Arjen P. de Vries | 2 Aug 14:52 2007
Picon
Picon

Re: Any efforts on using bitmap indices?


Hi,

We have this high on the wishlist, but had no time to do it yet.

The direction we expect to take is to make the underlying text indexing stuff 
from PF/Tijah accessible from SQL.

Notice that we are not so much interested in creating another coupled architecture,
but rather plan to do all FT query processing in plain MonetDB code.
(We have named this approach a multi-model dbms architecture in the past)
So, no integration with Lucene nor Postgres's engine.

Many greetings,

Arjen

| Thanks for the link, i didnt know MonetDB had fts support. However i am
| looking for the full text search capability in the SQL. Something like
| 
| select * from table where column ='+include this text -but with out this'
| so that we can combine the boolean text query along with the normal joins.
| FastBit ( accorrding to the doc) has integrated this with MonetDB by
| overriding the = operator. 
| 
| Thanks
| Bharani

-------------------------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc.
(Continue reading)

Bharani | 2 Aug 14:59 2007
Picon

Re: Any efforts on using bitmap indices?


Hi,
The only comfort with lucene is the query syntax and the speed becuase of
inverted index ( might be faster in BAT but have to wait to compare the
performance). I will be happy to use it when it is available.

Thanks
Bharani

Arjen P. de Vries wrote:
> 
> 
> Hi,
> 
> We have this high on the wishlist, but had no time to do it yet.
> 
> The direction we expect to take is to make the underlying text indexing
> stuff 
> from PF/Tijah accessible from SQL.
> 
> Notice that we are not so much interested in creating another coupled
> architecture,
> but rather plan to do all FT query processing in plain MonetDB code.
> (We have named this approach a multi-model dbms architecture in the past)
> So, no integration with Lucene nor Postgres's engine.
> 
> Many greetings,
> 
> Arjen
> 
(Continue reading)

Stefan Manegold | 2 Aug 23:05 2007
Picon
Picon

Re: pathfinder/tests/BugsViaSourgeforce/Tests ID.1766414.stable.err, , NONE, 1.1 ID.1766414.stable.out, , NONE, 1.1 ID.1766414.xml, , NONE, 1.1 ID.1766414.xq, , NONE, 1.1 All, , 1.34, 1.35

On Thu, Aug 02, 2007 at 08:42:23PM +0000, Peter Boncz wrote:
> Update of /cvsroot/monetdb/pathfinder/tests/BugsViaSourgeforce/Tests
> In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv13362
> 
> Modified Files:
> 	All 
> Added Files:
> 	ID.1766414.stable.err ID.1766414.stable.out ID.1766414.xml 
> 	ID.1766414.xq 
> Log Message:
> adding stable output, produced on branch -- will still fail in head until 

Why hasn't this test been added to the release branch?
(Apparently, the bug occured and was fixed there...)

> I propagate the fix

Is there an urgend need for individual "by hand" propagation (with all
possible conflicts it might cause for the ustandard (semi-)automa[tgn]ic
propagation by Sjoerd and/or me or does the head-fix require more than the
usual standard (semi-)automa[tgn]ic propagation by Sjoerd and/or me?

If 2x no, I'd prefer to have this done during / via the usual standard
(semi-)automa[tgn]ic propagation by Sjoerd and/or me.

(Please feel free to request a usual standard (semi-)automa[tgn]ic
propagation by Sjoerd and/or me whenever you "need" to have changes
propagated from the latest release branch to the development head
"urgendly".)

(Continue reading)

Maurice van Keulen | 3 Aug 15:36 2007
Picon
Picon

Re: Unimplemented XQuery standard functions

Hi Djoerd,

I desperately needed tokenize() myself, so I wrote my own version in
XQuery. Incredably slow, but it works. Note that $sep should be a string
with only one character.

declare function pxml:poormansTokenize($s as xs:string, $sep as xs:string)
{
   let $slen := string-length($s)
      ,$split := for $i in (1 to $slen) return substring($s,$i,1)
      ,$pos := (0,(for $c at $p in $split where $c eq $sep return
$p),$slen+1)
      ,$len := for $i in (2 to count($pos))
               return exactly-one($pos[$i]) - exactly-one($pos[$i - 1]) - 1
   for $p at $i in $len
   return substring($s,exactly-one($pos[$i])+1,$p)
};

Hope this helps.

