Thread

xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2011-12-06T04:19:49Z

Hi,

A while ago when blogging about WAL [1], I noticed a function to deal with
xlog location arithmetic is wanted. I remembered Depez [2] mentioning it and
after some questions during trainings and conferences I decided to translate
my shell script function in C.

The attached patch implements the function pg_xlog_location_diff (bikeshed
colors are welcome). It calculates the difference between two given
transaction log locations. Now that we have pg_stat_replication view, it will
be easy to get the lag just passing columns as parameters. Also, the
monitoring tools could take advantage of it instead of relying on a fragile
routine to get the lag.

I noticed that pg_xlogfile_name* functions does not sanity check the xrecoff
boundaries but that is material for another patch.


[1] http://eulerto.blogspot.com/2011/11/understanding-wal-nomenclature.html
[2]
http://www.depesz.com/index.php/2011/01/24/waiting-for-9-1-pg_stat_replication/


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2011-12-06T10:14:02Z

On Tue, Dec 6, 2011 at 05:19, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> Hi,
>
> A while ago when blogging about WAL [1], I noticed a function to deal with
> xlog location arithmetic is wanted. I remembered Depez [2] mentioning it and
> after some questions during trainings and conferences I decided to translate
> my shell script function in C.
>
> The attached patch implements the function pg_xlog_location_diff (bikeshed
> colors are welcome). It calculates the difference between two given
> transaction log locations. Now that we have pg_stat_replication view, it will
> be easy to get the lag just passing columns as parameters. Also, the
> monitoring tools could take advantage of it instead of relying on a fragile
> routine to get the lag.

I've been considering similar things, as you can find in the archives,
but what I was thinking of was converting the number to just a plain
bigint, then letting the user apply whatever arithmetic wanted at the
SQL level. I never got around to acutally coding it, though. It could
easily be extracted from your patch of course - and I think that's a
more flexible approach. Is there some advantage to your method that
I'm missing?

Also, why do you use DirectFunctionCall to do the simple math, and not
just do the math right there in the function?

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2011-12-06T16:11:32Z

On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
> I've been considering similar things, as you can find in the archives,
> but what I was thinking of was converting the number to just a plain
> bigint, then letting the user apply whatever arithmetic wanted at the
> SQL level. I never got around to acutally coding it, though. It could
> easily be extracted from your patch of course - and I think that's a
> more flexible approach. Is there some advantage to your method that
> I'm missing?

I went so far as to put together an lsn data type.  I didn't actually
get all that far with it, which is why I haven't posted it sooner, but
here's what I came up with.  It's missing indexing support and stuff,
but that could be added if people like the approach.  It solves this
problem by implementing -(lsn,lsn) => numeric (not int8, that can
overflow since it is not unsigned), which allows an lsn => numeric
conversion by just subtracting '0/0'::lsn.

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

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2011-12-06T17:39:21Z

On 06-12-2011 07:14, Magnus Hagander wrote:
> On Tue, Dec 6, 2011 at 05:19, Euler Taveira de Oliveira
> <euler@timbira.com> wrote:
>> Hi,
>>
>> A while ago when blogging about WAL [1], I noticed a function to deal with
>> xlog location arithmetic is wanted. I remembered Depez [2] mentioning it and
>> after some questions during trainings and conferences I decided to translate
>> my shell script function in C.
>>
>> The attached patch implements the function pg_xlog_location_diff (bikeshed
>> colors are welcome). It calculates the difference between two given
>> transaction log locations. Now that we have pg_stat_replication view, it will
>> be easy to get the lag just passing columns as parameters. Also, the
>> monitoring tools could take advantage of it instead of relying on a fragile
>> routine to get the lag.
> 
> I've been considering similar things, as you can find in the archives,
> but what I was thinking of was converting the number to just a plain
> bigint, then letting the user apply whatever arithmetic wanted at the
> SQL level. I never got around to acutally coding it, though. It could
> easily be extracted from your patch of course - and I think that's a
> more flexible approach. Is there some advantage to your method that
> I'm missing?
> 
The only advantage is that you don't expose the arithmetic, e.g., user doesn't
need to know the xlog internals (like I described in a recent blog post). If
one day we consider changes in xlog arithmetic (for example, XLogFileSize), we
don't need to worry too much about external tools.

> Also, why do you use DirectFunctionCall to do the simple math, and not
> just do the math right there in the function?
> 
I use it because I don't want to duplicate the overflow code.


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2011-12-06T18:00:51Z

On 06-12-2011 13:11, Robert Haas wrote:
> On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
>> I've been considering similar things, as you can find in the archives,
>> but what I was thinking of was converting the number to just a plain
>> bigint, then letting the user apply whatever arithmetic wanted at the
>> SQL level. I never got around to acutally coding it, though. It could
>> easily be extracted from your patch of course - and I think that's a
>> more flexible approach. Is there some advantage to your method that
>> I'm missing?
> 
> I went so far as to put together an lsn data type.  I didn't actually
> get all that far with it, which is why I haven't posted it sooner, but
> here's what I came up with.  It's missing indexing support and stuff,
> but that could be added if people like the approach.  It solves this
> problem by implementing -(lsn,lsn) => numeric (not int8, that can
> overflow since it is not unsigned), which allows an lsn => numeric
> conversion by just subtracting '0/0'::lsn.
> 
Interesting approach. I don't want to go that far. If so, you want to change
all of those functions that deal with LSNs and add some implicit conversion
between text and lsn data types (for backward compatibility). As of int8, I'm
not aware of any modern plataform that int8 is not 64 bits. I'm not against
numeric use; I'm just saying that int8 is sufficient.


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2011-12-06T18:06:36Z

On Tue, Dec 6, 2011 at 1:00 PM, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 06-12-2011 13:11, Robert Haas wrote:
>> On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
>>> I've been considering similar things, as you can find in the archives,
>>> but what I was thinking of was converting the number to just a plain
>>> bigint, then letting the user apply whatever arithmetic wanted at the
>>> SQL level. I never got around to acutally coding it, though. It could
>>> easily be extracted from your patch of course - and I think that's a
>>> more flexible approach. Is there some advantage to your method that
>>> I'm missing?
>>
>> I went so far as to put together an lsn data type.  I didn't actually
>> get all that far with it, which is why I haven't posted it sooner, but
>> here's what I came up with.  It's missing indexing support and stuff,
>> but that could be added if people like the approach.  It solves this
>> problem by implementing -(lsn,lsn) => numeric (not int8, that can
>> overflow since it is not unsigned), which allows an lsn => numeric
>> conversion by just subtracting '0/0'::lsn.
>>
> Interesting approach. I don't want to go that far. If so, you want to change
> all of those functions that deal with LSNs and add some implicit conversion
> between text and lsn data types (for backward compatibility). As of int8, I'm
> not aware of any modern plataform that int8 is not 64 bits. I'm not against
> numeric use; I'm just saying that int8 is sufficient.

The point isn't that int8 might not be 64 bits - of course it has to
be 64 bits; that's why it's called int8 i.e. 8 bytes.  The point is
that a large enough LSN, represented as an int8, will come out as a
negative values.  int8 can only represent 2^63 *non-negative* values,
because one bit is reserved for sign.

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

Re: xlog location arithmetic

Jim Nasby <jim@nasby.net> — 2011-12-13T17:48:22Z

On Dec 6, 2011, at 12:06 PM, Robert Haas wrote:
> On Tue, Dec 6, 2011 at 1:00 PM, Euler Taveira de Oliveira
> <euler@timbira.com> wrote:
>> On 06-12-2011 13:11, Robert Haas wrote:
>>> On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
>>>> I've been considering similar things, as you can find in the archives,
>>>> but what I was thinking of was converting the number to just a plain
>>>> bigint, then letting the user apply whatever arithmetic wanted at the
>>>> SQL level. I never got around to acutally coding it, though. It could
>>>> easily be extracted from your patch of course - and I think that's a
>>>> more flexible approach. Is there some advantage to your method that
>>>> I'm missing?
>>> 
>>> I went so far as to put together an lsn data type.  I didn't actually
>>> get all that far with it, which is why I haven't posted it sooner, but
>>> here's what I came up with.  It's missing indexing support and stuff,
>>> but that could be added if people like the approach.  It solves this
>>> problem by implementing -(lsn,lsn) => numeric (not int8, that can
>>> overflow since it is not unsigned), which allows an lsn => numeric
>>> conversion by just subtracting '0/0'::lsn.
>>> 
>> Interesting approach. I don't want to go that far. If so, you want to change
>> all of those functions that deal with LSNs and add some implicit conversion
>> between text and lsn data types (for backward compatibility). As of int8, I'm
>> not aware of any modern plataform that int8 is not 64 bits. I'm not against
>> numeric use; I'm just saying that int8 is sufficient.
> 
> The point isn't that int8 might not be 64 bits - of course it has to
> be 64 bits; that's why it's called int8 i.e. 8 bytes.  The point is
> that a large enough LSN, represented as an int8, will come out as a
> negative values.  int8 can only represent 2^63 *non-negative* values,
> because one bit is reserved for sign.

I've often wondered about adding uint2/4/8... I suspect it's actually pretty uncommon for people to put negative numbers into int fields, since one of their biggest uses seems to be surrogate keys.

I realize that this opens a can of worms with casting, but perhaps that can be kept under control by not doing any implicit casting between int and uint... that just means that we'd have to be smart about casting from unknown, but hopefully that's doable since we already have a similar concern with casting unknown to int2/4/8 vs numeric?
--
Jim C. Nasby, Database Architect                   jim@nasby.net
512.569.9461 (cell)                         http://jim.nasby.net

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2011-12-13T20:37:10Z

On Tue, Dec 13, 2011 at 12:48 PM, Jim Nasby <jim@nasby.net> wrote:
> I've often wondered about adding uint2/4/8... I suspect it's actually pretty uncommon for people to put negative numbers into int fields, since one of their biggest uses seems to be surrogate keys.
>
> I realize that this opens a can of worms with casting, but perhaps that can be kept under control by not doing any implicit casting between int and uint... that just means that we'd have to be smart about casting from unknown, but hopefully that's doable since we already have a similar concern with casting unknown to int2/4/8 vs numeric?

I've wondered about it too, but it seems like too large a can of worms
to open just to address this case.  Returning the value as numeric is
hardly a disaster; the user can always downcast to int8 if they really
want, and as long as it's < 2^63 (which in practice it virtually
always will be) it will work.  It's not clear what the point of this
is since for typical values numeric is going to take up less storage
anyway (e.g. 1000001 is 7 bytes on disk as a numeric), not to mention
that it only requires 4-byte alignment rather than 8-byte alignment,
and probably no one does enough arithmetic with LSN values for any
speed penalty to matter even slightly, but it should work.

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

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2011-12-20T10:27:41Z

On Tue, Dec 6, 2011 at 19:06, Robert Haas <robertmhaas@gmail.com> wrote:
> On Tue, Dec 6, 2011 at 1:00 PM, Euler Taveira de Oliveira
> <euler@timbira.com> wrote:
>> On 06-12-2011 13:11, Robert Haas wrote:
>>> On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
>>>> I've been considering similar things, as you can find in the archives,
>>>> but what I was thinking of was converting the number to just a plain
>>>> bigint, then letting the user apply whatever arithmetic wanted at the
>>>> SQL level. I never got around to acutally coding it, though. It could
>>>> easily be extracted from your patch of course - and I think that's a
>>>> more flexible approach. Is there some advantage to your method that
>>>> I'm missing?
>>>
>>> I went so far as to put together an lsn data type.  I didn't actually
>>> get all that far with it, which is why I haven't posted it sooner, but
>>> here's what I came up with.  It's missing indexing support and stuff,
>>> but that could be added if people like the approach.  It solves this
>>> problem by implementing -(lsn,lsn) => numeric (not int8, that can
>>> overflow since it is not unsigned), which allows an lsn => numeric
>>> conversion by just subtracting '0/0'::lsn.
>>>
>> Interesting approach. I don't want to go that far. If so, you want to change
>> all of those functions that deal with LSNs and add some implicit conversion
>> between text and lsn data types (for backward compatibility). As of int8, I'm

As long as you have the conversion, you don't really need to change
them, do you? It might be nice in some ways, but this is still a
pretty internal operation, so I don't see it as critical.

>> not aware of any modern plataform that int8 is not 64 bits. I'm not against
>> numeric use; I'm just saying that int8 is sufficient.
>
> The point isn't that int8 might not be 64 bits - of course it has to
> be 64 bits; that's why it's called int8 i.e. 8 bytes.  The point is
> that a large enough LSN, represented as an int8, will come out as a
> negative values.  int8 can only represent 2^63 *non-negative* values,
> because one bit is reserved for sign.

Doing it in numeric should be perfectly fine. The only real reason to
pick int8 over in this context would be performance, but it's not like
this is something that's going to be called in really performance
critical paths...

