Обсуждение: Verbose output of pg_dump not show schema name

On Thu, Apr 17, 2014 at 11:41 AM, Fabrízio de Royes Mello
<fabriziomello@gmail.com> wrote:
> Hi all,
>
> There are some reason to verbose output of pg_dump don't show schema name?
>
> A output example of using "pg_dump -Fd -j8 -v"
Specifying a target directory with "-f" is better here...

> This database have a lot of different schemas with same structure and if I
> need do view the status of dump I don't know what schema the table are dump
> from.
Yes this may be helpful. The attached quick'n dirty patch implements it.
--
Michael

On Thu, Apr 17, 2014 at 11:29:03AM -0300, Fabrízio de Royes Mello wrote:
> > > This database have a lot of different schemas with same structure and if I
> > > need do view the status of dump I don't know what schema the table are dump
> > > from.
> > Yes this may be helpful. The attached quick'n dirty patch implements it.
> >
> 
> Very nice... thanks!!!
> 
> I add schema name do the following messages too:
> 
> pg_restore: processing data for table "public"."bar"
> pg_restore: processing data for table "public"."foo"
> pg_restore: processing data for table "s1"."bar"
> pg_restore: processing data for table "s1"."foo"
> pg_restore: processing data for table "s2"."bar"
> pg_restore: processing data for table "s2"."foo"
> pg_restore: processing data for table "s3"."bar"
> pg_restore: processing data for table "s3"."foo"

Can you get that to _conditionally_ double-quote the strings?  In fact,
maybe we don't even need the double-quotes.  How do we double-quote
other places?

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + Everyone has their own god. +

On Thu, Apr 17, 2014 at 12:07:39PM -0300, Fabrízio de Royes Mello wrote:
> > Can you get that to _conditionally_ double-quote the strings? 
> 
> Sorry, I didn't understand what you means? Your idea is to check if the
> namespace is available and then don't show the double-quote, is that?

The idea is that we only need quotes when there are odd characters in
the identifier.  We do that right now in some places, though I can't
find them in pg_dump.  I know psql does that, see quote_ident().

> > In fact,
> > maybe we don't even need the double-quotes.  How do we double-quote
> > other places?
> >
> 
> Checking that more deeply I found some other places that show the table name
> and all of them are double-quoted.

OK.

> Just the "dumping contents of table.." message isn't double-quoted:
> 
> $ grep 'table %s' src/bin/pg_dump/*.c
> src/bin/pg_dump/pg_dump.c:            write_msg(NULL, "dumping contents of
> table %s\n",
> 
> 
> So maybe we must double-quote of all string, i.e. "public.foo", including the
> missing bellow.

No, I think double-quoting each part is the correct way.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + Everyone has their own god. +

Bruce Momjian <bruce@momjian.us> writes:
> The idea is that we only need quotes when there are odd characters in
> the identifier.  We do that right now in some places, though I can't
> find them in pg_dump.  I know psql does that, see quote_ident().

I think our general style rule is that identifiers embedded in messages
are always double-quoted.  There's an exception for type names, but
not otherwise.  You're confusing the message case with printing SQL.
        regards, tom lane

On Thu, Apr 17, 2014 at 11:44:37AM -0400, Tom Lane wrote:
> Bruce Momjian <bruce@momjian.us> writes:
> > The idea is that we only need quotes when there are odd characters in
> > the identifier.  We do that right now in some places, though I can't
> > find them in pg_dump.  I know psql does that, see quote_ident().
> 
> I think our general style rule is that identifiers embedded in messages
> are always double-quoted.  There's an exception for type names, but
> not otherwise.  You're confusing the message case with printing SQL.

OK.  I was unclear if a status _display_ was a message like an error
message.

--  Bruce Momjian  <bruce@momjian.us>        http://momjian.us EnterpriseDB
http://enterprisedb.com
 + Everyone has their own god. +

On Fri, Apr 18, 2014 at 4:29 AM, Fabrízio de Royes Mello
<fabriziomello@gmail.com> wrote:
>
> On Thu, Apr 17, 2014 at 12:46 PM, Bruce Momjian <bruce@momjian.us> wrote:
>>
>> On Thu, Apr 17, 2014 at 11:44:37AM -0400, Tom Lane wrote:
>> > Bruce Momjian <bruce@momjian.us> writes:
>> > > The idea is that we only need quotes when there are odd characters in
>> > > the identifier.  We do that right now in some places, though I can't
>> > > find them in pg_dump.  I know psql does that, see quote_ident().
>> >
>> > I think our general style rule is that identifiers embedded in messages
>> > are always double-quoted.  There's an exception for type names, but
>> > not otherwise.  You're confusing the message case with printing SQL.
>>
>> OK.  I was unclear if a status _display_ was a message like an error
>> message.
>>
>
> The attached patch fix missing double-quoted in "dumping contents of
> table.." message and add schema name to other messages:
> - "reading indexes for table \"%s\".\"%s\"\n"
> - "reading foreign key constraints for table \"%s\".\"%s\"\n"
> - "reading triggers for table \"%s\".\"%s\"\n"
>
> - "finding the columns and types of table \"%s\".\"%s\"\n"
> - "finding default expressions of table \"%s\".\"%s\"\n"
> - "finding check constraints for table \"%s\".\"%s\"\n"
Cool additions. There may be a more elegant way to check if namespace
is NULL, but I couldn't come up with one myself. So patch may be fine.
--
Michael