Kind regards,
Maurice.

Djoerd Hiemstra wrote:
> Dear all,
>
> For demo purposes we are in need for implementations of some XQuery
> standard functions, especially fn:tokenize() and fn:base-uri().
>
> Are there plans for implementing more (or all) standard functions?
(Continue reading)

Sjoerd Mullender | 6 Aug 09:39 2007
Picon

Re: MonetDB/src/gdk gdk_bbp.mx,,1.201,1.202

Should this be back-ported to the stable branch?

Peter Boncz wrote:
> Update of /cvsroot/monetdb/MonetDB/src/gdk
> In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv30287/src/gdk
> 
> Modified Files:
> 	gdk_bbp.mx 
> Log Message:
> - race condition in BBOclear
> 
> 
> 
> Index: gdk_bbp.mx
> ===================================================================
> RCS file: /cvsroot/monetdb/MonetDB/src/gdk/gdk_bbp.mx,v
> retrieving revision 1.201
> retrieving revision 1.202
> diff -u -d -r1.201 -r1.202
> --- gdk_bbp.mx	22 Jun 2007 06:46:53 -0000	1.201
> +++ gdk_bbp.mx	5 Aug 2007 23:00:50 -0000	1.202
>  <at>  <at>  -1116,7 +1116,7  <at>  <at> 
>  		THRprintf(GDKout, "#clear %d (%s)\n", (int) i, BBPname(i));
>  	}
>  	BBPuncacheit_(i, TRUE);
> -	BBP_status_set(i, 0, "BBPclear");
> +	BBP_status_set(i, BBPUNLOADING, "BBPclear");
>  	BBP_refs(i) = 0;
>  	BBP_lrefs(i) = 0;
>  	if (lock)
(Continue reading)

Romulo Goncalves | 8 Aug 18:12 2007
Picon
Picon

Re: [Monetdb-checkins] clients/src/mapiclient MapiClient.mx, , 1.8, 1.9

Martin Kersten wrote:
> Update of /cvsroot/monetdb/clients/src/mapiclient
> In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv19893
> 
> Modified Files:
> 	MapiClient.mx 
> Log Message:
> MapiClient has been extended with a SQL rendering option based on
> a predefined columnwidth and pagesize.
> A columnwidth= -1 means old display structure.
> Columnwidth=0 indicates to respect the sizes in the table headers.
> Columnwidth>0 is an upperbound. Broad columns are pushed to make
> all fit. The long values are displayed over multiple lines.
If columnwidth is bigger than the table headers it equivalent to set it 
to zero.

For columnwidth=5 and columnswith=10 I do not see any differences in the 
output, why?

The error output is different.

[goncalve <at> amelia src]$ MapiClient -lsql --columnwidth=-1 --pagesize=-1
sql>select * from table;
MAPI  = monetdb <at> localhost:50000
QUERY = select * from table;
ERROR = !syntax error, unexpected TABLE in: "select * from table"
sql>\q

[goncalve <at> amelia src]$ MapiClient -lsql --columnwidth=0 --pagesize=-1
sql>select * from table;
(Continue reading)

Martin Kersten | 8 Aug 21:44 2007
Picon
Picon

Re: [Monetdb-checkins] clients/src/mapiclient MapiClient.mx, , 1.8, 1.9

Thanks for the first reaction.

Romulo Goncalves wrote:
> Martin Kersten wrote:
>> Update of /cvsroot/monetdb/clients/src/mapiclient
>> In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv19893
>>
>> Modified Files:
>> 	MapiClient.mx 
>> Log Message:
>> MapiClient has been extended with a SQL rendering option based on
>> a predefined columnwidth and pagesize.
>> A columnwidth= -1 means old display structure.
>> Columnwidth=0 indicates to respect the sizes in the table headers.
>> Columnwidth>0 is an upperbound. Broad columns are pushed to make
>> all fit. The long values are displayed over multiple lines.
> If columnwidth is bigger than the table headers it equivalent to set it 
> to zero.

yes that's true and intended. The pagewidth (=columnwidth) is an upperbound.
> 
> For columnwidth=5 and columnswith=10 I do not see any differences in the 
> output, why?
> 
Now it does.
 > mclient -lsql
sql>\w5
sql>select 1;
+----+
| si |
(Continue reading)