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2011-12-20T13:08:50Z

On 20-12-2011 07:27, Magnus Hagander wrote:
> On Tue, Dec 6, 2011 at 19:06, Robert Haas <robertmhaas@gmail.com> wrote:
>> On Tue, Dec 6, 2011 at 1:00 PM, Euler Taveira de Oliveira
>> <euler@timbira.com> wrote:
>>> On 06-12-2011 13:11, Robert Haas wrote:
>>>> On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
>>>>> I've been considering similar things, as you can find in the archives,
>>>>> but what I was thinking of was converting the number to just a plain
>>>>> bigint, then letting the user apply whatever arithmetic wanted at the
>>>>> SQL level. I never got around to acutally coding it, though. It could
>>>>> easily be extracted from your patch of course - and I think that's a
>>>>> more flexible approach. Is there some advantage to your method that
>>>>> I'm missing?
>>>>
>>>> I went so far as to put together an lsn data type.  I didn't actually
>>>> get all that far with it, which is why I haven't posted it sooner, but
>>>> here's what I came up with.  It's missing indexing support and stuff,
>>>> but that could be added if people like the approach.  It solves this
>>>> problem by implementing -(lsn,lsn) => numeric (not int8, that can
>>>> overflow since it is not unsigned), which allows an lsn => numeric
>>>> conversion by just subtracting '0/0'::lsn.
>>>>
>>> Interesting approach. I don't want to go that far. If so, you want to change
>>> all of those functions that deal with LSNs and add some implicit conversion
>>> between text and lsn data types (for backward compatibility). As of int8, I'm
> 
> As long as you have the conversion, you don't really need to change
> them, do you? It might be nice in some ways, but this is still a
> pretty internal operation, so I don't see it as critical.
> 
For correctness, yes.

At this point, my question is: do we want to support the lsn data type idea or
a basic function that implements the difference between LSNs?


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2011-12-23T11:54:22Z

On Tue, Dec 20, 2011 at 14:08, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 20-12-2011 07:27, Magnus Hagander wrote:
>> On Tue, Dec 6, 2011 at 19:06, Robert Haas <robertmhaas@gmail.com> wrote:
>>> On Tue, Dec 6, 2011 at 1:00 PM, Euler Taveira de Oliveira
>>> <euler@timbira.com> wrote:
>>>> On 06-12-2011 13:11, Robert Haas wrote:
>>>>> On Tue, Dec 6, 2011 at 5:14 AM, Magnus Hagander <magnus@hagander.net> wrote:
>>>>>> I've been considering similar things, as you can find in the archives,
>>>>>> but what I was thinking of was converting the number to just a plain
>>>>>> bigint, then letting the user apply whatever arithmetic wanted at the
>>>>>> SQL level. I never got around to acutally coding it, though. It could
>>>>>> easily be extracted from your patch of course - and I think that's a
>>>>>> more flexible approach. Is there some advantage to your method that
>>>>>> I'm missing?
>>>>>
>>>>> I went so far as to put together an lsn data type.  I didn't actually
>>>>> get all that far with it, which is why I haven't posted it sooner, but
>>>>> here's what I came up with.  It's missing indexing support and stuff,
>>>>> but that could be added if people like the approach.  It solves this
>>>>> problem by implementing -(lsn,lsn) => numeric (not int8, that can
>>>>> overflow since it is not unsigned), which allows an lsn => numeric
>>>>> conversion by just subtracting '0/0'::lsn.
>>>>>
>>>> Interesting approach. I don't want to go that far. If so, you want to change
>>>> all of those functions that deal with LSNs and add some implicit conversion
>>>> between text and lsn data types (for backward compatibility). As of int8, I'm
>>
>> As long as you have the conversion, you don't really need to change
>> them, do you? It might be nice in some ways, but this is still a
>> pretty internal operation, so I don't see it as critical.
>>
> For correctness, yes.
>
> At this point, my question is: do we want to support the lsn data type idea or
> a basic function that implements the difference between LSNs?

Personally I think a function is enough - it solves the only case that
I've actually seen. But a datatype would be a more complete solution,
of course - but it seems a bit of an overkill to me. Not really sure
which way we should go - I was hoping somebody else would comment as
well..

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2011-12-23T15:05:54Z

Magnus Hagander <magnus@hagander.net> writes:
>> At this point, my question is: do we want to support the lsn data type idea or
>> a basic function that implements the difference between LSNs?

> Personally I think a function is enough - it solves the only case that
> I've actually seen. But a datatype would be a more complete solution,
> of course - but it seems a bit of an overkill to me. Not really sure
> which way we should go - I was hoping somebody else would comment as
> well..

I too think a datatype is overkill, if we're only planning on providing
one function.  Just emit the values as numeric and have done.

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2011-12-23T15:13:41Z

On Fri, Dec 23, 2011 at 10:05 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I too think a datatype is overkill, if we're only planning on providing
> one function.

Are there any other functions we ought to provide?

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

Re: xlog location arithmetic

Andrew Dunstan <andrew@dunslane.net> — 2011-12-23T15:13:42Z


On 12/23/2011 10:05 AM, Tom Lane wrote:
> Magnus Hagander<magnus@hagander.net>  writes:
>>> At this point, my question is: do we want to support the lsn data type idea or
>>> a basic function that implements the difference between LSNs?
>> Personally I think a function is enough - it solves the only case that
>> I've actually seen. But a datatype would be a more complete solution,
>> of course - but it seems a bit of an overkill to me. Not really sure
>> which way we should go - I was hoping somebody else would comment as
>> well..
> I too think a datatype is overkill, if we're only planning on providing
> one function.  Just emit the values as numeric and have done.
>
> 			

+1.

cheers

andrew

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2011-12-23T15:18:21Z

Robert Haas <robertmhaas@gmail.com> writes:
> On Fri, Dec 23, 2011 at 10:05 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> I too think a datatype is overkill, if we're only planning on providing
>> one function.

> Are there any other functions we ought to provide?

Even if there are several, what exact advantage does a datatype offer
over representing LSN values as numerics?  It seems to me to be adding
complication and extra code (I/O converters at least) for very little
gain.

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2011-12-23T15:32:29Z

On Fri, Dec 23, 2011 at 10:18 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Fri, Dec 23, 2011 at 10:05 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> I too think a datatype is overkill, if we're only planning on providing
>>> one function.
>
>> Are there any other functions we ought to provide?
>
> Even if there are several, what exact advantage does a datatype offer
> over representing LSN values as numerics?  It seems to me to be adding
> complication and extra code (I/O converters at least) for very little
> gain.

I guess I'm just constitutionally averse to labeling things as "text"
when they really aren't.  I do it all the time in Perl, of course, but
in PostgreSQL we have strong data typing, and it seems like we might
as well use it.

Also, we've occasionally talked (in the light of Pavan's single-pass
vacuum patch, for example) about needing to store LSNs in system
catalogs; and we're certainly not going to want to do that as text.
I'll admit that it's not 100% clear that anything like this will ever
happen, though, so maybe it's premature to worry about it.

I can see I'm in the minority on this one, though...

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

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2011-12-23T15:59:35Z

Robert Haas <robertmhaas@gmail.com> writes:
> On Fri, Dec 23, 2011 at 10:18 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Even if there are several, what exact advantage does a datatype offer
>> over representing LSN values as numerics?  It seems to me to be adding
>> complication and extra code (I/O converters at least) for very little
>> gain.

> I guess I'm just constitutionally averse to labeling things as "text"
> when they really aren't.

Er ... text?  I thought the proposal was to use numeric.

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2011-12-23T16:15:07Z

On Fri, Dec 23, 2011 at 10:59 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Fri, Dec 23, 2011 at 10:18 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Even if there are several, what exact advantage does a datatype offer
>>> over representing LSN values as numerics?  It seems to me to be adding
>>> complication and extra code (I/O converters at least) for very little
>>> gain.
>
>> I guess I'm just constitutionally averse to labeling things as "text"
>> when they really aren't.
>
> Er ... text?  I thought the proposal was to use numeric.

The proposal is to make a function that takes a text argument (which
is really an LSN, but we choose to represent it as text) and returns
numeric.

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

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-01-14T14:06:34Z

On Tue, Dec 6, 2011 at 1:19 PM, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> Hi,
>
> A while ago when blogging about WAL [1], I noticed a function to deal with
> xlog location arithmetic is wanted. I remembered Depez [2] mentioning it and
> after some questions during trainings and conferences I decided to translate
> my shell script function in C.
>
> The attached patch implements the function pg_xlog_location_diff (bikeshed
> colors are welcome). It calculates the difference between two given
> transaction log locations. Now that we have pg_stat_replication view, it will
> be easy to get the lag just passing columns as parameters. Also, the
> monitoring tools could take advantage of it instead of relying on a fragile
> routine to get the lag.
>
> I noticed that pg_xlogfile_name* functions does not sanity check the xrecoff
> boundaries but that is material for another patch.
>
>
> [1] http://eulerto.blogspot.com/2011/11/understanding-wal-nomenclature.html
> [2]
> http://www.depesz.com/index.php/2011/01/24/waiting-for-9-1-pg_stat_replication/

I think that this function is very useful. Can you add the patch into
CommitFest 2012-1 ?

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2012-01-14T14:12:21Z

On 14-01-2012 11:06, Fujii Masao wrote:
> I think that this function is very useful. Can you add the patch into
> CommitFest 2012-1 ?
> 
Sure. But I must adjust the patch based on the thread comments (basically,
numeric output). I have a new patch but need to test it before submitting it.
I'll post this weekend.


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Greg Smith <greg@2ndquadrant.com> — 2012-01-15T01:18:31Z

On 01/14/2012 09:12 AM, Euler Taveira de Oliveira wrote:
> But I must adjust the patch based on the thread comments (basically,
> numeric output). I have a new patch but need to test it before submitting it.
> I'll post this weekend.

It's now at https://commitfest.postgresql.org/action/patch_view?id=776 
listed as waiting on you right now.  It's good to put patches into the 
CF application early.  Helps planning, and gives a safety net against 
all sorts of things.  We wouldn't want something this obviously useful 
to get kicked out if, for example, you lost your Internet connection 
over the weekend and then didn't technically qualify as having submitted 
it there before the deadline.  As someone who sweated today for two 
hours when my power at home was turned off to install a new circuit 
breaker, I'm feeling particularly paranoid right now about that sort of 
thing here.

The fact that you got some review feedback before the official CF start 
doesn't mean you can't be listed there right now.  In fact, those are 
things I like to see tracked.  Having links to all of the e-mail 
messages that were important turning points for a feature that changed 
during review is very helpful to reviewers and committers.  And the 
easiest way to keep up with that is to start as early as possible:  add 
it to the app right after the first patch submission.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com

Re: xlog location arithmetic

Gurjeet Singh <singh.gurjeet@gmail.com> — 2012-01-15T03:49:00Z

On Sat, Jan 14, 2012 at 8:18 PM, Greg Smith <greg@2ndquadrant.com> wrote:

> On 01/14/2012 09:12 AM, Euler Taveira de Oliveira wrote:
>
>> But I must adjust the patch based on the thread comments (basically,
>> numeric output). I have a new patch but need to test it before submitting
>> it.
>> I'll post this weekend.
>>
>
> It's now at https://commitfest.postgresql.**org/action/patch_view?id=776<https://commitfest.postgresql.org/action/patch_view?id=776>listed as waiting on you right now.  It's good to put patches into the CF
> application early.  Helps planning, and gives a safety net against all
> sorts of things.  We wouldn't want something this obviously useful to get
> kicked out if, for example, you lost your Internet connection over the
> weekend and then didn't technically qualify as having submitted it there
> before the deadline.  As someone who sweated today for two hours when my
> power at home was turned off to install a new circuit breaker, I'm feeling
> particularly paranoid right now about that sort of thing here.
> he patch
> The fact that you got some review feedback before the official CF start
> doesn't mean you can't be listed there right now.  In fact, those are
> things I like to see tracked.  Having links to all of the e-mail messages
> that were important turning points for a feature that changed during review
> is very helpful to reviewers and committers.  And the easiest way to keep
> up with that is to start as early as possible:  add it to the app right
> after the first patch submission.


I agree.

So lets make it easy for the patch submitter to start the process. I
propose that we have a page in the CF application where people can
upload/attach the patch, and the app posts the patch to -hackers and uses
the post URL to create the CF entry.

Regards,
-- 
Gurjeet Singh
EnterpriseDB Corporation
The Enterprise PostgreSQL Company

Re: automating CF submissions (was xlog location arithmetic)

Greg Smith <greg@2ndquadrant.com> — 2012-01-15T04:44:08Z

On 01/14/2012 10:49 PM, Gurjeet Singh wrote:
> So lets make it easy for the patch submitter to start the process. I 
> propose that we have a page in the CF application where people can 
> upload/attach the patch, and the app posts the patch to -hackers and 
> uses the post URL to create the CF entry.
>

That would be nice, but there's at least two serious problems with it, 
which I would guess are both unsolvable without adding an unsupportable 
amount of work to the current PostgreSQL web team.  First, it is 
technically risky for a web application hosted on postgresql.org to be 
e-mailing this list.  There are some things in the infrastructure that 
do that already--I believe the pgsql-commiters list being driven from 
commits is the busiest such bot.  But all of the ones that currently 
exist are either moderated, have a limited number of approved 
submitters, or both.

If it were possible for a bot to create a postgresql.org community 
account, then trigger an e-mail to pgsql-hackers just by filling out a 
web form, I'd give it maybe six months before it has to be turned off 
for a bit--because there are thousands messages queued up once the first 
bored spammer figures that out.  Securing web to e-mail gateways is a 
giant headache, and everyone working on the PostgreSQL infrastructure 
who might work on that is already overloaded with community volunteer 
work.  There's an element of zero-sum game here:  while this would 
provide some assistance to new contributors, the time to build and 
maintain the thing would be coming mainly out of senior contributors.  I 
see the gain+risk vs. reward here skewed the wrong way.

