Discussion:
[basex-talk] Strange behavior while adding existing elements into new element.
Jan Dölle
2017-06-20 18:09:38 UTC
Permalink
Hi,

just before you asking me. No we wasn't to isolate the Problem for now.

Working on a complex transaction with database involved, we get the
following phenomenon.

declare function m:paths($el as element())
as xs:string*
{
distinct-values(
for $e in $item/descendant::*
return
string-join($e/ancestor-or-self::*/local-name(.), "/")
)
};

This function should produce a list of paths to all child of $el. And it
works!

But sometimes we get some strange results. We created new element
constructed from others, may existing, elements.

let $sequence-of-elements := collect-some-elements($from-somewhere)
let $new-element := element el { $sequence-of-elements }
return paths($new-element)

Most of the time we get as expected results like:

el/c1
el/c1/c12
el/c2
el/c2/c21
el/c2/c22
el/..

But sometimes it looks like that parents of the copied elements are not
set correctly:

oldParentOfc1/c1
oldParentOfc1/c1/c12
otherOld/ParentOfc2/c2
otherOld/ParentOfc2/c2/c21
otherOld/ParentOfc2/c2/c22
...

If we do a copy of the result <el> like
parse-xml(serialize($new-element))/* and call paths() again, everything
looks now as expected.

We call the function what products such results, isolated within a
test-script with exact the same inputs. But suddenly the result was
correct.

Then we copied the inputs of the function within the integration
(parse-xml, serialize). But it failed again.

Also notable is, that the "bad result" is stable. It produces always the
same result on same inputs (no random).

So we have no idea to isolate the issue.

Any suggestions?

Best Regards

Jan
--
Jan Dölle
E-Mail: basex-***@id1.de Telefon +49-69-244502-0 Home: www.id1.de

Information Design One AG, Baseler Straße 10, 60329 Frankfurt am Main
Registereintrag: Amtsgericht Frankfurt am Main, HRB 52596
Vorstand: Robert Peters, Aufsichtsrat: Christian Hecht (Vorsitz)
Christian Grün
2017-06-20 18:21:34 UTC
Permalink
Hi Jan,

This reminds me of an open GitHub issue [1], but it’s just a guess.

Does your query work as expected if you append "update { }" to your
newly constructed element?

let $new-element := element el {
$sequence-of-elements } update { }

In BaseX, there are two different types of nodes: fragment nodes
(which are created by XQuery node constructors) and database nodes
(which are compact representations of XML trees). In your example, the
$sequence-of-elements (which, I guess, is a sequence of database
nodes) will be wrapped with an element constructor. By using "update
{}", your node will be copied and transformed to a database node. See
[2] for some more details on the update keyword.

Quite obviously this is just a workaround to test how your query
behaves. If it behaves correctly, I am pretty sure that [1] would need
to be fixed to get your query running correctly.

Cheers,
Christian

[1] https://github.com/BaseXdb/basex/issues/919
[2] http://docs.basex.org/wiki/Update#update
Post by Jan Dölle
Hi,
just before you asking me. No we wasn't to isolate the Problem for now.
Working on a complex transaction with database involved, we get the
following phenomenon.
declare function m:paths($el as element())
as xs:string*
{
distinct-values(
for $e in $item/descendant::*
return
string-join($e/ancestor-or-self::*/local-name(.), "/")
)
};
This function should produce a list of paths to all child of $el. And it
works!
But sometimes we get some strange results. We created new element
constructed from others, may existing, elements.
let $sequence-of-elements := collect-some-elements($from-somewhere)
let $new-element := element el { $sequence-of-elements }
return paths($new-element)
el/c1
el/c1/c12
el/c2
el/c2/c21
el/c2/c22
el/..
But sometimes it looks like that parents of the copied elements are not set
oldParentOfc1/c1
oldParentOfc1/c1/c12
otherOld/ParentOfc2/c2
otherOld/ParentOfc2/c2/c21
otherOld/ParentOfc2/c2/c22
...
If we do a copy of the result <el> like parse-xml(serialize($new-element))/*
and call paths() again, everything looks now as expected.
We call the function what products such results, isolated within a
test-script with exact the same inputs. But suddenly the result was correct.
Then we copied the inputs of the function within the integration (parse-xml,
serialize). But it failed again.
Also notable is, that the "bad result" is stable. It produces always the
same result on same inputs (no random).
So we have no idea to isolate the issue.
Any suggestions?
Best Regards
Jan
--
Jan Dölle
Information Design One AG, Baseler Straße 10, 60329 Frankfurt am Main
Registereintrag: Amtsgericht Frankfurt am Main, HRB 52596
Vorstand: Robert Peters, Aufsichtsrat: Christian Hecht (Vorsitz)
Jan Dölle
2017-06-21 13:54:54 UTC
Permalink
Hi Christian,

