PL/pgSQL function within an exception handler. Make sure we use the right
resource owner when we create the tuplestore to hold returned tuples.
Simplify tuplestore API so that the caller doesn't need to be in the right
memory context when calling tuplestore_put* functions. tuplestore.c
automatically switches to the memory context used when the tuplestore was
created. Tuplesort was already modified like this earlier. This patch also
removes the now useless MemoryContextSwitch calls from callers.
Report by Aleksei on pgsql-bugs on Dec 22 2009. Backpatch to 8.1, like
the previous patch that broke this.
not include postgres.h nor anything else it doesn't directly need. Add
#includes to calling files as needed to compensate. Per my proposal of
yesterday.
This should be noted as a source code change in the 8.4 release notes,
since it's likely to require changes in add-on modules.
value-per-call mode. This should be more efficient in normal usage,
but the real problem with the prior coding was that it returned with
a SPI call still active. That could cause problems if execution was
interleaved with anything else that might use SPI.
backwards scan could actually happen. In particular, pass a flag to
materialize-mode SRFs that tells them whether they need to require random
access. In passing, also suppress unneeded backward-scan overhead for a
Portal's holdStore tuplestore. Per my proposal about reducing I/O costs for
tuplestores.
via a tuplestore instead of value-per-call. Refactor a few things to reduce
ensuing code duplication with nodeFunctionscan.c. This represents the
reasonably noncontroversial part of my proposed patch to switch SQL functions
over to returning tuplestores. For the moment, SQL functions still do things
the old way. However, this change enables PL SRFs to be called in targetlists
(observe changes in plperl regression results).
strings. This patch introduces four support functions cstring_to_text,
cstring_to_text_with_len, text_to_cstring, and text_to_cstring_buffer, and
two macros CStringGetTextDatum and TextDatumGetCString. A number of
existing macros that provided variants on these themes were removed.
Most of the places that need to make such conversions now require just one
function or macro call, in place of the multiple notational layers that used
to be needed. There are no longer any direct calls of textout or textin,
and we got most of the places that were using handmade conversions via
memcpy (there may be a few still lurking, though).
This commit doesn't make any serious effort to eliminate transient memory
leaks caused by detoasting toasted text objects before they reach
text_to_cstring. We changed PG_GETARG_TEXT_P to PG_GETARG_TEXT_PP in a few
places where it was easy, but much more could be done.
Brendan Jurd and Tom Lane
The original coding leaked memory (at least 8K per crosstab_hash call)
because it allowed the hash table to be allocated as a child of
TopMemoryContext and then never freed it. Fix that by putting the
hash table under per_query_ctx, instead. Also get rid of use
of a static variable to point to the hash table. Aside from being
ugly, that would actively do the wrong thing in the case of re-entrant
calls to crosstab_hash, which are at least theoretically possible
since it was expecting the static variable to stay valid across
a SPI_execute call.
fixup various places in the tree that were clearing a StringInfo by hand.
Making this function a part of the API simplifies client code slightly,
and avoids needlessly peeking inside the StringInfo interface.
tuples with less header overhead than a regular HeapTuple, per my
recent proposal. Teach TupleTableSlot code how to deal with these.
As proof of concept, change tuplestore.c to store MinimalTuples instead
of HeapTuples. Future patches will expand the concept to other places
where it is useful.
are unnecessarily allocated on the heap rather than the stack. If the
StringInfo doesn't outlive the stack frame in which it is created,
there is no need to allocate it on the heap via makeStringInfo() --
stack allocation is faster. While it's not a big deal unless the
code is in a critical path, I don't see a reason not to save a few
cycles -- using stack allocation is not less readable.
I also cleaned up a bit of code along the way: moved variable
declarations into a more tightly-enclosing scope where possible,
fixed some pointless copying of strings in dblink, etc.
more compliant with the error message style guide. In particular,
errdetail should begin with a capital letter and end with a period,
whereas errmsg should not. I also fixed a few related issues in
passing, such as fixing the repeated misspelling of "lexeme" in
contrib/tsearch2 (per Tom's suggestion).
comment line where output as too long, and update typedefs for /lib
directory. Also fix case where identifiers were used as variable names
in the backend, but as typedefs in ecpg (favor the backend for
indenting).
Backpatch to 8.1.X.
and RelationNameGetTupleDesc() as deprecated; remove uses of the
latter in the contrib library. Along the way, clean up crosstab()
code and documentation a little.
spotted by Qingqing Zhou. The HASH_ENTER action now automatically
fails with elog(ERROR) on out-of-memory --- which incidentally lets
us eliminate duplicate error checks in quite a bunch of places. If
you really need the old return-NULL-on-out-of-memory behavior, you
can ask for HASH_ENTER_NULL. But there is now an Assert in that path
checking that you aren't hoping to get that behavior in a palloc-based
hash table.
Along the way, remove the old HASH_FIND_SAVE/HASH_REMOVE_SAVED actions,
which were not being used anywhere anymore, and were surely too ugly
and unsafe to want to see revived again.
mode see a fresh snapshot for each command in the function, rather than
using the latest interactive command's snapshot. Also, suppress fresh
snapshots as well as CommandCounterIncrement inside STABLE and IMMUTABLE
functions, instead using the snapshot taken for the most closely nested
regular query. (This behavior is only sane for read-only functions, so
the patch also enforces that such functions contain only SELECT commands.)
As per my proposal of 6-Sep-2004; I note that I floated essentially the
same proposal on 19-Jun-2002, but that discussion tailed off without any
action. Since 8.0 seems like the right place to be taking possibly
nontrivial backwards compatibility hits, let's get it done now.
results with tuples as ordinary varlena Datums. This commit does not
in itself do much for us, except eliminate the horrid memory leak
associated with evaluation of whole-row variables. However, it lays the
groundwork for allowing composite types as table columns, and perhaps
some other useful features as well. Per my proposal of a few days ago.