Second, e-mail provides some level of validation that patches being 
submitted are coming from the person they claim.  We currently reject 
patches that are only shared with the community on the web, via places 
like github.  The process around this mailing list tries to make it 
clear sending patches to here is a code submission under the PostgreSQL 
license.  And e-mail nowadays keeps increasing the number of checks that 
confirm it's coming from the person it claims sent it.  I can go check 
into the DKIM credentials your Gmail message to the list contained if 
I'd like, to help confirm it really came from your account.  E-mail 
headers are certainly not perfectly traceable and audit-able, but they 
are far better than what you'd get from a web submission.  Little audit 
trail there beyond "came from this IP address".

One unicorn I would like to have here would give the CF app a database 
of recent e-mails to pgsql-hackers.  I login to the CF app, click on 
"Add recent submission", and anything matching my e-mail address appears 
with a checkbox next to it.  Click on the patch submissions, and then 
something like you described would happen.  That would save me the 
annoying work around looking up message IDs so much.

The role CF manager would benefit even more from infrastructure like 
that too.  Something that listed all the recent e-mail messages for an 
existing submission, such that you could just click on the ones that you 
wanted added to the patch's e-mail history, would save me personally 
enough time that I could probably even justify writing it.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com

Re: automating CF submissions (was xlog location arithmetic)

Magnus Hagander <magnus@hagander.net> — 2012-01-15T08:17:54Z

On Sun, Jan 15, 2012 at 05:44, Greg Smith <greg@2ndquadrant.com> wrote:
> On 01/14/2012 10:49 PM, Gurjeet Singh wrote:
>>
>> So lets make it easy for the patch submitter to start the process. I
>> propose that we have a page in the CF application where people can
>> upload/attach the patch, and the app posts the patch to -hackers and uses
>> the post URL to create the CF entry.
>>
>
> That would be nice, but there's at least two serious problems with it, which
> I would guess are both unsolvable without adding an unsupportable amount of
> work to the current PostgreSQL web team.  First, it is technically risky for
> a web application hosted on postgresql.org to be e-mailing this list.  There
> are some things in the infrastructure that do that already--I believe the
> pgsql-commiters list being driven from commits is the busiest such bot.  But
> all of the ones that currently exist are either moderated, have a limited
> number of approved submitters, or both.

It's not really a problem from that perspective, as long as it
requires the user to be logged in. The mail would be sent from the
users account, with that one as a sender, and thus be exposed to the
same moderation rules as the rest of the list posts.

> If it were possible for a bot to create a postgresql.org community account,
> then trigger an e-mail to pgsql-hackers just by filling out a web form, I'd
> give it maybe six months before it has to be turned off for a bit--because
> there are thousands messages queued up once the first bored spammer figures

Said bot can already use the bug report form *without*  having to sign
up for an account.

Or said bot could submit news or events, which trigger an email to at
least some lists, which hasn't bene done.

It's supposedly not easy for a bot to sign up for a community account,
since it requires you to have access to the email address it's
registered on. If that doesn't work, it's a bug and needs to be fixed
regardless.

> that out.  Securing web to e-mail gateways is a giant headache, and everyone
> working on the PostgreSQL infrastructure who might work on that is already
> overloaded with community volunteer work.  There's an element of zero-sum

We've already solved that problem for other situtations, and given how
the infrastructure is built, that's fairly easy to replicate to
another node.

I think the bigger problem is "who'll write it". AFAIK, the CF app
*itself* is even more person- and time-constrained to senior
developers (Robert Haas only) than the infrastructure, and that's a
bigger problem. There are already a bunch of things that are a lot
simpler than this that has been pending on that one for well over half
a year.

> Second, e-mail provides some level of validation that patches being
> submitted are coming from the person they claim.  We currently reject
> patches that are only shared with the community on the web, via places like
> github.  The process around this mailing list tries to make it clear sending
> patches to here is a code submission under the PostgreSQL license.  And
> e-mail nowadays keeps increasing the number of checks that confirm it's
> coming from the person it claims sent it.  I can go check into the DKIM
> credentials your Gmail message to the list contained if I'd like, to help
> confirm it really came from your account.  E-mail headers are certainly not

I think DKIM was a bad example, because AFAIK our lists mangle DKIM
and thus actually show them as *invalid* for at least the majority of
messages...

> One unicorn I would like to have here would give the CF app a database of
> recent e-mails to pgsql-hackers.  I login to the CF app, click on "Add
> recent submission", and anything matching my e-mail address appears with a
> checkbox next to it.  Click on the patch submissions, and then something
> like you described would happen.  That would save me the annoying work
> around looking up message IDs so much.

That would be neat.

And FWIW, I'd find it a lot more useful for the CF app to have the
ability to post *reviews* in it, that would end up being properly
threaded. The way it is now, half the reviewers create a *new* thread
to post their reviews on, making it a PITA to keep track of those
patches on the list at all, which somewhat takes away the whole idea
of "mail being the primary way to track it". Not saying it's critical,
but I'd put it a lot higher on the list than being able to post the
initial patch.

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: automating CF submissions (was xlog location arithmetic)

Greg Smith <greg@2ndquadrant.com> — 2012-01-15T08:37:30Z

On 01/15/2012 03:17 AM, Magnus Hagander wrote:
> And FWIW, I'd find it a lot more useful for the CF app to have the
> ability to post *reviews* in it, that would end up being properly
> threaded.

Next you'll be saying we should have some sort of web application to 
help with the whole review process, show the work on an integrated sort 
of Review Board or something.  What crazy talk.

My contribution toward patch review ease for this week is that peg is 
quite a bit smarter about referring to the correct part of the origin 
git repo now, when you've been working on a branch for a while then 
create a new one:  https://github.com/gregs1104/peg

Last week's was that I confirmed that on a Mac using Homebrew for 
package management, after "brew install postgresql" to get the 
dependencies in, you can then use peg to setup a PostgreSQL in your home 
directory for patch testing or development.  Works fine out of the box, 
you just won't have things like all the PLs installed.

Yes, I am aware I'm going at this bottom-up.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com

Re: automating CF submissions (was xlog location arithmetic)

Magnus Hagander <magnus@hagander.net> — 2012-01-15T08:45:25Z

On Sun, Jan 15, 2012 at 09:37, Greg Smith <greg@2ndquadrant.com> wrote:
> On 01/15/2012 03:17 AM, Magnus Hagander wrote:
>>
>> And FWIW, I'd find it a lot more useful for the CF app to have the
>> ability to post *reviews* in it, that would end up being properly
>> threaded.
>
>
> Next you'll be saying we should have some sort of web application to help
> with the whole review process, show the work on an integrated sort of Review
> Board or something.  What crazy talk.

Well, it's early in the morning for being sunday, I blame that.


> My contribution toward patch review ease for this week is that peg is quite
> a bit smarter about referring to the correct part of the origin git repo
> now, when you've been working on a branch for a while then create a new one:
>  https://github.com/gregs1104/peg

Being able to refer to a git branch is one of those things that have
been on the todo list for the cf app since pgcon last year...

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: automating CF submissions (was xlog location arithmetic)

Robert Haas <robertmhaas@gmail.com> — 2012-01-16T17:57:25Z

On Sun, Jan 15, 2012 at 3:45 AM, Magnus Hagander <magnus@hagander.net> wrote:
> On Sun, Jan 15, 2012 at 09:37, Greg Smith <greg@2ndquadrant.com> wrote:
>> On 01/15/2012 03:17 AM, Magnus Hagander wrote:
>>> And FWIW, I'd find it a lot more useful for the CF app to have the
>>> ability to post *reviews* in it, that would end up being properly
>>> threaded.
>>
>> Next you'll be saying we should have some sort of web application to help
>> with the whole review process, show the work on an integrated sort of Review
>> Board or something.  What crazy talk.
>
> Well, it's early in the morning for being sunday, I blame that.
>
>> My contribution toward patch review ease for this week is that peg is quite
>> a bit smarter about referring to the correct part of the origin git repo
>> now, when you've been working on a branch for a while then create a new one:
>>  https://github.com/gregs1104/peg
>
> Being able to refer to a git branch is one of those things that have
> been on the todo list for the cf app since pgcon last year...

Do we have an actual written TODO list for the cf app somewhere?  If
not, I think creating one would be a good idea.  I realize I've been
remiss in addressing some of the things people want, but the lack of
any centralized place where such items are collected doesn't make it
simpler.

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

Re: automating CF submissions (was xlog location arithmetic)

Magnus Hagander <magnus@hagander.net> — 2012-01-16T17:58:29Z

On Mon, Jan 16, 2012 at 18:57, Robert Haas <robertmhaas@gmail.com> wrote:
> On Sun, Jan 15, 2012 at 3:45 AM, Magnus Hagander <magnus@hagander.net> wrote:
>> On Sun, Jan 15, 2012 at 09:37, Greg Smith <greg@2ndquadrant.com> wrote:
>>> On 01/15/2012 03:17 AM, Magnus Hagander wrote:
>>>> And FWIW, I'd find it a lot more useful for the CF app to have the
>>>> ability to post *reviews* in it, that would end up being properly
>>>> threaded.
>>>
>>> Next you'll be saying we should have some sort of web application to help
>>> with the whole review process, show the work on an integrated sort of Review
>>> Board or something.  What crazy talk.
>>
>> Well, it's early in the morning for being sunday, I blame that.
>>
>>> My contribution toward patch review ease for this week is that peg is quite
>>> a bit smarter about referring to the correct part of the origin git repo
>>> now, when you've been working on a branch for a while then create a new one:
>>>  https://github.com/gregs1104/peg
>>
>> Being able to refer to a git branch is one of those things that have
>> been on the todo list for the cf app since pgcon last year...
>
> Do we have an actual written TODO list for the cf app somewhere?  If
> not, I think creating one would be a good idea.  I realize I've been
> remiss in addressing some of the things people want, but the lack of
> any centralized place where such items are collected doesn't make it
> simpler.

I don't think so - I've been keeping mine in your mailbox ;)

A simple wiki page is probably enough - going for an actual tracker or
anything seems vastly overkill...

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: automating CF submissions (was xlog location arithmetic)

Josh Berkus <josh@agliodbs.com> — 2012-01-16T20:48:41Z

On 1/14/12 8:44 PM, Greg Smith wrote:
> Second, e-mail provides some level of validation that patches being
> submitted are coming from the person they claim.  We currently reject
> patches that are only shared with the community on the web, via places
> like github.  The process around this mailing list tries to make it
> clear sending patches to here is a code submission under the PostgreSQL
> license.  And e-mail nowadays keeps increasing the number of checks that
> confirm it's coming from the person it claims sent it.  I can go check
> into the DKIM credentials your Gmail message to the list contained if
> I'd like, to help confirm it really came from your account.  E-mail
> headers are certainly not perfectly traceable and audit-able, but they
> are far better than what you'd get from a web submission.  Little audit
> trail there beyond "came from this IP address".

Putting submitters aside, I have to say based on teaching people how to
use the CF stuff on Thursday night that the process of submitting a
review of a patch is VERY unintuitive, or in the words of one reviewer
"astonishingly arcane".  Summing up:

1. Log into CF.  Claim the patch by editing it.

2. Write a review and email it to pgsql-hackers.

3. Dig the messageID out of your sent mail.

4. Add a comment to the patch, type "Review" with the messageID, and
ideally a short summary comment of the review.

5. Edit the patch to change its status as well as to remove yourself as
reviewer if you plan to do no further review.

There are so many things wrong with this workflow I wouldn't know where
to start.  The end result, though, is that it strongly discourages the
occasional reviewer by making the review process cumbersome and confusing.

I'll also point out that the process for *applying* a patch, if you
don't subscribe to hackers and keep archives around on your personal
machine for months, is also very cumbersome and error-prone.  Copy and
paste from a web page?  Really?

Certainly we could spend the next 6 years incrementally improving the CF
app "in our spare time".  But maybe it might be a better thing to look
at the code development tools which are already available?

-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-16T21:10:23Z

Excerpts from Josh Berkus's message of lun ene 16 17:48:41 -0300 2012:

> Putting submitters aside, I have to say based on teaching people how to
> use the CF stuff on Thursday night that the process of submitting a
> review of a patch is VERY unintuitive, or in the words of one reviewer
> "astonishingly arcane".  Summing up:
> 
> 1. Log into CF.  Claim the patch by editing it.
> 
> 2. Write a review and email it to pgsql-hackers.
> 
> 3. Dig the messageID out of your sent mail.
> 
> 4. Add a comment to the patch, type "Review" with the messageID, and
> ideally a short summary comment of the review.
> 
> 5. Edit the patch to change its status as well as to remove yourself as
> reviewer if you plan to do no further review.
> 
> There are so many things wrong with this workflow I wouldn't know where
> to start.

Other than having to figure out Message-Ids, which most MUAs seem to
hide as much as possible, is there anything here of substance?  I mean,
if getting a message-id from Gmail is all that complicated, please
complain to Google.

I mean, is email arcane?  Surely not.  Are summary lines arcane?  Give
me a break.  So the only real complain point here is message-id, which
normally people don't care about and don't even know they exist.  So
they have to learn about it.

Let's keep in mind that pgsql-hackers email is our preferred form of
communication.

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Josh Berkus <josh@agliodbs.com> — 2012-01-16T21:25:50Z

> I mean, is email arcane?  Surely not.  Are summary lines arcane?  Give
> me a break.  So the only real complain point here is message-id, which
> normally people don't care about and don't even know they exist.  So
> they have to learn about it.