Agustin Schapira | 8 Aug 21:52 2007
Picon

glib - double free or corruption

Dear Monet developers,

We're having trouble with a script that uses the {} operator to  
compute the avg of a large (4M rows) table: we get a 'double free'  
error from glib, and Monet crashes. This is using Monet 4.16.2  
compiled for 64-bits and 32-bit oids on Linux. (BTW, it also happens  
if we use {sum}, but not if we use {count})

The code takes two tables, links and attr, and does the equivalent of

SELECT avg(attr.value)
FROM   links, attr
WHERE  links.id = attr.id
GROUP BY links.from

Here's the MIL code:

# Get the BATs
var var_attr:=bat(bat("prox_link_attr").fetch(2)).find(oid(0));
var var_attr_id:=bat(bat(var_attr).fetch(0));
var var_attr_val:=bat(bat(var_attr).fetch(1));

var var_link_id:=bat(bat("prox_link").fetch(0));
var var_link_from:=bat(bat("prox_link").fetch(1));

# Join ATTR x LINK, keep LINK.from, ATTR.val
var var_1:=var_link_id.join(var_attr_id.reverse());
var var_2:=var_1.mark(0 <at> 0);
var var_3:=var_1.reverse().mark(0 <at> 0);
var var_5:=var_2.reverse().join(var_link_from);
(Continue reading)

Romulo Goncalves | 9 Aug 10:17 2007
Picon
Picon

Re: [Monetdb-checkins] clients/src/mapiclient MapiClient.mx, , 1.8, 1.9

Martin Kersten wrote:
> Thanks for the first reaction.
> 
> Romulo Goncalves wrote:
>> Martin Kersten wrote:
>>> Update of /cvsroot/monetdb/clients/src/mapiclient
>>> In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv19893
>>>
>>> Modified Files:
>>>     MapiClient.mx Log Message:
>>> MapiClient has been extended with a SQL rendering option based on
>>> a predefined columnwidth and pagesize.
>>> A columnwidth= -1 means old display structure.
>>> Columnwidth=0 indicates to respect the sizes in the table headers.
>>> Columnwidth>0 is an upperbound. Broad columns are pushed to make
>>> all fit. The long values are displayed over multiple lines.
>> If columnwidth is bigger than the table headers it equivalent to set 
>> it to zero.
> 
> yes that's true and intended. The pagewidth (=columnwidth) is an 
> upperbound.
>>
>> For columnwidth=5 and columnswith=10 I do not see any differences in 
>> the output, why?
>>
> Now it does.
>  > mclient -lsql
> sql>\w5
> sql>select 1;
> +----+
(Continue reading)

Sjoerd Mullender | 12 Aug 21:27 2007
Picon

Re: clients/src/mapiclient dump.c,,1.7,1.8

On 08/12/2007 08:39 PM, Martin Kersten wrote:
> Update of /cvsroot/monetdb/clients/src/mapiclient
> In directory sc8-pr-cvs16.sourceforge.net:/tmp/cvs-serv25741
> 
> Modified Files:
> 	dump.c 
> Log Message:
> Remove duplicate rows. This solves the duplicates seen
> when typing '\d idxs'.

Is this the right thing to do?  The problem, it seems to me, is that
idxs occurs twice in the tables table (select * from tables where name =
'idxs'; returns two entries).  Is that correct?

> Index: dump.c
> ===================================================================
> RCS file: /cvsroot/monetdb/clients/src/mapiclient/dump.c,v
> retrieving revision 1.7
> retrieving revision 1.8
> diff -u -d -r1.7 -r1.8
> --- dump.c	10 Aug 2007 11:48:59 -0000	1.7
> +++ dump.c	12 Aug 2007 18:39:52 -0000	1.8
>  <at>  <at>  -79,7 +79,7  <at>  <at> 
>  	fprintf(toConsole, " (\n");
>  
>  	snprintf(query, maxquerylen,
> -		 "SELECT \"c\".\"name\","		/* 0 */
> +		 "SELECT DISTINCT \"c\".\"name\","		/* 0 */
>  			"\"c\".\"type\","		/* 1 */
>  			"\"c\".\"type_digits\","	/* 2 */
(Continue reading)


Gmane