obscure kwic behaviour

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

obscure kwic behaviour

Immanuel Normann
Hi,

I observe a very obscure kwic:expand behaviour that seems to depend on the html namespace:
The kwic:expand function doesn't have any effect in some cases even though Lucene has found something. Yet, when I change the namespace from "http://www.w3.org/1999/xhtml" to something else (say "foo") it is in operation again.

This is the setting to replicate this strange behaviour:

collection.conf:

<collection xmlns="http://exist-db.org/collection-config/1.0">
    <index xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:foo="foo" xmlns:xlink="http://www.w3.org/1999/xlink">
        <fulltext default="none" attributes="false"/>
        <lucene>
            <analyzer class="org.apache.lucene.analysis.standard.StandardAnalyzer">
            </analyzer>
            <text qname="xhtml:div"/>
            <text qname="foo:div"/>
        </lucene>
    </index>
</collection>


namespace-html.xml:

<html xmlns="http://www.w3.org/1999/xhtml">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


namespace-foo.xml:

<html xmlns="foo">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


test-search.xql:

xquery version "3.0";

import module namespace kwic="http://exist-db.org/xquery/kwic";
declare namespace foo = "foo";
declare namespace html = "http://www.w3.org/1999/xhtml";
declare variable $namespace-foo := doc("/db/howto/lucene/data/namespace-foo.xml");
declare variable $namespace-html := doc("/db/howto/lucene/data/namespace-html.xml");

<test>
    <result query="Par*" ns="foo">{kwic:expand($namespace-foo//foo:div[ft:query(.,"Par*")])}</result>
    <result query="Par*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Par*")])}</result>
    <result query="Para*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Para*")])}</result>
</test>


Running test-search.xql:

$ curl http://localhost:8080/exist/rest/howto/lucene/test-search.xql
<test>
    <result query="Par*" ns="foo">
        <div xmlns="foo">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
    <result query="Par*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">Paragraph</div>
    </result>
    <result query="Para*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
</test>


The strange result is obviously <result query="Par*" ns="html"> as it doesn't surround "Paragraph" with <exist:match ...>!
Whereas in <result query="Par*" ns="html"> the <exist:match ...> is in place as expected. Note that the only difference here is the namespace! What makes it even more strange is <result query="Para*" ns="html"> which is in the same namespace as <result query="Par*" ns="html"> but with the little difference in the query string: now we search for "Para*" (instead of just "Par*") and this again returns a result with <exist:match ...> is in place.

In summary: <exist:match ...> should be in place for "Par*" as well as for "Para*" independent of the namespace. The bug is that it fails for "Par*" in the html namespace.

Of cause this strange observation does not depend on the word "Paragraph". There are vaious other words yielding same strange results.
For instance, assume we have "Berlin" instead of "Paragraph" in our XML-files. Then "Ber*" would fail like "Par*" fails in the html namespace - and again only in the html space. A hypothetical explanation could be that the kwic:expand always fails in the html namespace whenever the truncated search string is to short ("Ber*" fails for "Berlin" and "Par*" fails for "Paragraph" whereas "Para*" succeeds for "Paragraph").
Yet this hypothesis can be validated with this example: Assume we have "Perlin" instead of "Berlin" in our xml-files then "Per*" surprisingly succeeds for "Perlin" (whereas "Ber*" fails for "Berlin").
After all I don't have any clue what is going on behind the scene here. May be some of you?

Cheers
Immanuel

P.S.: My eXist build info:

project.version=2.2
project.built=20141120124006
scm.branch=master
scm.revision=5c5aadc





------------------------------------------------------------------------------

_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open
Reply | Threaded
Open this post in threaded view
|

Re: obscure kwic behaviour

Jens Østergaard Petersen-2
Hi Immanuel,

I get matches in all three tests, but then I am running a build from around last week of the develop branch of 3.0 …. And I call my collection config “collection.xconf”.

Best,

Jens

On 11 December 2015 at 19:15:02, Immanuel Normann ([hidden email]) wrote:

Hi,

I observe a very obscure kwic:expand behaviour that seems to depend on the html namespace:
The kwic:expand function doesn't have any effect in some cases even though Lucene has found something. Yet, when I change the namespace from "http://www.w3.org/1999/xhtml" to something else (say "foo") it is in operation again.