<div dir="ltr"><br />On Thu, Apr 17, 2014 at 9:15 PM, Michael Paquier <<a
href="mailto:michael.paquier@gmail.com">michael.paquier@gmail.com</a>>wrote:<br />><br />> On Fri, Apr 18,
2014at 4:29 AM, Fabrízio de Royes Mello<br /> > <<a
href="mailto:fabriziomello@gmail.com">fabriziomello@gmail.com</a>>wrote:<br />> ><br />> > On Thu, Apr
17,2014 at 12:46 PM, Bruce Momjian <<a href="mailto:bruce@momjian.us">bruce@momjian.us</a>> wrote:<br /> >
>><br/>> >> On Thu, Apr 17, 2014 at 11:44:37AM -0400, Tom Lane wrote:<br />> >> > Bruce
Momjian<<a href="mailto:bruce@momjian.us">bruce@momjian.us</a>> writes:<br />> >> > > The idea is
thatwe only need quotes when there are odd characters in<br /> > >> > > the identifier.  We do that
rightnow in some places, though I can't<br />> >> > > find them in pg_dump.  I know psql does that, see
quote_ident().<br/>> >> ><br />> >> > I think our general style rule is that identifiers
embeddedin messages<br /> > >> > are always double-quoted.  There's an exception for type names, but<br
/>>>> > not otherwise.  You're confusing the message case with printing SQL.<br />> >><br />>
>>OK.  I was unclear if a status _display_ was a message like an error<br /> > >> message.<br />>
>><br/>> ><br />> > The attached patch fix missing double-quoted in "dumping contents of<br />>
>table.." message and add schema name to other messages:<br />> > - "reading indexes for table
\"%s\".\"%s\"\n"<br/> > > - "reading foreign key constraints for table \"%s\".\"%s\"\n"<br />> > - "reading
triggersfor table \"%s\".\"%s\"\n"<br />> ><br />> > - "finding the columns and types of table
\"%s\".\"%s\"\n"<br/> > > - "finding default expressions of table \"%s\".\"%s\"\n"<br />> > - "finding
checkconstraints for table \"%s\".\"%s\"\n"<br />> Cool additions. There may be a more elegant way to check if
namespace<br/> > is NULL, but I couldn't come up with one myself. So patch may be fine.<br />><br /><br />Hi
all,<br/><br />I think this small patch was lost. There are something wrong?<br /><br />Regards,<br /><br />--<br
/>Fabríziode Royes Mello<br /> Consultoria/Coaching PostgreSQL<br />>> Timbira: <a
href="http://www.timbira.com.br">http://www.timbira.com.br</a><br/>>> Blog sobre TI: <a
href="http://fabriziomello.blogspot.com">http://fabriziomello.blogspot.com</a><br/> >> Perfil Linkedin: <a
href="http://br.linkedin.com/in/fabriziomello">http://br.linkedin.com/in/fabriziomello</a><br/>>> Twitter: <a
href="http://twitter.com/fabriziomello">http://twitter.com/fabriziomello</a></div>

