Building a Trie

Dear community,

Now I learned about Tries https://en.wikipedia.org/wiki/Trie which is possibly
a helpful method to tackle the left-over problem with my DemoJam contribution
at Markup UK (thanks a lot for the applause!).

Without spending too much time at the details, but at the heart of the
performance problem was this function:

<xsl:function name="my:wordMatches" as="xs:string*">
  <xsl:param name="start" as="xs:string"/>
  <xsl:sequence select="$words[starts-with(., $start)]"/>
</xsl:function>

With $words containing 25,000 entries, the performance was sufficient, but
with 160,000 entries it is just bad.

I already improved the performance by building a map, where the keys are the
first two letters of each word, and the value are all matching words.

<xsl:function name="my:wordMatches" as="xs:string*">
  <xsl:param name="start" as="xs:string"/>
  <xsl:variable name="key" select="substring($start, 1, 2)" as="xs:string"/>
  <xsl:variable name="words" select="map:get($wordsMap, $key)"
as="xs:string*"/>
  <xsl:sequence select="$words[starts-with(., $start)]"/>
</xsl:function>

Building that map takes some noticeable time, but then performance is great.
But only because I reduced the length of $words by factor 200b&300.

Before In look deeper myself, has anyone in the community already created a
Trie-like structure (map of maps?) or can share suggestions how to tackle this
in XSLT/XPath? (Because those are the hammers I plan to use for this
challenge/nail.)

Thanks,
- Michael

Michael MC<ller-Hillebrand
Senior Consultant
Phone +49 951-20859-752
Mobil +49 172-819 34 13
michael.mueller-hillebrand@xxxxxxxxx<mailto:michael.mueller-hillebrand@docufy
.de>
www.docufy.de<https://www.docufy.de/> |
DOCUFY@LinkedIN<https://www.linkedin.com/company/3845358/>
Datenschutz<https://www.docufy.de/datenschutz/>
DOCUFY GmbH | KirschC$ckerstr. 27 | 96052 Bamberg | Deutschland
CEO: Nadine Prill | Amtsgericht Bamberg HRB10571