Hi Constantine<div>Just my side note.<br><div>As far as I remember, I found this sort of performance discussion with most the XQuery processors I have tested (Sedna, Zorba, Saxon, eXist...). Each time related to sibling, preceding etc. and each time explaining, that indices are not containing pointers for tracking this relationship. My feeling is, this could lead to much bigger index file and slower performance when adding documents and indexing.</div>
<div>My conlcusion is - optimizing this sort of queries has its cost and mostly it is better (and somehow possible) to avoid these constructions.</div><div><br></div><div>With best regards</div><div><br></div><div>Jan</div>
<div><br clear="all"><div name="sig_b0a60a250d" style="margin-top:2em;margin-right:0pt;margin-bottom:2em;margin-left:0pt"><b>Ing. Jan Vlčinský</b><br><small>
CAD programy<br>
Slunečnicová 338/3,  734 01 Karviná Ráj, Czech Republic<br>
tel: +420-597 602 024;
mob: +420-608 979 040<br>
skype: janvlcinsky;
GoogleTalk: <a href="mailto:jan.vlcinsky@gmail.com" target="_blank">jan.vlcinsky@gmail.com</a><br>
<a href="http://cz.linkedin.com/in/vlcinsky" target="_blank">http://cz.linkedin.com/in/vlcinsky</a>
</small></div><br>
<br><br><div class="gmail_quote">On 20 October 2011 15:43, Christian Grün <span dir="ltr">&lt;<a href="mailto:christian.gruen@gmail.com">christian.gruen@gmail.com</a>&gt;</span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
Hi Constantine,<br>
<div class="im"><br>
&gt; The XPath I showed was just a simple example of using the preceding-sibling axis.<br>
<br>
</div>I guessed so already.. Thanks for persisting ;)<br>
<div class="im"><br>
&gt; I guess what I am asking is ... while most XPath queries I run are quick to resolve, anything using preceding::, preceding-sibling:: or following::, following-sibling:: essentially runs too slowly to be workable - at least, on my sample dataset with about 100K sibling children of the root element.<br>

&gt;<br>
&gt; So is this any sort of known issue? Should I be looking at any sort of configuration options? Is there any way to profile the execution of the Xpath?<br>
<br>
</div>It&#39;s true that BaseX is mostly optimized for descendant and ancestor<br>
operations. In particular, the preceding/preceding-sibling axes will<br>
always be slower, as our storage does not contain backward pointers to<br>
preceding nodes. I&#39;m pretty sure, though, that there are ways to<br>
either optimize relevant queries or the internal optimizer strategies,<br>
but we&#39;ll probably have to spend some more time on analyzing the<br>
relevant bottlenecks..<br>
<div class="im"><br>
Feel free to ask for more,<br>
Christian<br>
</div><div><div></div><div class="h5">_______________________________________________<br>
BaseX-Talk mailing list<br>
<a href="mailto:BaseX-Talk@mailman.uni-konstanz.de">BaseX-Talk@mailman.uni-konstanz.de</a><br>
<a href="https://mailman.uni-konstanz.de/mailman/listinfo/basex-talk" target="_blank">https://mailman.uni-konstanz.de/mailman/listinfo/basex-talk</a><br>
</div></div></blockquote></div><br></div></div>