On Wed, Jul 23, 2014 at 5:48 PM, Fabrízio de Royes Mello
<fabriziomello@gmail.com> wrote:
>
> On Thu, Apr 17, 2014 at 9:15 PM, Michael Paquier <michael.paquier@gmail.com>
> wrote:
>>
>> On Fri, Apr 18, 2014 at 4:29 AM, Fabrízio de Royes Mello
>> <fabriziomello@gmail.com> wrote:
>> >
>> > On Thu, Apr 17, 2014 at 12:46 PM, Bruce Momjian <bruce@momjian.us>
>> > wrote:
>> >>
>> >> On Thu, Apr 17, 2014 at 11:44:37AM -0400, Tom Lane wrote:
>> >> > Bruce Momjian <bruce@momjian.us> writes:
>> >> > > The idea is that we only need quotes when there are odd characters
>> >> > > in
>> >> > > the identifier.  We do that right now in some places, though I
>> >> > > can't
>> >> > > find them in pg_dump.  I know psql does that, see quote_ident().
>> >> >
>> >> > I think our general style rule is that identifiers embedded in
>> >> > messages
>> >> > are always double-quoted.  There's an exception for type names, but
>> >> > not otherwise.  You're confusing the message case with printing SQL.
>> >>
>> >> OK.  I was unclear if a status _display_ was a message like an error
>> >> message.
>> >>
>> >
>> > The attached patch fix missing double-quoted in "dumping contents of
>> > table.." message and add schema name to other messages:
>> > - "reading indexes for table \"%s\".\"%s\"\n"
>> > - "reading foreign key constraints for table \"%s\".\"%s\"\n"
>> > - "reading triggers for table \"%s\".\"%s\"\n"
>> >
>> > - "finding the columns and types of table \"%s\".\"%s\"\n"
>> > - "finding default expressions of table \"%s\".\"%s\"\n"
>> > - "finding check constraints for table \"%s\".\"%s\"\n"
>> Cool additions. There may be a more elegant way to check if namespace
>> is NULL, but I couldn't come up with one myself. So patch may be fine.
>>
>
> Hi all,
>
> I think this small patch was lost. There are something wrong?

Did it get added to a CommitFest?

I don't see it there.

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

On Fri, Jul 25, 2014 at 4:45 AM, Fabrízio de Royes Mello
<fabriziomello@gmail.com> wrote:
>
> Given this is a very small and simple patch I thought it's not necessary...
>
> Added to the next CommitFest.

I had a look at this patch, and here are a couple of comments:
1) Depending on how ArchiveEntry is called to register an object to
dump, namespace may be NULL, but it is not the case
namespace->dobj.name, so you could get the namespace name at the top
of the function that have their verbose output improved with something
like that:
const char *namespace = tbinfo->dobj.namespace ?
               tbinfo->dobj.namespace->dobj.name : NULL;
And then simplify the message output as follows:
if (namespace)
   write_msg("blah \"%s\".\"%s\" blah", namespace, classname);
else
   write_msg("blah \"%s\" blah", classname);
You can as well safely remove the checks on namespace->dobj.name.
2) I don't think that this is correct:
-                                       ahlog(AH, 1, "processing data
for table \"%s\"\n",
-                                                 te->tag);
+                                       ahlog(AH, 1, "processing data
for table \"%s\".\"%s\"\n",
+                                                 AH->currSchema, te->tag);
There are some code paths where AH->currSchema is set to NULL, and I
think that you should use te->namespace instead.
3) Changing only this message is not enough. The following verbose
messages need to be changed too for consistency:
- pg_dump: creating $tag $object
- pg_dump: setting owner and privileges for [blah]

I have been pondering as well about doing similar modifications to the
error message paths, but it did not seem worth it as this patch is
aimed only for the verbose output. Btw, I have basically fixed those
issues while doing the review, and finished with the attached patch.
Fabrizio, is this new version fine for you?
Regards,
--
Michael

On Thu, Aug 21, 2014 at 5:11 AM, Fabrízio de Royes Mello
<fabriziomello@gmail.com> wrote:
> I just change "if (tbinfo->dobj.namespace != NULL)" to "if
> (tbinfo->dobj.namespace)".
Fine for me. I am marking this patch as ready for committer.
--
Michael

<div dir="ltr"><div class="gmail_extra"><br />On Wed, Aug 20, 2014 at 8:24 PM, Michael Paquier <<a
href="mailto:michael.paquier@gmail.com">michael.paquier@gmail.com</a>>wrote:<br />><br />> On Thu, Aug 21,
2014at 5:11 AM, Fabrízio de Royes Mello<br /> > <<a
href="mailto:fabriziomello@gmail.com">fabriziomello@gmail.com</a>>wrote:<br />> > I just change "if
(tbinfo->dobj.namespace!= NULL)" to "if<br />> > (tbinfo->dobj.namespace)".<br /> > Fine for me. I am
markingthis patch as ready for committer.<br />><br /></div><div class="gmail_extra"><br />Thanks!<br /></div><div
class="gmail_extra"><br/>--<br />Fabrízio de Royes Mello<br />Consultoria/Coaching PostgreSQL<br /> >> Timbira:
<ahref="http://www.timbira.com.br">http://www.timbira.com.br</a><br />>> Blog: <a
href="http://fabriziomello.github.io">http://fabriziomello.github.io</a><br/>>> Linkedin: <a
href="http://br.linkedin.com/in/fabriziomello">http://br.linkedin.com/in/fabriziomello</a><br/> >> Twitter: <a
href="http://twitter.com/fabriziomello">http://twitter.com/fabriziomello</a><br/>>> Github: <a
href="http://github.com/fabriziomello">http://github.com/fabriziomello</a></div></div>