This is the setting to replicate this strange behaviour:

collection.conf:

<collection xmlns="http://exist-db.org/collection-config/1.0">
    <index xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:foo="foo" xmlns:xlink="http://www.w3.org/1999/xlink">
        <fulltext default="none" attributes="false"/>
        <lucene>
            <analyzer class="org.apache.lucene.analysis.standard.StandardAnalyzer">
            </analyzer>
            <text qname="xhtml:div"/>
            <text qname="foo:div"/>
        </lucene>
    </index>
</collection>


namespace-html.xml:

<html xmlns="http://www.w3.org/1999/xhtml">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


namespace-foo.xml:

<html xmlns="foo">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


test-search.xql:

xquery version "3.0";

import module namespace kwic="http://exist-db.org/xquery/kwic";
declare namespace foo = "foo";
declare namespace html = "http://www.w3.org/1999/xhtml";
declare variable $namespace-foo := doc("/db/howto/lucene/data/namespace-foo.xml");
declare variable $namespace-html := doc("/db/howto/lucene/data/namespace-html.xml");

<test>
    <result query="Par*" ns="foo">{kwic:expand($namespace-foo//foo:div[ft:query(.,"Par*")])}</result>
    <result query="Par*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Par*")])}</result>
    <result query="Para*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Para*")])}</result>
</test>


Running test-search.xql:

$ curl http://localhost:8080/exist/rest/howto/lucene/test-search.xql
<test>
    <result query="Par*" ns="foo">
        <div xmlns="foo">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
    <result query="Par*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">Paragraph</div>
    </result>
    <result query="Para*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
</test>


The strange result is obviously <result query="Par*" ns="html"> as it doesn't surround "Paragraph" with <exist:match ...>!
Whereas in <result query="Par*" ns="html"> the <exist:match ...> is in place as expected. Note that the only difference here is the namespace! What makes it even more strange is <result query="Para*" ns="html"> which is in the same namespace as <result query="Par*" ns="html"> but with the little difference in the query string: now we search for "Para*" (instead of just "Par*") and this again returns a result with <exist:match ...> is in place.

In summary: <exist:match ...> should be in place for "Par*" as well as for "Para*" independent of the namespace. The bug is that it fails for "Par*" in the html namespace.

Of cause this strange observation does not depend on the word "Paragraph". There are vaious other words yielding same strange results.
For instance, assume we have "Berlin" instead of "Paragraph" in our XML-files. Then "Ber*" would fail like "Par*" fails in the html namespace - and again only in the html space. A hypothetical explanation could be that the kwic:expand always fails in the html namespace whenever the truncated search string is to short ("Ber*" fails for "Berlin" and "Par*" fails for "Paragraph" whereas "Para*" succeeds for "Paragraph").
Yet this hypothesis can be validated with this example: Assume we have "Perlin" instead of "Berlin" in our xml-files then "Per*" surprisingly succeeds for "Perlin" (whereas "Ber*" fails for "Berlin").
After all I don't have any clue what is going on behind the scene here. May be some of you?

Cheers
Immanuel

P.S.: My eXist build info:

project.version=2.2
project.built=20141120124006
scm.branch=master
scm.revision=5c5aadc




------------------------------------------------------------------------------
_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open

------------------------------------------------------------------------------

_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open
Reply | Threaded
Open this post in threaded view
|

Re: obscure kwic behaviour

Immanuel Normann
Hi Jens,

so the problem seemed to disappear in 3.0 but it is present in 2.2. As 2.2 is distributed as stable release the community might still be interested to know about the problem reported in my first email.
Regarding "collection.conf": this was just a spelling error in my email. In fact I call it collection.xconf, too.

Cheers
Immaneuel

2015-12-12 11:48 GMT+01:00 Jens Østergaard Petersen <[hidden email]>:
Hi Immanuel,

I get matches in all three tests, but then I am running a build from around last week of the develop branch of 3.0 …. And I call my collection config “collection.xconf”.

Best,

Jens

On 11 December 2015 at 19:15:02, Immanuel Normann ([hidden email]) wrote:

Hi,

I observe a very obscure kwic:expand behaviour that seems to depend on the html namespace:
The kwic:expand function doesn't have any effect in some cases even though Lucene has found something. Yet, when I change the namespace from "http://www.w3.org/1999/xhtml" to something else (say "foo") it is in operation again.

