Mailing list for the Sedna XML Database ()

SourceForge.net | 8 Nov 2010 15:38

[ sedna-Bugs-3105299 ] Update functions can break XDM consistency!

Bugs item #3105299, was opened at 2010-11-08 17:38
Message generated for change (Tracker Item Submitted) made by mozinsur
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=713730&aid=3105299&group_id=129076

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: Update Language
Group: Sedna (current)
Status: Open
Resolution: None
Priority: 6
Private: No
Submitted By: Ilya Taranov (mozinsur)
Assigned to: Nobody/Anonymous (nobody)
Summary: Update functions can break XDM consistency!

Initial Comment:
The following code:

[code]
create document "a"&
update insert <a a="1" c="2" /> into doc("a")&
update replace doc("a")/a/ <at> a with (attribute b {"1.5"}, <x />)&
[/code]

leads to a broken data model. This can be told by 

[code]

(Continue reading)

headers

Zoltán Zörgő | 6 Nov 2010 21:37

Creating index on document in a collection

Dear all!

I am newbie in Sedna, and I am trying to create a value-index on a document in a collection. The statement is:

create index "book-title" on doc("books.xml", "testcollection")/catalog/book by title as xs:string

But I get the folowing error:

SEDNA Message: ERROR SE5049 Not a structural XPath supplied. Details: at (1:30), document-in-collection is not permitted in on-XPath

But doc("books.xml","testcollection")/catalog/book/title runs correctly.

I found no example in the documentation on this.

The index creation in collection is not permitted, not implemented or I am doing something wrong?

Thank You

Zoltan

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@...
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

headers

Mario Trojan | 19 Nov 2010 21:16

Handling of large XML files / performance

Hello,

i just wanted to know if it is also possible to load (and manipule)
large files, which can be hundreds of megabytes, or even multiple
gigabytes. I tried to load a file into sedna via php api (240 MB), and
it took 463 seconds to import it (on win 64). I recognized that the
se_sm.exe process was working all the time, but only using about 14MB of
memory.

How can the import be accelerated? Couldn't find anything in the
documentation about that topic.

I also noticed that there is only one function for loading xml
documents, and thats "sedna_load", where you need to pass the whole
document as a string. Meaning it is necessary to parse the whole 240MB
into a string (-> into memory), so that you can pass it to the function.
Or is there a better way to do this?

Best regards,
Mario

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

headers

Ivan Shcheklein | 19 Nov 2010 21:41

Re: Handling of large XML files / performance

Hi Mario,

i just wanted to know if it is also possible to load (and manipule)
large files, which can be hundreds of megabytes, or even multiple
gigabytes. I tried to load a file into sedna via php api (240 MB), and
it took 463 seconds to import it (on win 64). I recognized that the
se_sm.exe process was working all the time, but only using about 14MB of
memory.

Yes, it's possible. See wikixmldb.org for example - 36Gb XML file. And works pretty fine.

How can the import be accelerated? Couldn't find anything in the
documentation about that topic.

There are two things you can do:

1. Increase buffers (use -bufs-num SM's option - http://modis.ispras.ru/sedna/adminguide/AdminGuidesu3.html#x7-130002.3.4). Default value 100Mb is small for complex and big data.
2. Use log less mode (to turn it on in se_term, see LOG_LESS_MODE here http://modis.ispras.ru/sedna/adminguide/AdminGuidesu4.html#x8-150002.4). This option is also available via php API: SE OPTID LOGLESS.

I also noticed that there is only one function for loading xml
documents, and thats "sedna_load", where you need to pass the whole
document as a string. Meaning it is necessary to parse the whole 240MB
into a string (-> into memory), so that you can pass it to the function.
Or is there a better way to do this?

No, AFAIR. You should use some other way (se_term, C API, Java API) to perfrom initial bulk load of really big data.

Ivan Shcheklein,
Sedna Team

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@...
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

headers

Ivan Shcheklein | 19 Nov 2010 21:48

Re: Creating index on document in a collection

Hi Zoltán,

I am newbie in Sedna, and I am trying to create a value-index on a document in a collection. The statement is:

create index "book-title" on doc("books.xml", "testcollection")/catalog/book by title as xs:string

But I get the folowing error:

SEDNA Message: ERROR SE5049 Not a structural XPath supplied. Details: at (1:30), document-in-collection is not permitted in on-XPath

But doc("books.xml","testcollection")/catalog/book/title runs correctly.

It's not possible to create index on a single document within a collection. It's only possible to create an index on the whole collection, like:

create index "book-title" on collection("testcollection")/catalog/book by title as xs:string

If you need to create indexes on individual documents don't load them into collection.

Ivan Shcheklein,
Sedna Team

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@...
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

headers

Malcolm Davis | 19 Nov 2010 22:34

Java connection management & SE3007

From the FAQ section of Sedna website on Windows issues: "SE3007: Failed to receive a message" is blamed on Winsock2 issues.

I was able to recreate the SE3007 problem using Sedna’s Java ConnectionManager in a multi thread application.

I was able to resolve the SE3007 by writing my own connection wrapper around the ConnectionManager.

I most have spent time investigating project/code at www.cfoster.net rather than writing my own.

Malcolm

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@...
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

headers

Spencer Cheng | 21 Nov 2010 14:10

deadlock resolution procedure?