update {} as "quick copy" work.

Beside I checked the $sequence-of-elements for the special situation and
it only contains elements from a document created by parse-xml. But the
function which creates the sequence compares document elements against
some database elements and makes a choice to pick elements from document
or either db.

Thanks for your help

Jan
Post by Christian Grün
Hi Jan,
This reminds me of an open GitHub issue [1], but it’s just a guess.
Does your query work as expected if you append "update { }" to your
newly constructed element?
let $new-element := element el {
$sequence-of-elements } update { }
In BaseX, there are two different types of nodes: fragment nodes
(which are created by XQuery node constructors) and database nodes
(which are compact representations of XML trees). In your example, the
$sequence-of-elements (which, I guess, is a sequence of database
nodes) will be wrapped with an element constructor. By using "update
{}", your node will be copied and transformed to a database node. See
[2] for some more details on the update keyword.
Quite obviously this is just a workaround to test how your query
behaves. If it behaves correctly, I am pretty sure that [1] would need
to be fixed to get your query running correctly.
Cheers,
Christian
[1] https://github.com/BaseXdb/basex/issues/919
[2] http://docs.basex.org/wiki/Update#update
Post by Jan Dölle
Hi,
just before you asking me. No we wasn't to isolate the Problem for now.
Working on a complex transaction with database involved, we get the
following phenomenon.
declare function m:paths($el as element())
as xs:string*
{
distinct-values(
for $e in $item/descendant::*
return
string-join($e/ancestor-or-self::*/local-name(.), "/")
)
};
This function should produce a list of paths to all child of $el. And it
works!
But sometimes we get some strange results. We created new element
constructed from others, may existing, elements.
let $sequence-of-elements := collect-some-elements($from-somewhere)
let $new-element := element el { $sequence-of-elements }
return paths($new-element)
el/c1
el/c1/c12
el/c2
el/c2/c21
el/c2/c22
el/..
But sometimes it looks like that parents of the copied elements are not set
oldParentOfc1/c1
oldParentOfc1/c1/c12
otherOld/ParentOfc2/c2
otherOld/ParentOfc2/c2/c21
otherOld/ParentOfc2/c2/c22
...
If we do a copy of the result <el> like parse-xml(serialize($new-element))/*
and call paths() again, everything looks now as expected.
We call the function what products such results, isolated within a
test-script with exact the same inputs. But suddenly the result was correct.
Then we copied the inputs of the function within the integration (parse-xml,
serialize). But it failed again.
Also notable is, that the "bad result" is stable. It produces always the
same result on same inputs (no random).
So we have no idea to isolate the issue.
Any suggestions?
Best Regards
Jan
--
Jan Dölle
Information Design One AG, Baseler Straße 10, 60329 Frankfurt am Main
Registereintrag: Amtsgericht Frankfurt am Main, HRB 52596
Vorstand: Robert Peters, Aufsichtsrat: Christian Hecht (Vorsitz)
Loading...