This is the setting to replicate this strange behaviour:

collection.conf:

<collection xmlns="http://exist-db.org/collection-config/1.0">
    <index xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:foo="foo" xmlns:xlink="http://www.w3.org/1999/xlink">
        <fulltext default="none" attributes="false"/>
        <lucene>
            <analyzer class="org.apache.lucene.analysis.standard.StandardAnalyzer">
            </analyzer>
            <text qname="xhtml:div"/>
            <text qname="foo:div"/>
        </lucene>
    </index>
</collection>


namespace-html.xml:

<html xmlns="http://www.w3.org/1999/xhtml">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


namespace-foo.xml:

<html xmlns="foo">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


test-search.xql:

xquery version "3.0";

import module namespace kwic="http://exist-db.org/xquery/kwic";
declare namespace foo = "foo";
declare namespace html = "http://www.w3.org/1999/xhtml";
declare variable $namespace-foo := doc("/db/howto/lucene/data/namespace-foo.xml");
declare variable $namespace-html := doc("/db/howto/lucene/data/namespace-html.xml");

<test>
    <result query="Par*" ns="foo">{kwic:expand($namespace-foo//foo:div[ft:query(.,"Par*")])}</result>
    <result query="Par*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Par*")])}</result>
    <result query="Para*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Para*")])}</result>
</test>


Running test-search.xql:

$ curl http://localhost:8080/exist/rest/howto/lucene/test-search.xql
<test>
    <result query="Par*" ns="foo">
        <div xmlns="foo">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
    <result query="Par*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">Paragraph</div>
    </result>
    <result query="Para*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
</test>


The strange result is obviously <result query="Par*" ns="html"> as it doesn't surround "Paragraph" with <exist:match ...>!
Whereas in <result query="Par*" ns="html"> the <exist:match ...> is in place as expected. Note that the only difference here is the namespace! What makes it even more strange is <result query="Para*" ns="html"> which is in the same namespace as <result query="Par*" ns="html"> but with the little difference in the query string: now we search for "Para*" (instead of just "Par*") and this again returns a result with <exist:match ...> is in place.

In summary: <exist:match ...> should be in place for "Par*" as well as for "Para*" independent of the namespace. The bug is that it fails for "Par*" in the html namespace.

Of cause this strange observation does not depend on the word "Paragraph". There are vaious other words yielding same strange results.
For instance, assume we have "Berlin" instead of "Paragraph" in our XML-files. Then "Ber*" would fail like "Par*" fails in the html namespace - and again only in the html space. A hypothetical explanation could be that the kwic:expand always fails in the html namespace whenever the truncated search string is to short ("Ber*" fails for "Berlin" and "Par*" fails for "Paragraph" whereas "Para*" succeeds for "Paragraph").
Yet this hypothesis can be validated with this example: Assume we have "Perlin" instead of "Berlin" in our xml-files then "Per*" surprisingly succeeds for "Perlin" (whereas "Ber*" fails for "Berlin").
After all I don't have any clue what is going on behind the scene here. May be some of you?

Cheers
Immanuel

P.S.: My eXist build info:

project.version=2.2
project.built=20141120124006
scm.branch=master
scm.revision=5c5aadc




------------------------------------------------------------------------------
_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open


------------------------------------------------------------------------------

_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open
Reply | Threaded
Open this post in threaded view
|

Re: obscure kwic behaviour

Jens Østergaard Petersen-2
Hi Immanuel,

Yes, you are right, but I just tried out your tests on a fresh install of stable 2.2, and I get three matches there as well.

Best,

Jens

On 14 December 2015 at 08:19:39, Immanuel Normann ([hidden email]) wrote:

Hi Jens,

so the problem seemed to disappear in 3.0 but it is present in 2.2. As 2.2 is distributed as stable release the community might still be interested to know about the problem reported in my first email.
Regarding "collection.conf": this was just a spelling error in my email. In fact I call it collection.xconf, too.

Cheers
Immaneuel

2015-12-12 11:48 GMT+01:00 Jens Østergaard Petersen <[hidden email]>:
Hi Immanuel,

I get matches in all three tests, but then I am running a build from around last week of the develop branch of 3.0 …. And I call my collection config “collection.xconf”.

Best,

Jens

On 11 December 2015 at 19:15:02, Immanuel Normann ([hidden email]) wrote:

Hi,

I observe a very obscure kwic:expand behaviour that seems to depend on the html namespace:
The kwic:expand function doesn't have any effect in some cases even though Lucene has found something. Yet, when I change the namespace from "http://www.w3.org/1999/xhtml" to something else (say "foo") it is in operation again.