The complaint is that the reviewer is expected to use two different and
wholly incompatible methods of communication, each of which requires a
separate registration, to post the review.

-- 
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com

Re: automating CF submissions (was xlog location arithmetic)

Jeff Janes <jeff.janes@gmail.com> — 2012-01-16T21:37:59Z

On Mon, Jan 16, 2012 at 1:10 PM, Alvaro Herrera
<alvherre@commandprompt.com> wrote:
>
> Excerpts from Josh Berkus's message of lun ene 16 17:48:41 -0300 2012:
>
>> Putting submitters aside, I have to say based on teaching people how to
>> use the CF stuff on Thursday night that the process of submitting a
>> review of a patch is VERY unintuitive, or in the words of one reviewer
>> "astonishingly arcane".  Summing up:
>>
>> 1. Log into CF.  Claim the patch by editing it.
>>
>> 2. Write a review and email it to pgsql-hackers.
>>
>> 3. Dig the messageID out of your sent mail.
>>
>> 4. Add a comment to the patch, type "Review" with the messageID, and
>> ideally a short summary comment of the review.
>>
>> 5. Edit the patch to change its status as well as to remove yourself as
>> reviewer if you plan to do no further review.
>>
>> There are so many things wrong with this workflow I wouldn't know where
>> to start.
>
> Other than having to figure out Message-Ids, which most MUAs seem to
> hide as much as possible, is there anything here of substance?

I find it an annoyance, but can't get too worked up over it.

> I mean,
> if getting a message-id from Gmail is all that complicated, please
> complain to Google.

But after digging the message-id out of gmail and entering it into the
commitfest app, the resulting link is broken because the email has not
yet shown up in the archives.  So now I have to wonder if I did
something wrong, and keep coming back every few hours to see if will
start working.

>
> I mean, is email arcane?  Surely not.  Are summary lines arcane?

The way you have to set them is pretty arcane.  Again, I can't get too
worked over it, but if it were made simpler I'd be happier.

Cheers,

Jeff

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-16T22:08:09Z

Excerpts from Jeff Janes's message of lun ene 16 18:37:59 -0300 2012:

> > I mean,
> > if getting a message-id from Gmail is all that complicated, please
> > complain to Google.
> 
> But after digging the message-id out of gmail and entering it into the
> commitfest app, the resulting link is broken because the email has not
> yet shown up in the archives.  So now I have to wonder if I did
> something wrong, and keep coming back every few hours to see if will
> start working.

Hours?  Unless a message is delayed for moderation, it should show up in
archives within tem minutes.  If you have problems finding emails after
that period, by all means complain.

Now that we've moved archives to a new host, perhaps we could rerun the
archive script every two minutes instead of ten.

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Greg Smith <greg@2ndquadrant.com> — 2012-01-16T22:25:50Z

On 01/16/2012 03:48 PM, Josh Berkus wrote:
> 3. Dig the messageID out of your sent mail.
>
> 4. Add a comment to the patch, type "Review" with the messageID, and
> ideally a short summary comment of the review.

This is the time consuming part that would benefit the most from some 
automation.  The message-id digging is an obvious sore spot, which is 
why I focused on improvements to eliminate so much of that first in my 
suggestions.  The problem is that we don't actually want every message 
sent to the list on a thread to appear on the CF summary, and writing 
that short summary content is an important step.

Archived messages deemed notable enough that someone linked the two are 
the only ones that appear in the patch history.  That makes it possible 
to come up to speed on the most interesting history points of a patch in 
a reasonable period of time--even if you missed the earlier discussion.  
I think any of the other alternatives we might adopt would end up 
associating all of the e-mail history around a patch.  That's the 
firehose, and spraying the CF app with it makes the whole thing a lot 
less useful.

I don't think this is an unsolvable area to improve.  It's been stuck 
behind the major postgresql.org site overhaul, which is done now.  
Adding some web service style APIs to probe the archives for message IDs 
by a) ancestor and b) author would make it possible to sand off a whole 
lot of rough edges here.  While it's annoying in its current form, doing 
all my work based on message IDs has been a huge improvement over the 
old approach, where URLs into the archives were date based and not 
always permanent.

> I'll also point out that the process for *applying* a patch, if you
> don't subscribe to hackers and keep archives around on your personal
> machine for months, is also very cumbersome and error-prone.  Copy and
> paste from a web page?  Really?

The most reasonable answer to this is for people to publish a git repo 
URL in addition to the "official" submission of changes to the list in 
patch form.  And momentum toward doing that just keeps going up, even 
among longer term contributors who weren't git advocates at all a year 
during the transition.  I nudged Simon that way and he's pushing 
branches for major patches but not small ones yet, it looks like Andrew 
fully embraced bitbucket recently, etc.

We're 16 months into git adoption.  I'm pretty happy with how well 
that's going.  We don't need to add infrastructure to enable people to 
push code to github and link to their branch comparison repo viewer as a 
way to view the patch; that's already available to anyone who wants is.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-16T22:40:38Z

Excerpts from Greg Smith's message of lun ene 16 19:25:50 -0300 2012:
> On 01/16/2012 03:48 PM, Josh Berkus wrote:

> > I'll also point out that the process for *applying* a patch, if you
> > don't subscribe to hackers and keep archives around on your personal
> > machine for months, is also very cumbersome and error-prone.  Copy and
> > paste from a web page?  Really?
> 
> The most reasonable answer to this is for people to publish a git repo 
> URL in addition to the "official" submission of changes to the list in 
> patch form.

It's expected that we'll get a more reasonable interface to attachments,
one that will allow you to download patches separately.  (Currently,
attachments that have mime types other than text/plain are already
downloadable separately).

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Andrew Dunstan <andrew@dunslane.net> — 2012-01-16T23:23:33Z

On 01/16/2012 05:25 PM, Greg Smith wrote:
>
> The most reasonable answer to this is for people to publish a git repo 
> URL in addition to the "official" submission of changes to the list in 
> patch form.  And momentum toward doing that just keeps going up, even 
> among longer term contributors who weren't git advocates at all a year 
> during the transition.  I nudged Simon that way and he's pushing 
> branches for major patches but not small ones yet, it looks like 
> Andrew fully embraced bitbucket recently, etc.
>
>

If we're going to do that, the refspec to be pulled needs to be a tag, I 
think, not just a branch, and people would have to get into the habit of 
tagging commits and explicitly pushing tags.

I probably should be doing that, and it is now built into the buildfarm 
client release mechanism, but I usually don't when just publishing dev 
work. Guess I need to start. I'll probably use tag names like 
branch-YYYYMMDDHHMM.

I certainly like the idea of just being able to pull in a tag from a 
remote instead of applying a patch.

(BTW, I use both bitbucket and github. They both have advantages.)

cheers

andrew

Re: xlog location arithmetic

Peter Geoghegan <peter@2ndquadrant.com> — 2012-01-17T14:46:07Z

On 20 December 2011 10:27, Magnus Hagander <magnus@hagander.net> wrote:
> Doing it in numeric should be perfectly fine. The only real reason to
> pick int8 over in this context would be performance, but it's not like
> this is something that's going to be called in really performance
> critical paths...

FYI, my group commit patch has a little macro, in the spirit of
XLByteAdvance, to get the delta between two LSNs in bytes as an
uint64.

-- 
Peter Geoghegan       http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training and Services

Re: automating CF submissions (was xlog location arithmetic)

Matteo Beccati <php@beccati.com> — 2012-01-17T15:33:27Z

On 16/01/2012 23:40, Alvaro Herrera wrote:
> 
> Excerpts from Greg Smith's message of lun ene 16 19:25:50 -0300 2012:
>> On 01/16/2012 03:48 PM, Josh Berkus wrote:
> 
>>> I'll also point out that the process for *applying* a patch, if you
>>> don't subscribe to hackers and keep archives around on your personal
>>> machine for months, is also very cumbersome and error-prone.  Copy and
>>> paste from a web page?  Really?
>>
>> The most reasonable answer to this is for people to publish a git repo 
>> URL in addition to the "official" submission of changes to the list in 
>> patch form.
> 
> It's expected that we'll get a more reasonable interface to attachments,
> one that will allow you to download patches separately.  (Currently,
> attachments that have mime types other than text/plain are already
> downloadable separately).

My proof of concept archive for the hackers ML site is still online, in
case anyone has trouble downloading the patches or just wants to have
the full thread handy.

All you need to do is to swap postgresql.org with beccati.org in the
"message-id" link:

http://archives.postgresql.org/message-id/1320343602-sup-2290@alvh.no-ip.org

->

http://archives.beccati.org/message-id/1320343602-sup-2290@alvh.no-ip.org


Cheers
-- 
Matteo Beccati

Development & Consulting - http://www.beccati.com/

Re: automating CF submissions (was xlog location arithmetic)

Andrew Dunstan <andrew@dunslane.net> — 2012-01-17T16:50:20Z

On 01/16/2012 05:40 PM, Alvaro Herrera wrote:
> It's expected that we'll get a more reasonable interface to attachments,
> one that will allow you to download patches separately.  (Currently,
> attachments that have mime types other than text/plain are already
> downloadable separately).

Are you really sure about that? My recent JSON patch is at 
<http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net>. 
I don't see any download link for the patch there, yet my mailer set the 
attachment type to text/x-patch, not text/plain.

cheers

andrew

Re: automating CF submissions (was xlog location arithmetic)

Peter Eisentraut <peter_e@gmx.net> — 2012-01-17T16:50:26Z

On mån, 2012-01-16 at 17:25 -0500, Greg Smith wrote:
> The most reasonable answer to this is for people to publish a git repo
> URL in addition to the "official" submission of changes to the list in
> patch form. 

Note that the original complaint was that for the occasional reviewer,
the current system takes at least 5 partially redundant steps in two
different systems.  I doubt that adding a third system and more
partially redundant steps it going to help that.

I don't have anything against the general idea, but it won't address the
original point.

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-17T16:50:41Z

Excerpts from Matteo Beccati's message of mar ene 17 12:33:27 -0300 2012:
> 
> On 16/01/2012 23:40, Alvaro Herrera wrote:
> > 
> > Excerpts from Greg Smith's message of lun ene 16 19:25:50 -0300 2012:
> >> On 01/16/2012 03:48 PM, Josh Berkus wrote:
> > 
> >>> I'll also point out that the process for *applying* a patch, if you
> >>> don't subscribe to hackers and keep archives around on your personal
> >>> machine for months, is also very cumbersome and error-prone.  Copy and
> >>> paste from a web page?  Really?
> >>
> >> The most reasonable answer to this is for people to publish a git repo 
> >> URL in addition to the "official" submission of changes to the list in 
> >> patch form.
> > 
> > It's expected that we'll get a more reasonable interface to attachments,
> > one that will allow you to download patches separately.  (Currently,
> > attachments that have mime types other than text/plain are already
> > downloadable separately).
> 
> My proof of concept archive for the hackers ML site is still online, in
> case anyone has trouble downloading the patches or just wants to have
> the full thread handy.

I was going to ping you about this, because I tried it when I wrote this
message and it got stuck waiting for response.

Now that we've migrated the website, it's time to get back to our
conversations about migrating archives to your stuff too.  How confident
with Django are you?

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Matteo Beccati <php@beccati.com> — 2012-01-17T17:10:11Z

On 17/01/2012 17:50, Alvaro Herrera wrote:
> 
> Excerpts from Matteo Beccati's message of mar ene 17 12:33:27 -0300 2012:
>> My proof of concept archive for the hackers ML site is still online, in
>> case anyone has trouble downloading the patches or just wants to have
>> the full thread handy.
> 
> I was going to ping you about this, because I tried it when I wrote this
> message and it got stuck waiting for response.

Hmm, works for me, e.g. the recently cited message:

http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net


> Now that we've migrated the website, it's time to get back to our
> conversations about migrating archives to your stuff too.  How confident
> with Django are you?

I've never wrote a line of Python in my life, so someone else should
work on porting the web part, I'm afraid...


Cheers
-- 
Matteo Beccati

Development & Consulting - http://www.beccati.com/

Re: automating CF submissions (was xlog location arithmetic)

Matteo Beccati <php@beccati.com> — 2012-01-17T17:22:12Z

On 17/01/2012 18:10, Matteo Beccati wrote:
> On 17/01/2012 17:50, Alvaro Herrera wrote:
>>
>> Excerpts from Matteo Beccati's message of mar ene 17 12:33:27 -0300 2012:
>>> My proof of concept archive for the hackers ML site is still online, in
>>> case anyone has trouble downloading the patches or just wants to have
>>> the full thread handy.
>>
>> I was going to ping you about this, because I tried it when I wrote this
>> message and it got stuck waiting for response.
> 
> Hmm, works for me, e.g. the recently cited message:
> 
> http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net

Erm... I meant

http://archives.beccati.org/message-id/4F12F9E5.3090801@dunslane.net

which redirects to:

http://archives.beccati.org/pgsql-hackers/message/305925

for me.


Cheers
-- 
Matteo Beccati

Development & Consulting - http://www.beccati.com/

Re: automating CF submissions (was xlog location arithmetic)

Tom Lane <tgl@sss.pgh.pa.us> — 2012-01-17T17:24:05Z

Andrew Dunstan <andrew@dunslane.net> writes:
> On 01/16/2012 05:40 PM, Alvaro Herrera wrote:
>> It's expected that we'll get a more reasonable interface to attachments,
>> one that will allow you to download patches separately.  (Currently,
>> attachments that have mime types other than text/plain are already
>> downloadable separately).

