Thanks, Christian. Much appreciated advice. The only change needed was due to the fact that trials often have multiple ids of the same type (e.g., secondary_id) or empty ids (acronym is optional). Also, ids are sometimes actually lists of ids separated by commas (poor coding practice by sponsors). So the working code looks as follows:

let $maps :=
for $trial in db:open('CTGov')/clinical_study
return (
for $id in $trial/id_info/org_study_id for $t in tokenize($id, ',') return map:entry(functx:trim($t), $trial),
for $id in $trial/id_info/secondary_id for $t in tokenize($id, ',') return map:entry(functx:trim($t), $trial),
for $id in $trial/acronym for $t in tokenize($id, ',') return map:entry(functx:trim($t), $trial)
)

The new code completes in 10 seconds, a major improvement!

Best,
Ron

On December 30, 2015 at 1:23:44 PM, Christian Grün (christian.gruen@gmail.com) wrote:

A last hint: Performance is mostly a matter of how maps are
implemented in a particular XQuery processor. You may be able to speed
up your query if you only iterate over all nodes only once:

let $maps :=
for $trial in db:open('CTGov')/clinical_study
return (
map:entry(string($trial/id_info/org_study_id), $trial),
map:entry(string($trial/id_info/secondary_id), $trial),
map:entry(string($trial/acronym), $trial)
)
return map:merge(
for $map in $maps
for $key in map:keys($map)
group by $key
let $value := ($map ! .($key))/self::node()
return map { $key : $value }
)

If map keys are converted to strings, lookup will be a bit faster.
Next, the "/self::node()" step removes potential duplicate nodes (just
skip it if the IDs are unique assigned to a single "clinical_study"
element are distinct).




On Wed, Dec 30, 2015 at 7:13 PM, Ron Katriel <rkatriel@mdsol.com> wrote:
> Done (Bug 29353).
>
> Thanks,
> Ron
>
> On December 30, 2015 at 9:57:00 AM, Christian Grün
> (christian.gruen@gmail.com) wrote:
>
>> Any chance this could be made an option of map:merge? Doing it in Java
>> would
>> be faster and more elegant.
>
> It will be too late for XQuery 3.1, but feel free to motivate such
> enhancements in the W3 Bug Tracker [1].
>
> Christian
>
> [1] https://www.w3.org/Bugs/Public/
>
>
>>
>> Best,
>> Ron
>>
>> On December 29, 2015 at 3:12:28 AM, Christian Grün
>> (christian.gruen@gmail.com) wrote:
>>
>> Try this:
>>
>> let $maps := (
>> map:entry(0, "red"),
>> map:entry(1, "green"),
>> map:entry(1, "blue")
>> )
>> return map:merge(
>> for $map in $maps
>> for $key in map:keys($map)
>> group by $key
>> return map { $key : $map ! .($key) }
>> )
>>
>> This is an equivalent, possibly better readable, solution:
>>
>> let $maps := (
>> map:entry(0, "red"),
>> map:entry(1, "green"),
>> map:entry(1, "blue")
>> )
>> let $keys := distinct-values(
>> for $map in $maps
>> return map:keys($map)
>> )
>> return map:merge(
>> for $key in $keys
>> let $value :=
>> for $map in $maps
>> return $map($key)
>> return map { $key : $value }
>> )
>>
>>
>> On Wed, Dec 23, 2015 at 11:04 PM, Ron Katriel <rkatriel@mdsol.com> wrote:
>>> Hi,
>>>
>>> I am using map:merge to construct a map from smaller maps and would like
>>> to
>>> preserve values when keys agree. For example, when calling
>>>
>>> map:merge((map:entry(0, "red"), (map:entry(1, "green"), map:entry(1,
>>> "blue")))
>>>
>>> I would like to get back something like
>>>
>>> map { 0: "red", 1: ("green", "blue") }
>>>
>>> The default (W3C) behavior is to drop "green" in favor of "blue".
>>>
>>> Is there a simple way to accomplish this? I realize the above example is
>>> mixing types so presumably a solution would have all values as sets.
>>>
>>> Thanks,
>>> Ron
>>>