Parsing an array of dates in XQuery (MarkLogic v8 flavour) - xquery

Question

My question seems trivial but could not figure out how to parse a string that contains a list of dates separated by commas. The parsing part of individual dates is not an issue but the empty values are. The trouble is the order of the dates are significant and some dates can be omitted. The dates are expected to be formatted in YYYY-mm-dd
So, the following are valid inputs and expected return values:
,2000-12-12,2012-05-03, ➔ ( NULL, 2000-12-12, 2012-05-03, NULL )
2000-12-12,,2012-05-03 ➔ ( 2000-12-12, NULL, 2012-05-03 )
And here is my function signature
declare function local:assert-date-array-param(
$input as xs:string
, $accept-nulls as xs:boolean?
) as xs:date*
I recognised the problem after realizing that there seems to be no equivalent of of NULL in XQuery for the returned values as placeholders for omitted dates, if you want to return a sequence, that is. Because empty sequences wrapped inside sequences are flattened to nothing.
I suppose, my fallback would be to use date like 1900-01-01 as the placeholder or return a map instead of a sequence, but I sure hope to find a more elegant way
Thank you,
K.
PS. I am working with MarkLogic v8 (and v9 soon) and any solution should execute with their XQuery processor.
UPDATE: thanks for both answers, in the end I chose to go with a placeholder date as XQuery works so nicely with sequences and anything else would have required some changes in other places. But the problem remains for the cases where the required return values are numerics. In this case, use of placeholder values would probably be not feasible. A null literal for xs:anyAtomicType would have solved the problem nicely, but alas.

Answer 1

You could consider returning a json:array(), or an array-node{} with null-node{}'s inside. But perhaps a null-date placeholder is not as bad as it sounds:
declare variable $null-date := xs:date("0001-01-01");
declare function local:assert-date-array-param(
$input as xs:string,
$accept-nulls as xs:boolean?
) as xs:date*
{
for $d in fn:tokenize($input, "\s*,\s*")
return
if ($d eq "") then
if ($accept-nulls) then
$null-date
else
fn:error(xs:QName("NULL-NOT-ALLOWED"), "Date is required")
else
if ($d castable as xs:date) then
xs:date($d)
else if ($d castable as xs:dateTime) then
xs:date(xs:dateTime($d))
else
fn:error(xs:QName("INVALID-DATE"), "Invalid date format: " || $d)
};
declare function local:print-date-array($dates) {
string-join(for $d in $dates return if ($d eq $null-date) then "NULL" else fn:string($d), ", ")
};
local:print-date-array(
local:assert-date-array-param(",2000-12-12,2012-05-03,", fn:true())
),
local:print-date-array(
local:assert-date-array-param("2000-12-12,,2012-05-03", fn:true())
)
HTH!

Answer 2

Multiple options .. in addition to above.
return a sequence of functions which when invoked return dates
for $i in string-tokenize-to-sequence-of-strings()
let $dt := my-parse-date($i)
return function() { $dt ;}
or
return function() { return my-parse-date($i) ;
return tokenized and validated but not parsed strings. Use "" for 'not valid', e.g.:
( "2014-01-22","","2017-03-30","" )
then there's arrays', maps, arrays of maps, and ... XML
parseFunction() as xs:element()*:
for ... return <date>{ parse-and-validate($value) } </date>