Asking unrelated questions on a thread is risky - people don't
necessarily read the thread if the subject is not of interest to them.
Especially here when it is about the (extremely!) fine detail of how XSD
durations work and have evolved over versions.
The reason why public query benchmarks have faded is the realisation
that useful numbers come from (1) significant attention to the runtime
environment and (2) the use case of interest.
A benchmark execute by someone else on their use case and data has only
so much use.
Ultimately the question becomes "is it fast enough for my usage with the
configuration I want in the environment I will be using?" and the way to
know, is to try.
The same is true in the SQL world with TPC.
The fastest databases aren't the ones growing in usage like crazy at the
moment.
BSBM is better than LUMB in being a bit more realistic (LUBM is, as the
original authors have said, no longer that helpful) but it is quite
small and focuses on pattern-filter queries. There are two optimizations
that ARQ does that determine the efficiency - all the other
optimizations don't have much effect on the queries+data of BSBM but do
elsewhere.
Straight loads tests are more reliable as loading has less dimensions -
the shape of the data does matter but in query there is the whole
universe of query possibilities and features.
LDBC publish some benchmarks (data and query sets) by use case and they
do run professional benchmarking. It is on a pay-to-participate basis
for certified results (think TPC).
What are you interested in?
Inference?
Data publishing or data analytics?
...
Andy
On 28/01/2021 02:01, Fang, Xiao Shan wrote:
Dear all,
Does TDB has a latest benchmarking performance test result? I found a Berlin
sparql test benchmarking result. But it reports on 2008.
Thanks,
-xiaoshan
-----Original Message-----
From: Andy Seaborne [mailto:a...@apache.org]
Sent: 2021年1月28日 6:26
To: users@jena.apache.org
Subject: Re: XPath Duration functions in Jena
Hi Harri,
https://www.w3.org/TR/xpath-functions-3/#func-subtract-dates
The difference of two dates is the duration bwteeen the dates as points on the
timeline as a xs:dayTimeDuration.
?age is "P14484DT0H0M0.000S"^^xsd:dayTimeDuration
not PnnYnnMnnD
A duration is a defined, fixed length of time.
There two kinds of durations xs:dayTimeDuration and xs:yearMonthDuration
https://www.w3.org/TR/xpath-functions-3/#duration-subtypes
where arithmetic and ordering can be meaningfully defined at all.
A combination of months and days is not a fixed interval of time because the
number of days in a month isn't the same for all months.
Only xs:dayTimeDuration defines a duration that is of a well-defined difference
of two points on the timeline (ignoring negative leap seconds).
Once "xsd:date - xsd:date" is calculated, the duration is a fixed length of
time and does not tie back to which dates were used in the calculation.
So "P14484D.." is right not "P44Y6m??D"
[[
Confusingly the difference of two Gregorian year-months (gYearMonth) is also an
xs:dayTimeDuration, and not a xs:yearMonthDuration, not that it is really
defined on gYearMonth (section 9.1), because gYearMonth maps to a point on the
timeline and the difference is of two points on the timeline.
]]
With ?age is "P14484DT0H0M0.000S"^^xsd:dayTimeDuration
For the casting to xsd:yearMonthDuration: (had to check this one!)
Casting xsd:dayTimeDuration to xsd:yearMonthDuration is defined.
https://www.w3.org/TR/xpath-functions-31/#casting-from-primitive-to-primitive
https://www.w3.org/TR/xpath-functions-31/#casting-to-durations
"""
If ST is xs:yearMonthDuration and TT is xs:dayTimeDuration, the cast is
permitted and returns a xs:dayTimeDuration with value 0.0 seconds.
"""
ST is "source type" TT is target type".
I don't know the rationale - (a guess - treat like it is selecting fields from
PnnYnnMnnDT..)
Andy
On 27/01/2021 17:16, Harri Kiiskinen wrote:
Dear all,
This example run with Fuseki 3.17:
--------------------------------------------------------------------
PREFIX fn: <http://www.w3.org/2005/xpath-functions#>
PREFIX xsd: <http://www.w3.org/2001/XMLSchema#>
SELECT ?y ?d (xsd:dayTimeDuration(?age) as ?a1)
(xsd:yearMonthDuration(?age) as ?a2) where {
bind ( xsd:date("1984-12-01") as ?y)
bind ( xsd:date("1944-05-06") as ?d)
bind ((?y-?d) as ?age )
}
-------------------------------------------------------------------
Gives unexpected results:
------------------------------------------------------------------
y,d,a1,a2
1984-12-01,1944-05-06,P14484DT0H0M0.000S,P0M
------------------------------------------------------------------
I was expecting the variable "a2" to contain the duration expressed in
Years and Months, but instead it is zero.
What am I doing wrong here?
Best,
Harri Kiiskinen