On 08/20/2014 11:11 PM, Fabrízio de Royes Mello wrote:
> On Wed, Aug 20, 2014 at 2:43 AM, Michael Paquier <michael.paquier@gmail.com>
> wrote:
>>
>> I had a look at this patch, and here are a couple of comments:
>> 1) Depending on how ArchiveEntry is called to register an object to
>> dump, namespace may be NULL, but it is not the case
>> namespace->dobj.name, so you could get the namespace name at the top
>> of the function that have their verbose output improved with something
>> like that:
>> const char *namespace = tbinfo->dobj.namespace ?
>>                 tbinfo->dobj.namespace->dobj.name : NULL;
>> And then simplify the message output as follows:
>> if (namespace)
>>     write_msg("blah \"%s\".\"%s\" blah", namespace, classname);
>> else
>>     write_msg("blah \"%s\" blah", classname);
>> You can as well safely remove the checks on namespace->dobj.name.
>
> Ok

AFAICS, the namespace can never be NULL in any of these. There is a 
"selectSourceSchema(fout, tbinfo->dobj.namespace->dobj.name)" call 
before or after printing the message, so if tbinfo->dobj.namespace is 
NULL, you'll crash anyway. Please double-check, and remove the dead code 
if you agree.

- Heikki

On Tue, Aug 26, 2014 at 3:48 AM, Heikki Linnakangas
<hlinnakangas@vmware.com> wrote:
> AFAICS, the namespace can never be NULL in any of these. There is a
> "selectSourceSchema(fout, tbinfo->dobj.namespace->dobj.name)" call before or
> after printing the message, so if tbinfo->dobj.namespace is NULL, you'll
> crash anyway. Please double-check, and remove the dead code if you agree.
Ah right, this field is used in many places. Even for
pg_backup_archiver.c, the portion of code processing data always has
the namespace set. I am sure that Fabrizio would have done that
quickly, but as I was on this thread I simplified the patch as
attached.
Regards,
--
Michael

On 08/26/2014 10:10 AM, Michael Paquier wrote:
> On Tue, Aug 26, 2014 at 3:48 AM, Heikki Linnakangas
> <hlinnakangas@vmware.com> wrote:
>> AFAICS, the namespace can never be NULL in any of these. There is a
>> "selectSourceSchema(fout, tbinfo->dobj.namespace->dobj.name)" call before or
>> after printing the message, so if tbinfo->dobj.namespace is NULL, you'll
>> crash anyway. Please double-check, and remove the dead code if you agree.
> Ah right, this field is used in many places. Even for
> pg_backup_archiver.c, the portion of code processing data always has
> the namespace set. I am sure that Fabrizio would have done that
> quickly, but as I was on this thread I simplified the patch as

Ok thanks, committed.

- Heikki

<div dir="ltr"><div class="gmail_extra"><br />On Tue, Aug 26, 2014 at 4:10 AM, Michael Paquier <<a
href="mailto:michael.paquier@gmail.com">michael.paquier@gmail.com</a>>wrote:<br />><br />> On Tue, Aug 26,
2014at 3:48 AM, Heikki Linnakangas<br /> > <<a
href="mailto:hlinnakangas@vmware.com">hlinnakangas@vmware.com</a>>wrote:<br />> > AFAICS, the namespace can
neverbe NULL in any of these. There is a<br />> > "selectSourceSchema(fout, tbinfo->dobj.namespace-><a
href="http://dobj.name">dobj.name</a>)"call before or<br /> > > after printing the message, so if
tbinfo->dobj.namespaceis NULL, you'll<br />> > crash anyway. Please double-check, and remove the dead code if
youagree.<br />> Ah right, this field is used in many places. Even for<br /> > pg_backup_archiver.c, the portion
ofcode processing data always has<br />> the namespace set. I am sure that Fabrizio would have done that<br />>
quickly,but as I was on this thread I simplified the patch as<br />> attached.<br /> ><br /><br /></div><div
class="gmail_extra">ThanksMichael... last night I was working on a refactoring in "tablecmds.c".<br /></div><div
class="gmail_extra"><br/></div><div class="gmail_extra">Regards,<br /><br /></div><div class="gmail_extra">--<br
/>Fabríziode Royes Mello<br />Consultoria/Coaching PostgreSQL<br />>> Timbira: <a
href="http://www.timbira.com.br">http://www.timbira.com.br</a><br/>>> Blog: <a
href="http://fabriziomello.github.io">http://fabriziomello.github.io</a><br/> >> Linkedin: <a
href="http://br.linkedin.com/in/fabriziomello">http://br.linkedin.com/in/fabriziomello</a><br/>>> Twitter: <a
href="http://twitter.com/fabriziomello">http://twitter.com/fabriziomello</a><br/>>> Github: <a
href="http://github.com/fabriziomello">http://github.com/fabriziomello</a></div></div>