> Are you really sure about that? My recent JSON patch is at 
> <http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net>. 
> I don't see any download link for the patch there, yet my mailer set the 
> attachment type to text/x-patch, not text/plain.

Yeah, AFAICT the archives treat text/x-patch the same as text/plain.
I tend to send stuff that way if I mean it primarily to be read in the
email.  If I'm thinking people will download and apply it, it's better
to gzip the patch and pick a mime type appropriate to that, because that
makes it much easier to pull the patch off the archives at need, at the
cost that you can't just eyeball it in your mail reader.

Anyway, I agree with the general tenor of this thread that it'd be nice
to reduce the impedance mismatches a bit.  Don't have any great ideas
about specific ways to do that.

			regards, tom lane

Re: automating CF submissions (was xlog location arithmetic)

Greg Smith <greg@2ndquadrant.com> — 2012-01-17T18:19:13Z

On 01/17/2012 11:50 AM, Peter Eisentraut wrote:
> On mån, 2012-01-16 at 17:25 -0500, Greg Smith wrote:
>> The most reasonable answer to this is for people to publish a git repo
>> URL in addition to the "official" submission of changes to the list in
>> patch form.
> Note that the original complaint was that for the occasional reviewer,
> the current system takes at least 5 partially redundant steps in two
> different systems.  I doubt that adding a third system and more
> partially redundant steps it going to help that.

Publishing the submission via git is an extra step for the patch 
submitter.  If that happens, the reviewer can test just be cloning that, 
instead of first closing the PostgreSQL one then applying the patch.  It 
removes the "how do I fish the patch out of the archives?" problem from 
the reviewer's side of things.

-- 
Greg Smith   2ndQuadrant US    greg@2ndQuadrant.com   Baltimore, MD
PostgreSQL Training, Services, and 24x7 Support www.2ndQuadrant.com

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-17T20:17:05Z

Excerpts from Tom Lane's message of mar ene 17 14:24:05 -0300 2012:
> 
> Andrew Dunstan <andrew@dunslane.net> writes:
> > On 01/16/2012 05:40 PM, Alvaro Herrera wrote:
> >> It's expected that we'll get a more reasonable interface to attachments,
> >> one that will allow you to download patches separately.  (Currently,
> >> attachments that have mime types other than text/plain are already
> >> downloadable separately).
> 
> > Are you really sure about that? My recent JSON patch is at 
> > <http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net>. 
> > I don't see any download link for the patch there, yet my mailer set the 
> > attachment type to text/x-patch, not text/plain.
> 
> Yeah, AFAICT the archives treat text/x-patch the same as text/plain.

Right, maybe it's text/* or something like that.

> I tend to send stuff that way if I mean it primarily to be read in the
> email.  If I'm thinking people will download and apply it, it's better
> to gzip the patch and pick a mime type appropriate to that, because that
> makes it much easier to pull the patch off the archives at need, at the
> cost that you can't just eyeball it in your mail reader.

Maybe we could find a way to convince Mhonarc to present links to
download all mime parts separately, not only those that are
undisplayable.

> Anyway, I agree with the general tenor of this thread that it'd be nice
> to reduce the impedance mismatches a bit.  Don't have any great ideas
> about specific ways to do that.

I'm hopeful that the migration to the Archivopteryx stuff by Matteo will
improve things a bit.

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-18T01:23:13Z

Excerpts from Andrew Dunstan's message of mar ene 17 13:50:20 -0300 2012:
> 
> On 01/16/2012 05:40 PM, Alvaro Herrera wrote:
> > It's expected that we'll get a more reasonable interface to attachments,
> > one that will allow you to download patches separately.  (Currently,
> > attachments that have mime types other than text/plain are already
> > downloadable separately).
> 
> 
> Are you really sure about that? My recent JSON patch is at 
> <http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net>. 
> I don't see any download link for the patch there, yet my mailer set the 
> attachment type to text/x-patch, not text/plain.

I tweaked the Mhonarc config and now this attachment (as well as many
others) is shown as a downloadable link.  Please give it a look.

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-18T02:46:36Z

Excerpts from Alvaro Herrera's message of mar ene 17 22:23:13 -0300 2012:
> Excerpts from Andrew Dunstan's message of mar ene 17 13:50:20 -0300 2012:
> > 
> > On 01/16/2012 05:40 PM, Alvaro Herrera wrote:
> > > It's expected that we'll get a more reasonable interface to attachments,
> > > one that will allow you to download patches separately.  (Currently,
> > > attachments that have mime types other than text/plain are already
> > > downloadable separately).
> > 
> > 
> > Are you really sure about that? My recent JSON patch is at 
> > <http://archives.postgresql.org/message-id/4F12F9E5.3090801@dunslane.net>. 
> > I don't see any download link for the patch there, yet my mailer set the 
> > attachment type to text/x-patch, not text/plain.
> 
> I tweaked the Mhonarc config and now this attachment (as well as many
> others) is shown as a downloadable link.  Please give it a look.

Hm, I notice it works almost every patch I've checked, except the ones
from Tom such as this one:
http://archives.postgresql.org/message-id/4643.1326776814@sss.pgh.pa.us

The problem is that this one doesn't have the
Content-Disposition: attachment
line in the MIME header.  I don't know what we can do about it.

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Alex Shulgin <ash@commandprompt.com> — 2012-01-19T17:59:35Z

Greg Smith <greg@2ndQuadrant.com> writes:

> One unicorn I would like to have here would give the CF app a database
> of recent e-mails to pgsql-hackers.  I login to the CF app, click on
> "Add recent submission", and anything matching my e-mail address
> appears with a checkbox next to it.  Click on the patch submissions,
> and then something like you described would happen.  That would save
> me the annoying work around looking up message IDs so much.

Another idea: introduce some simple tag system in mails sent to -hackers
to be treated specially, e.g:

@fest add-to-current

to add new patch to the commit fest currently in progress, or

@fest add-to-next

to add it to the next scheduled fest.

Attribute your mail with

@fest comment COMMENT TEXT

or

@fest comment <<EOF
...
EOF

to add a (long) comment, ditto for "patch" and "review".

How does that sound?

--
Alex

Re: automating CF submissions (was xlog location arithmetic)

Andrew Dunstan <andrew@dunslane.net> — 2012-01-19T18:14:02Z


On 01/19/2012 12:59 PM, Alex Shulgin wrote:
> Greg Smith<greg@2ndQuadrant.com>  writes:
>
>> One unicorn I would like to have here would give the CF app a database
>> of recent e-mails to pgsql-hackers.  I login to the CF app, click on
>> "Add recent submission", and anything matching my e-mail address
>> appears with a checkbox next to it.  Click on the patch submissions,
>> and then something like you described would happen.  That would save
>> me the annoying work around looking up message IDs so much.
> Another idea: introduce some simple tag system in mails sent to -hackers
> to be treated specially, e.g:
>
> @fest add-to-current
>
> to add new patch to the commit fest currently in progress, or
>
> @fest add-to-next
>
> to add it to the next scheduled fest.
>
> Attribute your mail with
>
> @fest comment COMMENT TEXT
>
> or
>
> @fest comment<<EOF
> ...
> EOF
>
> to add a (long) comment, ditto for "patch" and "review".
>
> How does that sound?
>

Like a recipe for something that requires constant fixups, to be honest.

Seriously, adding something to the CF isn't *that* hard. I like Greg's 
idea of a list of recent emails that you could choose from.

cheers

andrew

Re: automating CF submissions (was xlog location arithmetic)

Alex Shulgin <ash@commandprompt.com> — 2012-01-19T18:41:54Z

Andrew Dunstan <andrew@dunslane.net> writes:

> On 01/19/2012 12:59 PM, Alex Shulgin wrote:
>
>> Another idea: introduce some simple tag system in mails sent to -hackers
>> to be treated specially, e.g:
>>
>> @fest add-to-current
>>
>> to add new patch to the commit fest currently in progress, or
>>
>> @fest add-to-next
>>
>> to add it to the next scheduled fest.
>>
>> Attribute your mail with
>>
>> @fest comment COMMENT TEXT
>>
>> or
>>
>> @fest comment<<EOF
>> ...
>> EOF
>>
>> to add a (long) comment, ditto for "patch" and "review".
>>
>> How does that sound?
>>
>
> Like a recipe for something that requires constant fixups, to be honest.
>
> Seriously, adding something to the CF isn't *that* hard. I like Greg's
> idea of a list of recent emails that you could choose from.

I've just added a comment about a patch and it took me to:

a. Login to commitfest app
b. Locate the patch and review I was replying to
c. Fetch archives thread index, refresh the index page for ~10 minutes
to see my reply appear
d. Copy message id and finally register comment in the commitfest app

(IIRC, something close to that was already described in this thread)

With the proposed approach it would only take me to include

@fest comment "Patch applies cleanly"

and possibly

@fest status Needs Review

to update the patch status and that'd be it.

--
Alex

PS: yes, I could just copy message id from the sent mail in my MUA, but
I like to make sure links I post aren't broke, so still I'll need to
wait until archives catches up to double check.

Re: automating CF submissions (was xlog location arithmetic)

Alvaro Herrera <alvherre@commandprompt.com> — 2012-01-19T19:04:07Z

Excerpts from Alex Shulgin's message of jue ene 19 15:41:54 -0300 2012:

> PS: yes, I could just copy message id from the sent mail in my MUA, but
> I like to make sure links I post aren't broke, so still I'll need to
> wait until archives catches up to double check.

I find this a bad excuse.  If you're a pgsql-hackers regular, then you
already know your posts are going to show up with the correct
message-id.  The links might be broken for the next 10 minutes, but
links that stay broken for a longer period than that should be rare.
Surely you don't change your MUA once a month or anything.

I know I don't waste time waiting for my posts to show up in the
archives before adding links to the CF app.

-- 
Álvaro Herrera <alvherre@commandprompt.com>
The PostgreSQL Company - Command Prompt, Inc.
PostgreSQL Replication, Consulting, Custom Development, 24x7 support

Re: automating CF submissions (was xlog location arithmetic)

Andrew Dunstan <andrew@dunslane.net> — 2012-01-19T19:10:00Z


On 01/19/2012 01:41 PM, Alex Shulgin wrote:
>
> With the proposed approach it would only take me to include
>
> @fest comment "Patch applies cleanly"
>
> and possibly
>
> @fest status Needs Review
>
> to update the patch status and that'd be it.
>

It will be easy if you get it right. My point was that it's way too easy 
to get it wrong.

cheers

andrew

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2012-01-21T16:13:41Z

On 23-12-2011 12:05, Tom Lane wrote:
> I too think a datatype is overkill, if we're only planning on providing
> one function.  Just emit the values as numeric and have done.
> 
Here it is. Output changed to numeric. While the output was int8 I could use
pg_size_pretty but now I couldn't. I attached another patch that implements
pg_size_pretty(numeric).


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: automating CF submissions (was xlog location arithmetic)

Dimitri Fontaine <dimitri@2ndquadrant.fr> — 2012-01-22T18:53:39Z

Alvaro Herrera <alvherre@commandprompt.com> writes:
> The problem is that this one doesn't have the
> Content-Disposition: attachment
> line in the MIME header.  I don't know what we can do about it.

It's sent with an inline attachment AFAICT, some MA will make it easy
to process the attachment and some others will just make the content
appear within the mail. It seems the vast majority falls into the
unhelpful second category.

Regards,
-- 
Dimitri Fontaine
http://2ndQuadrant.fr     PostgreSQL : Expertise, Formation et Support

Re: automating CF submissions (was xlog location arithmetic)

Dimitri Fontaine <dimitri@2ndquadrant.fr> — 2012-01-22T18:54:30Z

Alex Shulgin <ash@commandprompt.com> writes:
> Another idea: introduce some simple tag system in mails sent to -hackers
> to be treated specially, e.g:
[...]
> How does that sound?

Very much like what debbugs does already.

Regards,
-- 
Dimitri Fontaine
http://2ndQuadrant.fr     PostgreSQL : Expertise, Formation et Support

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-01-26T09:19:28Z

On Sun, Jan 22, 2012 at 1:13 AM, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 23-12-2011 12:05, Tom Lane wrote:
>> I too think a datatype is overkill, if we're only planning on providing
>> one function.  Just emit the values as numeric and have done.
>>
> Here it is. Output changed to numeric.

Thanks!

When I compiled the source with xlogdiff.patch, I got the following warnings.

xlogfuncs.c:511:2: warning: format '%lX' expects argument of type
'long unsigned int *', but argument 3 has type 'uint64 *' [-Wformat]
xlogfuncs.c:511:2: warning: format '%lX' expects argument of type
'long unsigned int *', but argument 4 has type 'uint64 *' [-Wformat]
xlogfuncs.c:515:2: warning: format '%lX' expects argument of type
'long unsigned int *', but argument 3 has type 'uint64 *' [-Wformat]
xlogfuncs.c:515:2: warning: format '%lX' expects argument of type
'long unsigned int *', but argument 4 has type 'uint64 *' [-Wformat]
xlogfuncs.c:524:3: warning: format '%lX' expects argument of type
'long unsigned int', but argument 2 has type 'uint64' [-Wformat]
xlogfuncs.c:528:3: warning: format '%lX' expects argument of type
'long unsigned int', but argument 2 has type 'uint64' [-Wformat]

When I tested the patch, I got the following error:

postgres=# SELECT pg_current_xlog_location();
 pg_current_xlog_location
--------------------------
 0/2000074
(1 row)

postgres=# SELECT pg_xlog_location_diff('0/2000074', '0/2000074');
ERROR:  xrecoff "2000074" is out of valid range, 0..A4A534C

In func.sgml
   <para>
    The functions shown in <xref
    linkend="functions-admin-backup-table"> assist in making on-line backups.
    These functions cannot be executed during recovery.
   </para>

Since pg_xlog_location_diff() can be executed during recovery,
the above needs to be updated.

