidated the plan.
This is pretty inefficient as well as undercommented, but it's all going
to go away shortly, so I didn't try hard. This commit is just another
attempt to use the buildfarm to see if we've missed anything in the plan
to simplify the executor's table management.
Note that the change needed here in relation_open() exposes that
parallel workers now really are accessing tables without holding any
lock of their own, whereas they were not doing that before this commit.
This does not give me a warm fuzzy feeling about that aspect of parallel
query; it does not seem like a good design, and we now know that it's
had exactly no actual testing. I think that we should modify parallel
query so that that change can be reverted.
Discussion: https://postgr.es/m/468c85d9-540e-66a2-1dde-fec2b741e688@lab.ntt.co.jp
I (Andres) was more than a bit hasty in committing 33001fd7a7
after last minute changes, leading to a number of problems (jit output
was only shown for JIT in parallel workers, and just EXPLAIN without
ANALYZE didn't work). Lukas luckily found these issues quickly.
Instead of combining instrumentation in in standard_ExecutorEnd(), do
so on demand in the new ExplainPrintJITSummary().
Also update a documentation example of the JIT output, changed in
52050ad8eb.
Author: Lukas Fittl, with minor changes by me
Discussion: https://postgr.es/m/CAP53PkxmgJht69pabxBXJBM+0oc6kf3KHMborLP7H2ouJ0CCtQ@mail.gmail.com
Backpatch: 11, where JIT compilation was introduced
Although all known platforms define "long long" as 64 bits, it still feels
a bit shaky to be using "va_arg(args, int64)" to pull out an argument that
the caller thought was declared "long long". The reason it was coded like
this, way back in commit 3311c7669, was to work around the possibility that
the compiler had no type named "long long" --- and, at the time, that it
maybe didn't have 64-bit ints at all. Now that we're requiring compilers
to support C99, those concerns are moot. Let's make the code clearer and
more bulletproof by writing "long long" where we mean "long long".
This does introduce a hazard that we'd inefficiently use 128-bit arithmetic
to convert plain old integers. The way to tackle that would be to provide
two versions of fmtint(), one for "long long" and one for narrower types.
Since, as of today, no platforms require that, we won't bother with the
extra code for now.
Discussion: https://postgr.es/m/1680.1538587115@sss.pgh.pa.us
This case occurs often enough (around 45% of conversion specs executed
in our regression tests are just "%s") that it's worth an extra test
per conversion spec to allow skipping all the logic associated with
field widths and padding when it happens.
Discussion: https://postgr.es/m/26193.1538582367@sss.pgh.pa.us
In combination, these changes make our version of snprintf as fast
or faster than most platforms' native snprintf, except for cases
involving floating-point conversion (which we still delegate to
the native sprintf). The speed penalty for a float conversion
is down to around 10% though, much better than before.
Notable changes:
* Rather than always parsing the format twice to see if it contains
instances of %n$, do the extra scan only if we actually find a $.
This obviously wins for non-localized formats, and even when there
is use of %n$, we can avoid scanning text before the first % twice.
* Use strchrnul() if available to find the next %, and emit the
literal text between % escapes as strings rather than char-by-char.
* Create a bespoke function (dopr_outchmulti) for the common case
of emitting N copies of the same character, in place of writing
loops around dopr_outch.
* Simplify construction of the format string for invocations of sprintf
for floats.
* Const-ify some internal functions, and avoid unnecessary use of
pass-by-reference arguments.
Patch by me, reviewed by Andres Freund
Discussion: https://postgr.es/m/11787.1534530779@sss.pgh.pa.us
The API (EOH_flatten_into) that flattens the expanded value representation
expects the target address to be maxaligned. All it's usage adhere to that
principle except when serializing datums for parallel query. Fix that
usage.
Diagnosed-by: Tom Lane
Author: Tom Lane and Amit Kapila
Backpatch-through: 9.6
Discussion: https://postgr.es/m/11629.1536550032@sss.pgh.pa.us
Cygwin has been building and linking against static libraries. Although
a bug this has been relatively harmless until now, when this has caused
errors due to changes in the way we build certain libraries. So this
patch makes things work the way we always intended, namely that we would
link against the dynamic libraries (cygpq.dll etc.) and just not build
the static libraries. The downstream packagers have been doing this for
some time, so this just aligns with their practice.
Extracted from a patch by Marco Atzeri, with a suggestion from Tom Lane.
Discussion: https://postgr.es/m/1056.1538235347@sss.pgh.pa.us
Andres Freund
Peter Eisentraut
Tom Lane
Greg Stark
Michael Paquier
Thomas Munro
Alvaro Herrera
Magnus Hagander
Dean Rasheed
Bruce Momjian
Amit Kapila
Andrew Dunstan