Subject: An interesting first beast...
From: AnHai Doan (anhai@cs.washington.edu)
Date: Mon Apr 09 2001 - 12:30:06 PDT
Lorel is one of the first query languages for
semi-structured data (UnQL being another one), predated the
XML age. In reading the paper, I was trying to find out what
are the main features of Lorel, where they go in subsequent
languages, and what features are missing from Lorel.
The OEM model is a simple and influential one. Most
subsequent languages use this model or a variation of it.
(UnQL uses a slightly different model, though.)
The implicit coercions is an interesting feature. Presumably
it can be interpreted in a way similar to implicit coercion
in Perl: Given an expression, first we determine a context,
then use the context to decide how to quietly convert the
values. Equality thus is context-based. In "05" = 5 and
5 = "5", it is in integer context, but in "05" = "5",
it is in string context. Technically, we cannot relate these
two kinds of equality because they are from different
contexts. So I don't think the claim of the paper that
equality is no longer transitive is valid.
Implicit coercion between a set and an element is
interesting, and I'm trying to find out where this feature
goes in subsequent languages...
The path variable stuff is also interesting. The issue here
is that when we care about the paths themselves, and there
are a number of practical situations where we do, there
maybe an infinite number of paths. No elegant solution is
proposed here and in subsequent languages.
Interesting features missing from Lorel are pattern
matching, template, and Skolem functions. My guess is that
both pattern matching and template are syntactic sugar: any
query expressed with them can be converted into one in Lorel
(albeit much more verbose). A more serious problem is that
Lorel lacks powerful restructuring facilities as in XML-QL.
All in all, Lorel makes for an interesting and reasonable
first beast in the world of query languages for
semi-structured data.
Hai.
This archive was generated by hypermail 2b25 : Mon Apr 09 2001 - 12:25:59 PDT