> While the output was int8 I could use
> pg_size_pretty but now I couldn't. I attached another patch that implements
> pg_size_pretty(numeric).

I agree it's necessary.

 * Note: every entry in pg_proc.h is expected to have a DESCR() comment,
 * except for functions that implement pg_operator.h operators and don't
 * have a good reason to be called directly rather than via the operator.

According to the above source code comment in pg_proc.h, ISTM
pg_size_pretty() for numeric also needs to have its own DESCR().

+			buf = DatumGetCString(DirectFunctionCall1(numeric_out,
NumericGetDatum(size)));
+			result = strcat(buf, " kB");

According to "man strcat", the dest string must have enough space for
the result.
"buf" has enough space?

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2012-02-07T17:29:16Z

On 26-01-2012 06:19, Fujii Masao wrote:

Thanks for your review. Comments below.

> When I compiled the source with xlogdiff.patch, I got the following warnings.
> 
> xlogfuncs.c:511:2: warning: format '%lX' expects argument of type
> 'long unsigned int *', but argument 3 has type 'uint64 *' [-Wformat]
> 
What is your compiler? I'm using gcc 4.6.2. I refactored the patch so I'm now
using XLogRecPtr and %X.

> postgres=# SELECT pg_xlog_location_diff('0/2000074', '0/2000074');
> ERROR:  xrecoff "2000074" is out of valid range, 0..A4A534C
> 
Ugh? I can't reproduce that. It seems to be related to long int used by the
prior version.

> Since pg_xlog_location_diff() can be executed during recovery,
> the above needs to be updated.
> 
Fixed.

>> While the output was int8 I could use
>> pg_size_pretty but now I couldn't. I attached another patch that implements
>> pg_size_pretty(numeric).
> 
I realized that it collides with the pg_size_pretty(int8) if we don't specify
a type. Hence, I decided to drop the pg_size_pretty(int8) in favor of
pg_size_pretty(numeric). It is slower than the former but it is not a
performance critical function.

> According to the above source code comment in pg_proc.h, ISTM
> pg_size_pretty() for numeric also needs to have its own DESCR().
> 
Fixed.

> According to "man strcat", the dest string must have enough space for
> the result.
> "buf" has enough space?
> 
Ops. Fixed.


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-02-08T12:35:15Z

On Wed, Feb 8, 2012 at 2:29 AM, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 26-01-2012 06:19, Fujii Masao wrote:
>
> Thanks for your review. Comments below.
>
>> When I compiled the source with xlogdiff.patch, I got the following warnings.
>>
>> xlogfuncs.c:511:2: warning: format '%lX' expects argument of type
>> 'long unsigned int *', but argument 3 has type 'uint64 *' [-Wformat]
>>
> What is your compiler? I'm using gcc 4.6.2. I refactored the patch so I'm now
> using XLogRecPtr and %X.

gcc version 4.6.1 (Ubuntu/Linaro 4.6.1-9ubuntu3)

$ uname -a
Linux hermes 3.0.0-15-generic #26-Ubuntu SMP Fri Jan 20 15:59:53 UTC
2012 i686 i686 i386 GNU/Linux

In the updated version of the patch, I got no warnings at the compile time.
But initdb failed because the OID which you assigned to pg_xlog_location_diff()
has already been used for other function. So you need to update pg_proc.h.

>> postgres=# SELECT pg_xlog_location_diff('0/2000074', '0/2000074');
>> ERROR:  xrecoff "2000074" is out of valid range, 0..A4A534C
>>
> Ugh? I can't reproduce that. It seems to be related to long int used by the
> prior version.

Maybe.

But another problem happened. When I changed pg_proc.h so that the unused
OID was assigned to pg_xlog_location_diff(), and executed the above again,
I encountered the segmentation fault:

LOG:  server process (PID 14384) was terminated by signal 11: Segmentation fault
DETAIL:  Failed process was running: SELECT
pg_xlog_location_diff('0/2000074', '0/2000074');
LOG:  terminating any other active server processes

ISTM that the cause is that int8_numeric() is executed for uint32 value. We
should use int4_numeric(), instead?

>>> While the output was int8 I could use
>>> pg_size_pretty but now I couldn't. I attached another patch that implements
>>> pg_size_pretty(numeric).
>>
> I realized that it collides with the pg_size_pretty(int8) if we don't specify
> a type. Hence, I decided to drop the pg_size_pretty(int8) in favor of
> pg_size_pretty(numeric). It is slower than the former but it is not a
> performance critical function.

I'm OK with this.

-DATA(insert OID = 2288 ( pg_size_pretty			PGNSP PGUID 12 1 0 0 0 f f
f t f v 1 0 25 "20" _null_ _null_ _null_ _null_ pg_size_pretty _null_
_null_ _null_ ));
-DESCR("convert a long int to a human readable text using size units");
+DATA(insert OID = 3158 ( pg_size_pretty			PGNSP PGUID 12 1 0 0 0 f f
f t f v 1 0 25 "1700" _null_ _null_ _null_ _null_ pg_size_pretty
_null_ _null_ _null_ ));
+DESCR("convert a numeric to a human readable text using size units");

Why OID needs to be reassigned?

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2012-02-09T22:00:30Z

On 08-02-2012 09:35, Fujii Masao wrote:

Fujii, new patch attached. Thanks for your tests.

> But another problem happened. When I changed pg_proc.h so that the unused
> OID was assigned to pg_xlog_location_diff(), and executed the above again,
> I encountered the segmentation fault:
> 
I reproduced the problems in my old 32-bit laptop. I fixed it casting to
int64. I also updated the duplicated OID.

> Why OID needs to be reassigned?
> 
There isn't a compelling reason. It is just a way to say: "hey, it is another
function with the same old name".

I'll not attach another version for pg_size_pretty because it is a matter of
updating a duplicated OID.


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-02-10T08:32:46Z

On Fri, Feb 10, 2012 at 7:00 AM, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 08-02-2012 09:35, Fujii Masao wrote:
>
> Fujii, new patch attached. Thanks for your tests.

Thanks for the new patch!

>> But another problem happened. When I changed pg_proc.h so that the unused
>> OID was assigned to pg_xlog_location_diff(), and executed the above again,
>> I encountered the segmentation fault:
>>
> I reproduced the problems in my old 32-bit laptop. I fixed it casting to
> int64. I also updated the duplicated OID.

Yep, in the updated patch, I could confirm that the function works fine without
any error in my machine. The patch looks fine to me except the following minor
comments:

In the document, it's better to explain clearly that the function subtracts the
second argument from the first.

-    These functions cannot be executed during recovery.
+	These functions cannot be executed during recovery (except
+	<function>pg_xlog_location_diff</function>).

+	<function>pg_xlog_location_diff</> calculates the difference in bytes
+	between two transaction log locations. It can be used with
+	<structname>pg_stat_replication</structname> or some functions shown in
+	<xref linkend="functions-admin-backup-table"> to get the replication lag.

Very minor comment: you should use spaces rather than a tab to indent each line.

>> Why OID needs to be reassigned?
>>
> There isn't a compelling reason. It is just a way to say: "hey, it is another
> function with the same old name".
>
> I'll not attach another version for pg_size_pretty because it is a matter of
> updating a duplicated OID.

Okay, I reviewed the previous patch again. That looks fine to me.

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2012-02-25T12:23:25Z

On Fri, Feb 10, 2012 at 09:32, Fujii Masao <masao.fujii@gmail.com> wrote:
> On Fri, Feb 10, 2012 at 7:00 AM, Euler Taveira de Oliveira
> <euler@timbira.com> wrote:
>> On 08-02-2012 09:35, Fujii Masao wrote:
>>
>> Fujii, new patch attached. Thanks for your tests.
>
> Thanks for the new patch!
>
>>> But another problem happened. When I changed pg_proc.h so that the unused
>>> OID was assigned to pg_xlog_location_diff(), and executed the above again,
>>> I encountered the segmentation fault:
>>>
>> I reproduced the problems in my old 32-bit laptop. I fixed it casting to
>> int64. I also updated the duplicated OID.
>
> Yep, in the updated patch, I could confirm that the function works fine without
> any error in my machine. The patch looks fine to me except the following minor
> comments:

I started working on this one to commit it, and came up with a few things more.

Do we even *need* the validate_xlog_location() function? If we just
remove those calls, won't we still catch all the incorrectly formatted
ones in the errors of the sscanf() calls? Or am I too deep into
weekend-mode and missing something obvious?

I've also removed tabs in the documentation, fixed the merge confllict
in pg_proc.h that happened during the wait, and fixed some indentation
(updated patch with these changes attached).

But I'm going to hold off committing it until someone confirms I'm not
caught too deeply in weekend-mode and am missing something obvious in
the comment above about validate_xlog_location.

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Euler Taveira de Oliveira <euler@timbira.com> — 2012-02-25T23:53:19Z

On 25-02-2012 09:23, Magnus Hagander wrote:
> Do we even *need* the validate_xlog_location() function? If we just
> remove those calls, won't we still catch all the incorrectly formatted
> ones in the errors of the sscanf() calls? Or am I too deep into
> weekend-mode and missing something obvious?
> 
sscanf() is too fragile for input sanity check. Try
pg_xlog_location_diff('12/3', '-10/0'), for example. I won't object removing
that function if you protect xlog location input from silly users.


-- 
   Euler Taveira de Oliveira - Timbira       http://www.timbira.com.br/
   PostgreSQL: Consultoria, Desenvolvimento, Suporte 24x7 e Treinamento

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-02-28T06:21:12Z

On Sun, Feb 26, 2012 at 8:53 AM, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 25-02-2012 09:23, Magnus Hagander wrote:
>> Do we even *need* the validate_xlog_location() function? If we just
>> remove those calls, won't we still catch all the incorrectly formatted
>> ones in the errors of the sscanf() calls? Or am I too deep into
>> weekend-mode and missing something obvious?
>>
> sscanf() is too fragile for input sanity check. Try
> pg_xlog_location_diff('12/3', '-10/0'), for example. I won't object removing
> that function if you protect xlog location input from silly users.

After this patch will have been committed, it would be better to change
pg_xlogfile_name() and pg_xlogfile_name_offset() so that they use
the validate_xlog_location() function to validate the input.

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2012-03-04T11:26:36Z

On Sun, Feb 26, 2012 at 00:53, Euler Taveira de Oliveira
<euler@timbira.com> wrote:
> On 25-02-2012 09:23, Magnus Hagander wrote:
>> Do we even *need* the validate_xlog_location() function? If we just
>> remove those calls, won't we still catch all the incorrectly formatted
>> ones in the errors of the sscanf() calls? Or am I too deep into
>> weekend-mode and missing something obvious?
>>
> sscanf() is too fragile for input sanity check. Try
> pg_xlog_location_diff('12/3', '-10/0'), for example. I won't object removing
> that function if you protect xlog location input from silly users.

Ah, good point. No, that's the reason I was missing :-)

Patch applied, thanks!

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2012-03-04T11:26:56Z

On Tue, Feb 28, 2012 at 07:21, Fujii Masao <masao.fujii@gmail.com> wrote:
> On Sun, Feb 26, 2012 at 8:53 AM, Euler Taveira de Oliveira
> <euler@timbira.com> wrote:
>> On 25-02-2012 09:23, Magnus Hagander wrote:
>>> Do we even *need* the validate_xlog_location() function? If we just
>>> remove those calls, won't we still catch all the incorrectly formatted
>>> ones in the errors of the sscanf() calls? Or am I too deep into
>>> weekend-mode and missing something obvious?
>>>
>> sscanf() is too fragile for input sanity check. Try
>> pg_xlog_location_diff('12/3', '-10/0'), for example. I won't object removing
>> that function if you protect xlog location input from silly users.
>
> After this patch will have been committed, it would be better to change
> pg_xlogfile_name() and pg_xlogfile_name_offset() so that they use
> the validate_xlog_location() function to validate the input.

And I've done this part as well.

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-03-09T12:11:13Z

On Sun, Mar 4, 2012 at 8:26 PM, Magnus Hagander <magnus@hagander.net> wrote:
> On Sun, Feb 26, 2012 at 00:53, Euler Taveira de Oliveira
> <euler@timbira.com> wrote:
>> On 25-02-2012 09:23, Magnus Hagander wrote:
>>> Do we even *need* the validate_xlog_location() function? If we just
>>> remove those calls, won't we still catch all the incorrectly formatted
>>> ones in the errors of the sscanf() calls? Or am I too deep into
>>> weekend-mode and missing something obvious?
>>>
>> sscanf() is too fragile for input sanity check. Try
>> pg_xlog_location_diff('12/3', '-10/0'), for example. I won't object removing
>> that function if you protect xlog location input from silly users.
>
> Ah, good point. No, that's the reason I was missing :-)
>
> Patch applied, thanks!

Thanks for committing the patch!