This is the setting to replicate this strange behaviour:

collection.conf:

<collection xmlns="http://exist-db.org/collection-config/1.0">
    <index xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:foo="foo" xmlns:xlink="http://www.w3.org/1999/xlink">
        <fulltext default="none" attributes="false"/>
        <lucene>
            <analyzer class="org.apache.lucene.analysis.standard.StandardAnalyzer">
            </analyzer>
            <text qname="xhtml:div"/>
            <text qname="foo:div"/>
        </lucene>
    </index>
</collection>


namespace-html.xml:

<html xmlns="http://www.w3.org/1999/xhtml">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


namespace-foo.xml:

<html xmlns="foo">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


test-search.xql:

xquery version "3.0";

import module namespace kwic="http://exist-db.org/xquery/kwic";
declare namespace foo = "foo";
declare namespace html = "http://www.w3.org/1999/xhtml";
declare variable $namespace-foo := doc("/db/howto/lucene/data/namespace-foo.xml");
declare variable $namespace-html := doc("/db/howto/lucene/data/namespace-html.xml");

<test>
    <result query="Par*" ns="foo">{kwic:expand($namespace-foo//foo:div[ft:query(.,"Par*")])}</result>
    <result query="Par*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Par*")])}</result>
    <result query="Para*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Para*")])}</result>
</test>


Running test-search.xql:

$ curl http://localhost:8080/exist/rest/howto/lucene/test-search.xql
<test>
    <result query="Par*" ns="foo">
        <div xmlns="foo">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
    <result query="Par*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">Paragraph</div>
    </result>
    <result query="Para*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
</test>


The strange result is obviously <result query="Par*" ns="html"> as it doesn't surround "Paragraph" with <exist:match ...>!
Whereas in <result query="Par*" ns="html"> the <exist:match ...> is in place as expected. Note that the only difference here is the namespace! What makes it even more strange is <result query="Para*" ns="html"> which is in the same namespace as <result query="Par*" ns="html"> but with the little difference in the query string: now we search for "Para*" (instead of just "Par*") and this again returns a result with <exist:match ...> is in place.

In summary: <exist:match ...> should be in place for "Par*" as well as for "Para*" independent of the namespace. The bug is that it fails for "Par*" in the html namespace.

Of cause this strange observation does not depend on the word "Paragraph". There are vaious other words yielding same strange results.
For instance, assume we have "Berlin" instead of "Paragraph" in our XML-files. Then "Ber*" would fail like "Par*" fails in the html namespace - and again only in the html space. A hypothetical explanation could be that the kwic:expand always fails in the html namespace whenever the truncated search string is to short ("Ber*" fails for "Berlin" and "Par*" fails for "Paragraph" whereas "Para*" succeeds for "Paragraph").
Yet this hypothesis can be validated with this example: Assume we have "Perlin" instead of "Berlin" in our xml-files then "Per*" surprisingly succeeds for "Perlin" (whereas "Ber*" fails for "Berlin").
After all I don't have any clue what is going on behind the scene here. May be some of you?

Cheers
Immanuel

P.S.: My eXist build info:

project.version=2.2
project.built=20141120124006
scm.branch=master
scm.revision=5c5aadc




------------------------------------------------------------------------------
_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open


------------------------------------------------------------------------------

_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open
Reply | Threaded
Open this post in threaded view
|

Re: obscure kwic behaviour

Immanuel Normann
Hi Jens,

You are right! Now, I also installed a fresh and stable 2.2 release and now everything works fine!
It was just the built that I used which apparently wasn't stable:

project.version=2.2
project.built=20141120124006
scm.branch=master
scm.revision=5c5aadc

Thanks for your effort!

Best,
Immanuel

2015-12-14 13:51 GMT+01:00 Jens Østergaard Petersen <[hidden email]>:
Hi Immanuel,

Yes, you are right, but I just tried out your tests on a fresh install of stable 2.2, and I get three matches there as well.

Best,

Jens

On 14 December 2015 at 08:19:39, Immanuel Normann ([hidden email]) wrote:

Hi Jens,

so the problem seemed to disappear in 3.0 but it is present in 2.2. As 2.2 is distributed as stable release the community might still be interested to know about the problem reported in my first email.
Regarding "collection.conf": this was just a spelling error in my email. In fact I call it collection.xconf, too.