Make btree index creation and initial validation of foreign-key constraints
use maintenance_work_mem rather than work_mem as their memory limit.
Add some code to guc.c to allow these variables to be referenced by their
old names in SHOW and SET commands, for backwards compatibility.
Lars Boegild Thomsen (full email below) and also corrects the regression
expected output for a recent backend message adjustment. Please apply.
Joe Conway
This was the last piece of code that took it upon itself to reset the
random number sequence --- now we only have srandom() in postmaster start,
backend start, and explicit setseed() operations.
>>Sounds like all that's needed for your case. But to be complete, in
>>addition to changing tablefunc.c we'd have to:
>>1) come up with a new function call signature that makes sense and does
>>not cause backward compatibility problems for other people
>>2) make needed changes to tablefunc.sql.in
>>3) adjust the README.tablefunc appropriately
>>4) adjust the regression test for new functionality
>>5) be sure we don't break any of the old cases
>>
>>If you want to submit a complete patch, it would be gratefully accepted
>>-- for review at least ;-)
>
> Here's the patch, at least for steps 1-3
Nabil Sayegh
Joe Conway
(materialization into a tuple store) discussed on pgsql-hackers earlier.
I've updated the documentation and the regression tests.
Notes on the implementation:
- I needed to change the tuple store API slightly -- it assumes that it
won't be used to hold data across transaction boundaries, so the temp
files that it uses for on-disk storage are automatically reclaimed at
end-of-transaction. I added a flag to tuplestore_begin_heap() to control
this behavior. Is changing the tuple store API in this fashion OK?
- in order to store executor results in a tuple store, I added a new
CommandDest. This works well for the most part, with one exception: the
current DestFunction API doesn't provide enough information to allow the
Executor to store results into an arbitrary tuple store (where the
particular tuple store to use is chosen by the call site of
ExecutorRun). To workaround this, I've temporarily hacked up a solution
that works, but is not ideal: since the receiveTuple DestFunction is
passed the portal name, we can use that to lookup the Portal data
structure for the cursor and then use that to get at the tuple store the
Portal is using. This unnecessarily ties the Portal code with the
tupleReceiver code, but it works...
The proper fix for this is probably to change the DestFunction API --
Tom suggested passing the full QueryDesc to the receiveTuple function.
In that case, callers of ExecutorRun could "subclass" QueryDesc to add
any additional fields that their particular CommandDest needed to get
access to. This approach would work, but I'd like to think about it for
a little bit longer before deciding which route to go. In the mean time,
the code works fine, so I don't think a fix is urgent.
- (semi-related) I added a NO SCROLL keyword to DECLARE CURSOR, and
adjusted the behavior of SCROLL in accordance with the discussion on
-hackers.
- (unrelated) Cleaned up some SGML markup in sql.sgml, copy.sgml
Neil Conway
version of crosstab. This fixes a major deficiency in real-world use of
the original version. Easiest to undestand with an illustration:
Data:
-------------------------------------------------------------------
select * from cth;
id | rowid | rowdt | attribute | val
----+-------+---------------------+----------------+---------------
1 | test1 | 2003-03-01 00:00:00 | temperature | 42
2 | test1 | 2003-03-01 00:00:00 | test_result | PASS
3 | test1 | 2003-03-01 00:00:00 | volts | 2.6987
4 | test2 | 2003-03-02 00:00:00 | temperature | 53
5 | test2 | 2003-03-02 00:00:00 | test_result | FAIL
6 | test2 | 2003-03-02 00:00:00 | test_startdate | 01 March 2003
7 | test2 | 2003-03-02 00:00:00 | volts | 3.1234
(7 rows)
Original crosstab:
-------------------------------------------------------------------
SELECT * FROM crosstab(
'SELECT rowid, attribute, val FROM cth ORDER BY 1,2',4)
AS c(rowid text, temperature text, test_result text, test_startdate
text, volts text);
rowid | temperature | test_result | test_startdate | volts
-------+-------------+-------------+----------------+--------
test1 | 42 | PASS | 2.6987 |
test2 | 53 | FAIL | 01 March 2003 | 3.1234
(2 rows)
Hashed crosstab:
-------------------------------------------------------------------
SELECT * FROM crosstab(
'SELECT rowid, attribute, val FROM cth ORDER BY 1',
'SELECT DISTINCT attribute FROM cth ORDER BY 1')
AS c(rowid text, temperature int4, test_result text, test_startdate
timestamp, volts float8);
rowid | temperature | test_result | test_startdate | volts
-------+-------------+-------------+---------------------+--------
test1 | 42 | PASS | | 2.6987
test2 | 53 | FAIL | 2003-03-01 00:00:00 | 3.1234
(2 rows)
Notice that the original crosstab slides data over to the left in the
result tuple when it encounters missing data. In order to work around
this you have to be make your source sql do all sorts of contortions
(cartesian join of distinct rowid with distinct attribute; left join
that back to the real source data). The new version avoids this by
building a hash table using a second distinct attribute query.
The new version also allows for "extra" columns (see the README) and
allows the result columns to be coerced into differing datatypes if they
are suitable (as shown above).
In testing a "real-world" data set (69 distinct rowid's, 27 distinct
categories/attributes, multiple missing data points) I saw about a
5-fold improvement in execution time (from about 2200 ms old, to 440 ms
new).
I left the original version intact because: 1) BC, 2) it is probably
slightly faster if you know that you have no missing attributes.
README and regression test adjustments included. If there are no
objections, please apply.
Joe Conway