Euler proposed one more patch upthread, which replaces pg_size_pretty(bigint)
with pg_size_pretty(numeric) so that pg_size_pretty(pg_xlog_location_diff())
succeeds. It's also worth committing this patch?
http://archives.postgresql.org/message-id/4F315F6C.8030700@timbira.com

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T14:37:35Z

Fujii Masao <masao.fujii@gmail.com> writes:
> Euler proposed one more patch upthread, which replaces pg_size_pretty(bigint)
> with pg_size_pretty(numeric) so that pg_size_pretty(pg_xlog_location_diff())
> succeeds. It's also worth committing this patch?

Why would it be useful to use pg_size_pretty on xlog locations?
-1 because of the large expense of bigint->numeric->whatever conversion
that would be added to existing uses.

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-03-09T14:44:08Z

On Fri, Mar 9, 2012 at 9:37 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Fujii Masao <masao.fujii@gmail.com> writes:
>> Euler proposed one more patch upthread, which replaces pg_size_pretty(bigint)
>> with pg_size_pretty(numeric) so that pg_size_pretty(pg_xlog_location_diff())
>> succeeds. It's also worth committing this patch?
>
> Why would it be useful to use pg_size_pretty on xlog locations?
> -1 because of the large expense of bigint->numeric->whatever conversion
> that would be added to existing uses.

The point is that it would be useful to use it on the difference
between two xlog locations, but that is a numeric value, not int8,
because of signedness issues.

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

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T14:55:09Z

I wrote:
> Fujii Masao <masao.fujii@gmail.com> writes:
>> Euler proposed one more patch upthread, which replaces pg_size_pretty(bigint)
>> with pg_size_pretty(numeric) so that pg_size_pretty(pg_xlog_location_diff())
>> succeeds. It's also worth committing this patch?

> Why would it be useful to use pg_size_pretty on xlog locations?
> -1 because of the large expense of bigint->numeric->whatever conversion
> that would be added to existing uses.

Actually ... now that I look at it, isn't it completely bogus to be
using numeric for the result of pg_xlog_location_diff?  There's no way
for the difference of two xlog locations to be anywhere near as wide as
64 bits.  That'd only be possible if XLogFileSize exceeded 1GB, which we
don't let it get anywhere near.

			regards, tom lane

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T15:00:55Z

Robert Haas <robertmhaas@gmail.com> writes:
> On Fri, Mar 9, 2012 at 9:37 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Why would it be useful to use pg_size_pretty on xlog locations?

> The point is that it would be useful to use it on the difference
> between two xlog locations,

Um, that is exactly the claim I was questioning.  Why is that useful?

> but that is a numeric value, not int8, because of signedness issues.

See my followup --- this statement appears factually incorrect,
whatever you may feel about the usefulness issue.

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-03-09T15:31:26Z

On Fri, Mar 9, 2012 at 9:55 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> I wrote:
>> Fujii Masao <masao.fujii@gmail.com> writes:
>>> Euler proposed one more patch upthread, which replaces pg_size_pretty(bigint)
>>> with pg_size_pretty(numeric) so that pg_size_pretty(pg_xlog_location_diff())
>>> succeeds. It's also worth committing this patch?
>
>> Why would it be useful to use pg_size_pretty on xlog locations?
>> -1 because of the large expense of bigint->numeric->whatever conversion
>> that would be added to existing uses.
>
> Actually ... now that I look at it, isn't it completely bogus to be
> using numeric for the result of pg_xlog_location_diff?  There's no way
> for the difference of two xlog locations to be anywhere near as wide as
> 64 bits.  That'd only be possible if XLogFileSize exceeded 1GB, which we
> don't let it get anywhere near.

rhaas=# select pg_xlog_location_diff('ffffffff/0', '0/0');
 pg_xlog_location_diff
-----------------------
  18374686475393433600
(1 row)

rhaas=# select pg_xlog_location_diff('ffffffff/0', '0/0')::int8;
ERROR:  bigint out of range
STATEMENT:  select pg_xlog_location_diff('ffffffff/0', '0/0')::int8;

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

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T15:37:42Z

Robert Haas <robertmhaas@gmail.com> writes:
> On Fri, Mar 9, 2012 at 9:55 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Actually ... now that I look at it, isn't it completely bogus to be
>> using numeric for the result of pg_xlog_location_diff?

> rhaas=# select pg_xlog_location_diff('ffffffff/0', '0/0')::int8;
> ERROR:  bigint out of range

Oh ... I see my mistake.  I was looking at this:

	/*
	 * result = XLogFileSize * (xlogid1 - xlogid2) + xrecoff1 - xrecoff2
	 */

and confusing XLogFileSize with XLogSegSize.  Not the best choice of
names.

			regards, tom lane

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2012-03-09T17:13:26Z

On Fri, Mar 9, 2012 at 16:37, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Fri, Mar 9, 2012 at 9:55 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Actually ... now that I look at it, isn't it completely bogus to be
>>> using numeric for the result of pg_xlog_location_diff?
>
>> rhaas=# select pg_xlog_location_diff('ffffffff/0', '0/0')::int8;
>> ERROR:  bigint out of range
>
> Oh ... I see my mistake.  I was looking at this:
>
>        /*
>         * result = XLogFileSize * (xlogid1 - xlogid2) + xrecoff1 - xrecoff2
>         */
>
> and confusing XLogFileSize with XLogSegSize.  Not the best choice of
> names.

Yeah, the use of XLogFile to mean something other than, well a file in
the xlog, is greatly annoying.. I guess we could change it, but it
goes pretty deep in the system so it's not a small change...

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2012-03-09T17:14:00Z

On Fri, Mar 9, 2012 at 15:37, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Fujii Masao <masao.fujii@gmail.com> writes:
>> Euler proposed one more patch upthread, which replaces pg_size_pretty(bigint)
>> with pg_size_pretty(numeric) so that pg_size_pretty(pg_xlog_location_diff())
>> succeeds. It's also worth committing this patch?
>
> Why would it be useful to use pg_size_pretty on xlog locations?
> -1 because of the large expense of bigint->numeric->whatever conversion
> that would be added to existing uses.

Given the expense, perhaps we need to different (overloaded) functions instead?

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T17:18:24Z

Magnus Hagander <magnus@hagander.net> writes:
> On Fri, Mar 9, 2012 at 15:37, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Why would it be useful to use pg_size_pretty on xlog locations?
>> -1 because of the large expense of bigint->numeric->whatever conversion
>> that would be added to existing uses.

> Given the expense, perhaps we need to different (overloaded) functions instead?

That would be a workable solution, but I continue to not believe that
this is useful enough to be worth the trouble.

			regards, tom lane

Re: xlog location arithmetic

Magnus Hagander <magnus@hagander.net> — 2012-03-09T17:38:28Z

On Fri, Mar 9, 2012 at 18:18, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Magnus Hagander <magnus@hagander.net> writes:
>> On Fri, Mar 9, 2012 at 15:37, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Why would it be useful to use pg_size_pretty on xlog locations?
>>> -1 because of the large expense of bigint->numeric->whatever conversion
>>> that would be added to existing uses.
>
>> Given the expense, perhaps we need to different (overloaded) functions instead?
>
> That would be a workable solution, but I continue to not believe that
> this is useful enough to be worth the trouble.

There's certainly some use to being able to prettify it. Wouldn't a
pg_size_pretty(numeric) also be useful if you want to pg_size_() a
sum() of something? Used on files it doesn't make too much sense,
given how big those files have to be, but it can be used on other
things as well...

I can see a usecase for having a pg_size_pretty(numeric) as an option.
Not necessarily a very big one, but a >0 one.

-- 
 Magnus Hagander
 Me: http://www.hagander.net/
 Work: http://www.redpill-linpro.com/

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-03-09T18:23:38Z

On Fri, Mar 9, 2012 at 12:38 PM, Magnus Hagander <magnus@hagander.net> wrote:
> On Fri, Mar 9, 2012 at 18:18, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Magnus Hagander <magnus@hagander.net> writes:
>>> On Fri, Mar 9, 2012 at 15:37, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>>> Why would it be useful to use pg_size_pretty on xlog locations?
>>>> -1 because of the large expense of bigint->numeric->whatever conversion
>>>> that would be added to existing uses.
>>
>>> Given the expense, perhaps we need to different (overloaded) functions instead?
>>
>> That would be a workable solution, but I continue to not believe that
>> this is useful enough to be worth the trouble.
>
> There's certainly some use to being able to prettify it. Wouldn't a
> pg_size_pretty(numeric) also be useful if you want to pg_size_() a
> sum() of something? Used on files it doesn't make too much sense,
> given how big those files have to be, but it can be used on other
> things as well...
>
> I can see a usecase for having a pg_size_pretty(numeric) as an option.
> Not necessarily a very big one, but a >0 one.

+1.

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

Re: xlog location arithmetic

Peter Eisentraut <peter_e@gmx.net> — 2012-03-09T19:23:28Z

On fre, 2012-03-09 at 18:13 +0100, Magnus Hagander wrote:
> > and confusing XLogFileSize with XLogSegSize.  Not the best choice of
> > names.
> 
> Yeah, the use of XLogFile to mean something other than, well a file in
> the xlog, is greatly annoying.. I guess we could change it, but it
> goes pretty deep in the system so it's not a small change...
> 
The whole thing was built around the lack of 64 bit integers.  If we bit
the bullet and changed the whole thing to be just a single 64-bit
counter, we could probably delete thousands of lines of code.

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T19:26:44Z

Peter Eisentraut <peter_e@gmx.net> writes:
>> Yeah, the use of XLogFile to mean something other than, well a file in
>> the xlog, is greatly annoying.. I guess we could change it, but it
>> goes pretty deep in the system so it's not a small change...

> The whole thing was built around the lack of 64 bit integers.  If we bit
> the bullet and changed the whole thing to be just a single 64-bit
> counter, we could probably delete thousands of lines of code.

Hm.  I think "thousands" is an overestimate, but yeah the logic could be
greatly simplified.  However, I'm not sure we could avoid breaking the
existing naming convention for WAL files.  How much do we care about
that?

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-03-09T19:34:04Z

On Fri, Mar 9, 2012 at 2:26 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Peter Eisentraut <peter_e@gmx.net> writes:
>>> Yeah, the use of XLogFile to mean something other than, well a file in
>>> the xlog, is greatly annoying.. I guess we could change it, but it
>>> goes pretty deep in the system so it's not a small change...
>
>> The whole thing was built around the lack of 64 bit integers.  If we bit
>> the bullet and changed the whole thing to be just a single 64-bit
>> counter, we could probably delete thousands of lines of code.
>
> Hm.  I think "thousands" is an overestimate, but yeah the logic could be
> greatly simplified.  However, I'm not sure we could avoid breaking the
> existing naming convention for WAL files.  How much do we care about
> that?

Probably not very much, since WAL files aren't portable across major
versions anyway.  But I don't see why you couldn't keep the naming
convention - there's nothing to prevent you from converting a 64-bit
integer back into two 32-bit integers if and where needed.

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

Re: xlog location arithmetic

Kevin Grittner <kevin.grittner@wicourts.gov> — 2012-03-09T19:37:43Z

Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Peter Eisentraut <peter_e@gmx.net> writes:

>> The whole thing was built around the lack of 64 bit integers.  If
>> we bit the bullet and changed the whole thing to be just a single
>> 64-bit counter, we could probably delete thousands of lines of
>> code.
> 
> Hm.  I think "thousands" is an overestimate, but yeah the logic
> could be greatly simplified.  However, I'm not sure we could avoid
> breaking the existing naming convention for WAL files.  How much
> do we care about that?

We have a few scripts in our backup area that are based around the
current WAL file naming convention, so there would be some impact;
but I have to believe it would be pretty minor.  Most of the pain
would be related to the need to support both naming conventions for
some transition period.  If it simplifies the WAL-related logic, it
seems well worth it to me.  We just have to know it's coming and be
clear on what the new naming rules are.

-Kevin

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-03-09T19:39:26Z

On Fri, Mar 9, 2012 at 2:34 PM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Fri, Mar 9, 2012 at 2:26 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> Peter Eisentraut <peter_e@gmx.net> writes:
>>>> Yeah, the use of XLogFile to mean something other than, well a file in
>>>> the xlog, is greatly annoying.. I guess we could change it, but it
>>>> goes pretty deep in the system so it's not a small change...
>>
>>> The whole thing was built around the lack of 64 bit integers.  If we bit
>>> the bullet and changed the whole thing to be just a single 64-bit
>>> counter, we could probably delete thousands of lines of code.
>>
>> Hm.  I think "thousands" is an overestimate, but yeah the logic could be
>> greatly simplified.  However, I'm not sure we could avoid breaking the
>> existing naming convention for WAL files.  How much do we care about
>> that?
>
> Probably not very much, since WAL files aren't portable across major
> versions anyway.  But I don't see why you couldn't keep the naming
> convention - there's nothing to prevent you from converting a 64-bit
> integer back into two 32-bit integers if and where needed.

On further reflection, this seems likely to break quite a few
third-party tools.  Maybe it'd be worth it anyway, but it definitely
seems like it would be worth going to at least some minor trouble to
avoid it.

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

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-09T20:04:23Z

Robert Haas <robertmhaas@gmail.com> writes:
> On Fri, Mar 9, 2012 at 2:34 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>> On Fri, Mar 9, 2012 at 2:26 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Hm. I think "thousands" is an overestimate, but yeah the logic could be
>>> greatly simplified. However, I'm not sure we could avoid breaking the
>>> existing naming convention for WAL files. How much do we care about
>>> that?

