Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

I have a custom query language which expresses containment relations between variables. Containment in this context is simply an object (A) in programming language X (java/C#/python etc: a language with support for objects) having a child object (B)

The statements of interest are like these

A CONTAINS B CONTAINS C AND D

The statement above describes all objects of type A that has a child object of type B which must have child objects of both C AND D. So this language defines containment constraints on objects, with support for boolean operators (AND, OR, NOT). Another example where parenthesis are used to define precedence would be:

A CONTAINS B CONTAINS (C AND (D CONTAINS E OR F))

I need to turn these statements to boolean expressions, and then I need to transform the boolean expressions to disjunctive normal form. I've came up with a method to do so, it is just that my method is ad hoc, and I don't really now if I can provide a mathematical definition of my transformation. Here is an attempt:

I'm taking every X CONTAINS Y statement on its own, and creating a variable in propositional logic represented by Y, that is X CONTAINS Y is a proposition that is represented by Y variable. To include X, I simply create a root proposition such as "Current data context CONTAINS X" which gives me X propositional logic variable. Therefore A CONTAINS B CONTAINS C AND D is transformed into the following set of propositions which in turn give me the variables:

CONTEXT CONTAINS A -> Boolean variable A    
A CONTAINS B ->Boolean variable B
B CONTAINS C ->Boolean variable C
B CONTAINS D -> Boolean variable D

and these propositions are turned into boolean variables and stitched together via boolean and/or/not operators:

A && B && (C && D) 

(I'm using && and || above to distinguish between AND keyword in the query statements which is an AND for containment.)

When boolean expressions are turned into disjunctive normal form, I end up re-writing the original containment statements so that OR operators end up as the root of operands which contain only AND/NOT operators.

Given what I'm doing, can I say that I am transforming the statements to propositional logic and then to boolean expressions? I would not like to use a mathematical definition of what I'm doing unless I'm sure my definition fits what I'm doing.

UPDATE In response to Henning Makholm's comments, here is more detail for anyone who would like to read: The query language I'm (partly) transforming is Archetype Query Language, a domain specific language for constructing queries on electronic health records. see: http://www.openehr.org/wiki/display/spec/Archetype+Query+Language+Description A, B and C above correspond to types of data elements. A real life example to containment query I've written above would be:

EHR e CONTAINS COMPOSITION c [openEHR-EHR-COMPOSITION.referral.v1] CONTAINS
(OBSERVATION o openEHR-EHR-OBSERVATION-laboratory-hba1c.v1 AND
OBSERVATION o1 openEHR-EHR-OBSERVATION-laboratory-glucose.v1)

My transformation pulls OR operators in these queries to top level, because it helps later when I generate SQL to pull data from a quite tricky database schema.

share|improve this question
add comment

1 Answer 1

up vote 1 down vote accepted

I have some issues:

First, if you "re-writ[e] the original containment statements so that OR operators end up as the root of operands which contain only AND/NOT operators", then the result is in disjunctive normal form, not conjunctive ditto.

Second, it seems to be a somewhat pointless detour to invent new propositional/Boolean variables for your atomic "this-contains-that" formulas simply in order to put it into disjunctive (or conjunctive) normal form. It is quite common to do such manipulation directly on logical formulae with atoms that are not just propositional variables. It is of course quite alright to let some valuables represent each "this-contains-that" atom, but you should describe those valuables at ranging over such atoms, not over truth values. (And please select them from another end of the alphabet, or from a different alphabet, than the A, B, C you're already using as type names. Otherwise it becomes very confusing to read).

Third, it looks to me like your description is conflating names of element types with variables that range over concrete elements. For example, I suppose that the query A contains B contains A ought to be valid and describe an A that contains some B which again contains some A, not necessarily the original one. (Indeed, the word "contains" seems to suggest that the relation is acyclic). So the query is not just a question of finding one $A$ and one $B$ such that ($A$ contains $B$) and ($B$ contains $A$).

Therefore I think when you translate A CONTAINS (B CONTAINS (C AND D)) you ought to invent new variables to stand for the subexpressions and generate the following clauses:

p is an A
q is a B
r is a C
s is a D
(context) contains p
p contains q
q contains r
q contains s

Furthermore, if you have control over the design of the query language, I wonder whether the intended semantics wouldn't be clearer if the keyword were CONTAINING rather than CONTAINS.

share|improve this answer
    
regarding your third point, you are right. I've tried to kept the query language related part abstract. I have no control over the language. My question is, is my definition of what I'm doing correct? –  sarikan Nov 10 '12 at 19:54
    
@sarikan: What I'm trying to say is that your use of variables for atomic formulas should be something that happens in your description of what you're doing -- but when you're actually doing it, it is simpler just to manipulate representations of the atomic formulas themselves. That way it is fine to write: "... now let $\phi$ stand for p contains q ...", Just don't say "$\phi$ is a Boolean variable", but rather "$\phi$ is a variable that stands for an atomic formula". –  Henning Makholm Nov 10 '12 at 19:56
    
would using "this contains that" atoms be wrong? I'd really like to hear your input about any issues you'd see with this approach. Thanks a lot for the detailed response. –  sarikan Nov 10 '12 at 19:57
    
My third point is that it sounds like you're doing something that won't work. That may either be because your description is badly written, or because you're actually doing something wrong. Or because I'm misunderstanding what you're trying to achieve. I can't tell. Using "this contains that" as atoms is fine as far as it goes, but the crucial point is what kind of things "this" and "that" are here. –  Henning Makholm Nov 10 '12 at 19:59
    
Thanks a lot. I am aware that I'm probably not doing a good job of describing the problem in a way that'd make others such as you easily comment. I've added some more details. Does that help? I'd be happy to go into more details if it makes it easier to comment, but I'd need guidance for that. How can I make the question easier to understand, especially regarding your comment "but the crucial point is what kind of things "this" and "that" are here" ? –  sarikan Nov 10 '12 at 20:54
show 3 more comments

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.