- From: thomas lörtsch <tl@rat.io>
- Date: Sat, 8 Jan 2022 22:13:30 +0100
- To: public-rdf-star@w3.org
Hi Peter! > Am 12.12.2021 um 06:19 schrieb Peter F. Patel-Schneider <pfpschneider@gmail.com>: > > Semantics for linear logic can be based on multisets. > Here is one such treatment: > http://users.auth.gr/tzouvara/Texfiles.htm/multlog.pdf > > There is a treatment of SPARQL that uses multisets: > https://arxiv.org/abs/1610.04315 Thanks a lot for the pointers! As was to be expected this is way over my head… I read both papers and they both seem excellent and well written, yet two or three pages in I’m lost, very much. I still managed to gather something from those first pages. The first, more foundational paper is not cited very often. The topic doesn’t seem to be very hotly debated. It makes this curious statemet quite at the beginning: "For reasons hidden in the early history of set theory, a set came to mean a collection of types of objects rather than of concrete tokens of them." I was of course a bit disappointed that the paper didn’t get into these hidden reasons but the second paper, by Angles and Gutiérrez, starts with an answer to that question: "The incorporation of multisets (also called 'duplicates' or 'bags') into the semantics of query languages like SQL or SPARQL is essentially due to practical concerns: duplicate elimination is expensive and duplicates might be required for some applications, e.g. for aggregation. […] The theory behind these query languages is relational algebra or equivalently, relational calculus, formalisms that for sets have a clean and intuitive semantics for users, developers and theoreticians. The same cannot be said of their extensions to multisets, whose theory is complex (particular containment of queries) and their practical use not always clear for users and developers. Worst, there exist several possible ways of extending set relational operators to multisets and one can find them in practice." - followed by some very compelling examples of how a UNION might be interpreted in quite different ways - all very intuitive, yet incompatible with each other. They then explain how SPARQL multiset semantics straightforwardly match to one specific interpretation of UNION grounded in non-recursive Datalog with safe negation and "come up with a Multiset Relational Algebra (MRA) that captures precisely the multiset semantics of the core relational patterns of SPARQL" - quite impressive indeed. Thanks again, Thomas
Received on Saturday, 8 January 2022 21:13:47 UTC