>> Probably not very much, since WAL files aren't portable across major
>> versions anyway. But I don't see why you couldn't keep the naming
>> convention - there's nothing to prevent you from converting a 64-bit
>> integer back into two 32-bit integers if and where needed.

> On further reflection, this seems likely to break quite a few
> third-party tools.  Maybe it'd be worth it anyway, but it definitely
> seems like it would be worth going to at least some minor trouble to
> avoid it.

The main actual simplification would be in getting rid of the "hole"
at the end of each 4GB worth of WAL, cf this bit in xlog_internal.h:

/*
 * We break each logical log file (xlogid value) into segment files of the
 * size indicated by XLOG_SEG_SIZE.  One possible segment at the end of each
 * log file is wasted, to ensure that we don't have problems representing
 * last-byte-position-plus-1.
 */
#define XLogSegSize		((uint32) XLOG_SEG_SIZE)
#define XLogSegsPerFile (((uint32) 0xffffffff) / XLogSegSize)
#define XLogFileSize	(XLogSegsPerFile * XLogSegSize)

If we can't get rid of that and have a continuous 64-bit WAL address
space then it's unlikely we can actually simplify any logic.

Now, doing that doesn't break the naming convention exactly; what it
changes is that there will be WAL files numbered xxxFFFF (for some
number of trailing-1-bits I'm too lazy to work out at the moment) where
before there were not.  So the question really is how much external code
there is that is aware of that specific noncontiguous numbering behavior
and would be broken if things stopped being that way.

			regards, tom lane

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-03-09T20:12:20Z

On Fri, Mar 9, 2012 at 3:04 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Fri, Mar 9, 2012 at 2:34 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>>> On Fri, Mar 9, 2012 at 2:26 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>>> Hm.  I think "thousands" is an overestimate, but yeah the logic could be
>>>> greatly simplified.  However, I'm not sure we could avoid breaking the
>>>> existing naming convention for WAL files.  How much do we care about
>>>> that?
>
>>> Probably not very much, since WAL files aren't portable across major
>>> versions anyway.  But I don't see why you couldn't keep the naming
>>> convention - there's nothing to prevent you from converting a 64-bit
>>> integer back into two 32-bit integers if and where needed.
>
>> On further reflection, this seems likely to break quite a few
>> third-party tools.  Maybe it'd be worth it anyway, but it definitely
>> seems like it would be worth going to at least some minor trouble to
>> avoid it.
>
> The main actual simplification would be in getting rid of the "hole"
> at the end of each 4GB worth of WAL, cf this bit in xlog_internal.h:
>
> /*
>  * We break each logical log file (xlogid value) into segment files of the
>  * size indicated by XLOG_SEG_SIZE.  One possible segment at the end of each
>  * log file is wasted, to ensure that we don't have problems representing
>  * last-byte-position-plus-1.
>  */
> #define XLogSegSize             ((uint32) XLOG_SEG_SIZE)
> #define XLogSegsPerFile (((uint32) 0xffffffff) / XLogSegSize)
> #define XLogFileSize    (XLogSegsPerFile * XLogSegSize)
>
> If we can't get rid of that and have a continuous 64-bit WAL address
> space then it's unlikely we can actually simplify any logic.
>
> Now, doing that doesn't break the naming convention exactly; what it
> changes is that there will be WAL files numbered xxxFFFF (for some
> number of trailing-1-bits I'm too lazy to work out at the moment) where
> before there were not.  So the question really is how much external code
> there is that is aware of that specific noncontiguous numbering behavior
> and would be broken if things stopped being that way.

I would expect that most things would NOT know about that particular
foible, and just be matching pathnames on an RE, which should be fine.

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

Re: xlog location arithmetic

Bruce Momjian <bruce@momjian.us> — 2012-03-13T00:47:24Z

On Fri, Mar 09, 2012 at 03:04:23PM -0500, Tom Lane wrote:
> The main actual simplification would be in getting rid of the "hole"
> at the end of each 4GB worth of WAL, cf this bit in xlog_internal.h:
> 
> /*
>  * We break each logical log file (xlogid value) into segment files of the
>  * size indicated by XLOG_SEG_SIZE.  One possible segment at the end of each
>  * log file is wasted, to ensure that we don't have problems representing
>  * last-byte-position-plus-1.
>  */
> #define XLogSegSize		((uint32) XLOG_SEG_SIZE)
> #define XLogSegsPerFile (((uint32) 0xffffffff) / XLogSegSize)
> #define XLogFileSize	(XLogSegsPerFile * XLogSegSize)
> 
> If we can't get rid of that and have a continuous 64-bit WAL address
> space then it's unlikely we can actually simplify any logic.
> 
> Now, doing that doesn't break the naming convention exactly; what it
> changes is that there will be WAL files numbered xxxFFFF (for some
> number of trailing-1-bits I'm too lazy to work out at the moment) where
> before there were not.  So the question really is how much external code
> there is that is aware of that specific noncontiguous numbering behavior
> and would be broken if things stopped being that way.

Our current WAL naming is hopelessly arcane, and we would certainly be
benfitting users to simplify it.  Is this a TODO?

-- 
  Bruce Momjian  <bruce@momjian.us>        http://momjian.us
  EnterpriseDB                             http://enterprisedb.com

  + It's impossible for everything to be true. +

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-03-13T02:34:10Z

On Sat, Mar 10, 2012 at 3:23 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Fri, Mar 9, 2012 at 12:38 PM, Magnus Hagander <magnus@hagander.net> wrote:
>> On Fri, Mar 9, 2012 at 18:18, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> Magnus Hagander <magnus@hagander.net> writes:
>>>> On Fri, Mar 9, 2012 at 15:37, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>>>> Why would it be useful to use pg_size_pretty on xlog locations?
>>>>> -1 because of the large expense of bigint->numeric->whatever conversion
>>>>> that would be added to existing uses.
>>>
>>>> Given the expense, perhaps we need to different (overloaded) functions instead?

Agreed. Attached patch introduces the overloaded funtion
pg_size_pretty(numeric).

>>> That would be a workable solution, but I continue to not believe that
>>> this is useful enough to be worth the trouble.
>>
>> There's certainly some use to being able to prettify it. Wouldn't a
>> pg_size_pretty(numeric) also be useful if you want to pg_size_() a
>> sum() of something? Used on files it doesn't make too much sense,
>> given how big those files have to be, but it can be used on other
>> things as well...
>>
>> I can see a usecase for having a pg_size_pretty(numeric) as an option.
>> Not necessarily a very big one, but a >0 one.
>
> +1.

+1, too.

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-03-13T02:51:50Z

On Sat, Mar 10, 2012 at 5:04 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Robert Haas <robertmhaas@gmail.com> writes:
>> On Fri, Mar 9, 2012 at 2:34 PM, Robert Haas <robertmhaas@gmail.com> wrote:
>>> On Fri, Mar 9, 2012 at 2:26 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>>> Hm.  I think "thousands" is an overestimate, but yeah the logic could be
>>>> greatly simplified.  However, I'm not sure we could avoid breaking the
>>>> existing naming convention for WAL files.  How much do we care about
>>>> that?
>
>>> Probably not very much, since WAL files aren't portable across major
>>> versions anyway.  But I don't see why you couldn't keep the naming
>>> convention - there's nothing to prevent you from converting a 64-bit
>>> integer back into two 32-bit integers if and where needed.
>
>> On further reflection, this seems likely to break quite a few
>> third-party tools.  Maybe it'd be worth it anyway, but it definitely
>> seems like it would be worth going to at least some minor trouble to
>> avoid it.
>
> The main actual simplification would be in getting rid of the "hole"
> at the end of each 4GB worth of WAL, cf this bit in xlog_internal.h:
>
> /*
>  * We break each logical log file (xlogid value) into segment files of the
>  * size indicated by XLOG_SEG_SIZE.  One possible segment at the end of each
>  * log file is wasted, to ensure that we don't have problems representing
>  * last-byte-position-plus-1.
>  */
> #define XLogSegSize             ((uint32) XLOG_SEG_SIZE)
> #define XLogSegsPerFile (((uint32) 0xffffffff) / XLogSegSize)
> #define XLogFileSize    (XLogSegsPerFile * XLogSegSize)
>
> If we can't get rid of that and have a continuous 64-bit WAL address
> space then it's unlikely we can actually simplify any logic.
>
> Now, doing that doesn't break the naming convention exactly; what it
> changes is that there will be WAL files numbered xxxFFFF (for some
> number of trailing-1-bits I'm too lazy to work out at the moment) where
> before there were not.  So the question really is how much external code
> there is that is aware of that specific noncontiguous numbering behavior
> and would be broken if things stopped being that way.

A page header contains WAL location, so getting rid of "hole" seems to
break pg_upgrade. No? Unless pg_upgrade converts noncontinuous
location to continuous one, we still need to handle noncontinuous one
after upgrade.

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Tom Lane <tgl@sss.pgh.pa.us> — 2012-03-13T03:03:42Z

Fujii Masao <masao.fujii@gmail.com> writes:
> On Sat, Mar 10, 2012 at 5:04 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>> The main actual simplification would be in getting rid of the "hole"
>> at the end of each 4GB worth of WAL, cf this bit in xlog_internal.h:
>> If we can't get rid of that and have a continuous 64-bit WAL address
>> space then it's unlikely we can actually simplify any logic.
>> ...
>> Now, doing that doesn't break the naming convention exactly; what it
>> changes is that there will be WAL files numbered xxxFFFF (for some
>> number of trailing-1-bits I'm too lazy to work out at the moment) where
>> before there were not.  So the question really is how much external code
>> there is that is aware of that specific noncontiguous numbering behavior
>> and would be broken if things stopped being that way.

> A page header contains WAL location, so getting rid of "hole" seems to
> break pg_upgrade. No?

No, why would it do that?  The meaning and ordering of WAL addresses is
the same as before.  The only difference is that after the upgrade, the
system will stop skipping over 16MB of potentially usable WAL addresses
at the end of each subsequently-used 4GB of space.  The holes before
the switchover point are still holes, but that doesn't matter.

			regards, tom lane

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-03-13T03:36:45Z

On Tue, Mar 13, 2012 at 12:03 PM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Fujii Masao <masao.fujii@gmail.com> writes:
>> On Sat, Mar 10, 2012 at 5:04 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
>>> The main actual simplification would be in getting rid of the "hole"
>>> at the end of each 4GB worth of WAL, cf this bit in xlog_internal.h:
>>> If we can't get rid of that and have a continuous 64-bit WAL address
>>> space then it's unlikely we can actually simplify any logic.
>>> ...
>>> Now, doing that doesn't break the naming convention exactly; what it
>>> changes is that there will be WAL files numbered xxxFFFF (for some
>>> number of trailing-1-bits I'm too lazy to work out at the moment) where
>>> before there were not.  So the question really is how much external code
>>> there is that is aware of that specific noncontiguous numbering behavior
>>> and would be broken if things stopped being that way.
>
>> A page header contains WAL location, so getting rid of "hole" seems to
>> break pg_upgrade. No?
>
> No, why would it do that?  The meaning and ordering of WAL addresses is
> the same as before.  The only difference is that after the upgrade, the
> system will stop skipping over 16MB of potentially usable WAL addresses
> at the end of each subsequently-used 4GB of space.  The holes before
> the switchover point are still holes, but that doesn't matter.

Oh, I see. You're right.

Regards,

-- 
Fujii Masao
NIPPON TELEGRAPH AND TELEPHONE CORPORATION
NTT Open Source Software Center

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-04-13T20:30:16Z

On Mon, Mar 12, 2012 at 10:34 PM, Fujii Masao <masao.fujii@gmail.com> wrote:
>>> I can see a usecase for having a pg_size_pretty(numeric) as an option.
>>> Not necessarily a very big one, but a >0 one.
>>
>> +1.
>
> +1, too.

I did some beautification of this patch.  I think the attached version
is cleaner and easier to read.  Thoughts?

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

Re: xlog location arithmetic

Fujii Masao <masao.fujii@gmail.com> — 2012-04-14T11:25:42Z

On Sat, Apr 14, 2012 at 5:30 AM, Robert Haas <robertmhaas@gmail.com> wrote:
> On Mon, Mar 12, 2012 at 10:34 PM, Fujii Masao <masao.fujii@gmail.com> wrote:
>>>> I can see a usecase for having a pg_size_pretty(numeric) as an option.
>>>> Not necessarily a very big one, but a >0 one.
>>>
>>> +1.
>>
>> +1, too.
>
> I did some beautification of this patch.  I think the attached version
> is cleaner and easier to read.  Thoughts?

Looks good to me. Thanks for polishing the patch!

Regards,

-- 
Fujii Masao

Re: xlog location arithmetic

Robert Haas <robertmhaas@gmail.com> — 2012-04-14T12:08:12Z

On Sat, Apr 14, 2012 at 7:25 AM, Fujii Masao <masao.fujii@gmail.com> wrote:
> On Sat, Apr 14, 2012 at 5:30 AM, Robert Haas <robertmhaas@gmail.com> wrote:
>> On Mon, Mar 12, 2012 at 10:34 PM, Fujii Masao <masao.fujii@gmail.com> wrote:
>>>>> I can see a usecase for having a pg_size_pretty(numeric) as an option.
>>>>> Not necessarily a very big one, but a >0 one.
>>>>
>>>> +1.
>>>
>>> +1, too.
>>
>> I did some beautification of this patch.  I think the attached version
>> is cleaner and easier to read.  Thoughts?
>
> Looks good to me. Thanks for polishing the patch!

You're welcome.  Committed.

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