[Exist-open] Possible optimization problem

[Anthony M. <hi...@sp...>]

First of all I am using eXist 1.4.1-rev15155-20110815 hosted on a Windows 7
box using jdk1.7.0_01.

I noticed that I was getting results from a function call that were
incorrect and I had been under the impression that the function call was
working just fine.  In troubleshooting I soon found that if I called the
function directly I'd get one result, but the code that I was developing
persisted in returning a different result.  I finally narrowed it down to
the following test case:

If I execute the query:

xquery version "1.0";
import module namespace mtrc="http://xquery.xtellus.com/modules/mtrc" at
"xmldb:exist:///db/dev/modules/mtrc.xqm";

let $name := 'by-trace'
return mtrc:get-colors($name)


I get the following result (expected):

<colors>
    <color>#FF0000</color>
    <color>#00FF00</color>
    <color>#0000FF</color>
    <color>#FF00FF</color>
    <color>#EEC900</color>
    <color>#B0C4DE</color>
    <color>#2E8B57</color>
    <color>#A020F0</color>
    <color>#6B8E23</color>
    <color>#F5DEB3</color>
</colors>

If I execute the query:

xquery version "1.0";
import module namespace mtrc="http://xquery.xtellus.com/modules/mtrc" at
"xmldb:exist:///db/dev/modules/mtrc.xqm";

let $name := 'by-trace'
return typeswitch (element { 'foo' } {}) case element(foo) return
mtrc:get-colors($name) default return ()

I get the following result (un-expected):

<colors>
    <color>#FF0000</color>
    <color>#00FF00</color>
    <color>#0000FF</color>
    <color>#FF00FF</color>
    <color>#EEC900</color>
    <color>#</color>
    <color>#</color>
    <color>#A020F0</color>
    <color>#698B22</color>
    <color>#8B7E66</color>
</colors>

The mtrc:get-colors function:

declare function mtrc:get-colors($name as xs:string) as node() {
    let $doc  := doc('/db/dev/color-sets.xml')/*
    let $name := if ($doc/color-set[@name=$name]) then $name else 'by-trace'
    return
    <colors>{
        for $color in co:hex-values($doc/color-set[@name=$name]/string())
return <color>{$color}</color>
    }</colors>
};

The relevant part of the co: module:

declare function co:first-color($name as xs:string, $list as xs:string) as
node()? {
    let $nodes :=
        if ($list ne '') then

doc($co:color-path)/*/color-list[@src=$list]/color[ngram:starts-with(@name,
$name)]
        else
            doc($co:color-path)//color[ngram:starts-with(@name, $name)]
    return $nodes[1]
};

declare function co:hex-values($names as item()*) as xs:string* {
    co:hex-values($names, 'x11')
};

declare function co:hex-values($names as item()*, $list as xs:string) as
xs:string* {
    let $names := if ($names instance of xs:string) then tokenize($names,
',') else for $n in $names return string($n)
    return for $name in $names return concat('#', co:first-color($name,
$list)/@hex/string())
};


Basically, the call to mtrc:get-colors winds up calling co:hex-values with
a string argument that looks like:
'red,green,blue,magenta,gold2,lightsteelblue,sea green,purple,olive
drab,wheat' and co:hex-values looks them up in a big list of colors and
returns hex values.

The same problem exists in version 1.4.2-rev16251-20120416

Other than a comment that the query to 8ms to execute there is nothing in
the log to indicate what happened.

What I am worried about is that it is a common pattern for me to end my
queries with a typeswitch to return different formats of the results of the
query, I can work around this case easy enough, but now I'm worried about
the integrity of a considerable body of queries.

Anthony