Could anyone point me to more documentation on the subject?

I have 2 Ruby processes accessing Sedna via the Ruby binding. After I run the processes for for a few minutes,
one of the process dies with "The transaction is a victim of deadlock resolution procedure
(Sedna::Exception)". 

Regards,
Spencer

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

headers

Ivan Shcheklein | 21 Nov 2010 14:37

Re: deadlock resolution procedure?

Hi Spencer,

Could anyone point me to more documentation on the subject?
I have 2 Ruby processes accessing Sedna via the Ruby binding. After I run the processes for for a few minutes, one of the process dies with "The transaction is a victim of deadlock resolution procedure (Sedna::Exception)".

See http://en.wikipedia.org/wiki/Two-phase_locking, Deadlocks in 2PL . Sedna aborts transaction using wait-for-graph as described in that article. You should manually restart it if you need.

Ivan Shcheklein,
Sedna Team

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@...
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

headers

Ivan Shcheklein | 21 Nov 2010 14:40

Re: Java connection management & SE3007

Hi Malcolm,

>From the FAQ section of Sedna website on Windows issues: "SE3007: Failed to receive a message" is blamed on Winsock2 issues.

I was able to recreate the SE3007 problem using Sedna’s Java ConnectionManager in a multi thread application.

I was able to resolve the SE3007 by writing my own connection wrapper around the ConnectionManager.

I most have spent time investigating project/code at www.cfoster.net rather than writing my own.

It this error still issue for you?
Most likely, you should just properly close connection in a finally { } block.

Ivan Shcheklein,
Sedna Team

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

_______________________________________________
Sedna-discussion mailing list
Sedna-discussion@...
https://lists.sourceforge.net/lists/listinfo/sedna-discussion

headers

Spencer Cheng | 21 Nov 2010 16:39

Re: deadlock resolution procedure?

Hi Ivan,

On Nov 21, 2010, at 8:37, Ivan Shcheklein wrote:
> 
> See http://en.wikipedia.org/wiki/Two-phase_locking, Deadlocks in 2PL . Sedna
> aborts transaction using wait-for-graph as described in that article. You
> should manually restart it if you need.
> 

I can't quite figure out why I am getting the deadlock though.

My (FIFO queue-like) code looks like this in ruby'esque code

Process 1
---------

sedna.execute("update insert <data>...</data> preceding doc('somedoc')/a/b/tail")

Process 2
----------

sedna.transaction {
  sedna.execute("doc('somedoc')/a/b/data[1]")
  sedna.execute("update delete doc('somedoc')/a/b/data[1]")
}

My assumption was that Process 1 using only a single query would be atomic and Process 2 requires the
transaction wrapper to maintain atomicity. If Process 2 gets context switched after the 1st query and
Process 1 tries to execute a query, Process 1 will block and Process 2 will get to run and complete the
transaction before Process 1 can execute it's query.

Is here an error in my logic?

Regards,
Spencer

------------------------------------------------------------------------------
Beautiful is writing same markup. Internet Explorer 9 supports
standards for HTML5, CSS3, SVG 1.1,  ECMAScript5, and DOM L2 & L3.
Spend less time writing and  rewriting code and more time creating great
experiences on the web. Be a part of the beta today
http://p.sf.net/sfu/msIE9-sfdev2dev

7	June 2013
6	May 2013
27	April 2013
24	March 2013
34	February 2013
36	January 2013
6	December 2012
15	November 2012
12	October 2012
29	September 2012
16	August 2012
21	July 2012
13	June 2012
7	May 2012
19	April 2012
36	March 2012
26	February 2012
22	January 2012
28	December 2011
21	November 2011
20	October 2011
9	September 2011
26	August 2011
24	July 2011
41	June 2011
33	May 2011
46	April 2011
41	March 2011
32	February 2011
5	January 2011
44	December 2010
12	November 2010
60	October 2010
13	September 2010
13	August 2010
16	July 2010
29	June 2010
33	May 2010
18	April 2010
21	March 2010
16	February 2010
24	January 2010
18	December 2009
28	November 2009
18	October 2009
29	September 2009
67	August 2009
39	July 2009
37	June 2009
8	May 2009
18	April 2009
49	March 2009
51	February 2009
15	January 2009
54	December 2008
55	November 2008
68	October 2008
79	September 2008
33	August 2008
20	July 2008
34	June 2008
23	May 2008
48	April 2008
41	March 2008
24	February 2008
71	January 2008
16	December 2007
25	November 2007
4	October 2007
8	September 2007
15	August 2007
34	July 2007
62	June 2007
12	May 2007
35	April 2007
13	March 2007
49	February 2007
9	January 2007
2	December 2006
6	November 2006
18	October 2006
73	September 2006
30	August 2006
28	July 2006
28	June 2006
19	May 2006
30	April 2006
2	March 2006
22	February 2006
54	January 2006
35	December 2005
36	November 2005
27	October 2005
10	September 2005
7	August 2005
20	July 2005
52	June 2005
73	May 2005
38	April 2005
2	February 2005

Mon	Tue	Wed	Thu	Fri	Sat	Sun

1	2	3	4	5	6	7
8	9	10	11	12	13	14
15	16	17	18	19	20	21
22	23	24	25	26	27	28
29	30