Cheers
Immaneuel

2015-12-12 11:48 GMT+01:00 Jens Østergaard Petersen <[hidden email]>:
Hi Immanuel,

I get matches in all three tests, but then I am running a build from around last week of the develop branch of 3.0 …. And I call my collection config “collection.xconf”.

Best,

Jens

On 11 December 2015 at 19:15:02, Immanuel Normann ([hidden email]) wrote:

Hi,

I observe a very obscure kwic:expand behaviour that seems to depend on the html namespace:
The kwic:expand function doesn't have any effect in some cases even though Lucene has found something. Yet, when I change the namespace from "http://www.w3.org/1999/xhtml" to something else (say "foo") it is in operation again.

This is the setting to replicate this strange behaviour:

collection.conf:

<collection xmlns="http://exist-db.org/collection-config/1.0">
    <index xmlns:xhtml="http://www.w3.org/1999/xhtml" xmlns:foo="foo" xmlns:xlink="http://www.w3.org/1999/xlink">
        <fulltext default="none" attributes="false"/>
        <lucene>
            <analyzer class="org.apache.lucene.analysis.standard.StandardAnalyzer">
            </analyzer>
            <text qname="xhtml:div"/>
            <text qname="foo:div"/>
        </lucene>
    </index>
</collection>


namespace-html.xml:

<html xmlns="http://www.w3.org/1999/xhtml">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


namespace-foo.xml:

<html xmlns="foo">
    <head><title/></head>
    <body>
        <div>Paragraph</div>
    </body>
</html>


test-search.xql:

xquery version "3.0";

import module namespace kwic="http://exist-db.org/xquery/kwic";
declare namespace foo = "foo";
declare namespace html = "http://www.w3.org/1999/xhtml";
declare variable $namespace-foo := doc("/db/howto/lucene/data/namespace-foo.xml");
declare variable $namespace-html := doc("/db/howto/lucene/data/namespace-html.xml");

<test>
    <result query="Par*" ns="foo">{kwic:expand($namespace-foo//foo:div[ft:query(.,"Par*")])}</result>
    <result query="Par*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Par*")])}</result>
    <result query="Para*" ns="html">{kwic:expand($namespace-html//html:div[ft:query(.,"Para*")])}</result>
</test>


Running test-search.xql:

$ curl http://localhost:8080/exist/rest/howto/lucene/test-search.xql
<test>
    <result query="Par*" ns="foo">
        <div xmlns="foo">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
    <result query="Par*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">Paragraph</div>
    </result>
    <result query="Para*" ns="html">
        <div xmlns="http://www.w3.org/1999/xhtml">
            <exist:match xmlns:exist="http://exist.sourceforge.net/NS/exist">Paragraph</exist:match>
        </div>
    </result>
</test>


The strange result is obviously <result query="Par*" ns="html"> as it doesn't surround "Paragraph" with <exist:match ...>!
Whereas in <result query="Par*" ns="html"> the <exist:match ...> is in place as expected. Note that the only difference here is the namespace! What makes it even more strange is <result query="Para*" ns="html"> which is in the same namespace as <result query="Par*" ns="html"> but with the little difference in the query string: now we search for "Para*" (instead of just "Par*") and this again returns a result with <exist:match ...> is in place.

In summary: <exist:match ...> should be in place for "Par*" as well as for "Para*" independent of the namespace. The bug is that it fails for "Par*" in the html namespace.

Of cause this strange observation does not depend on the word "Paragraph". There are vaious other words yielding same strange results.
For instance, assume we have "Berlin" instead of "Paragraph" in our XML-files. Then "Ber*" would fail like "Par*" fails in the html namespace - and again only in the html space. A hypothetical explanation could be that the kwic:expand always fails in the html namespace whenever the truncated search string is to short ("Ber*" fails for "Berlin" and "Par*" fails for "Paragraph" whereas "Para*" succeeds for "Paragraph").
Yet this hypothesis can be validated with this example: Assume we have "Perlin" instead of "Berlin" in our xml-files then "Per*" surprisingly succeeds for "Perlin" (whereas "Ber*" fails for "Berlin").
After all I don't have any clue what is going on behind the scene here. May be some of you?

Cheers
Immanuel

P.S.: My eXist build info:

project.version=2.2
project.built=20141120124006
scm.branch=master
scm.revision=5c5aadc




------------------------------------------------------------------------------
_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open



------------------------------------------------------------------------------

_______________________________________________
Exist-open mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/exist-open