Closed szarnyasg closed 6 years ago
I'd prefer calling them ((Non)?Incremental)?QueryPlan. If you have to explain what the name means it's not a very good name.
Well, good point, but then we'd have other lousy class names like NonIncremental2IncrementalQueryPlanTransformation
. I think all (current and future) ingraph developers are well aware what Rete is, so we can safely use that term in the code - let's see what others think on this.
A point against incremental and nonincremental - they have the same postfix which makes it more difficult to search for them.
I think that refacoring of metamodels is much more important than the naming. We should separate the common base, the metamodel for the non-incremental plan and for the incremental plan, i.e. create 3 xcore
files.
An other point is that rete is an implementation detail (which is unlikely to change, though), so I'd avoid to call it as such.
For the naming: as also the operators in the incremental plan encapsulates graph relational algebraic operators, I'd propose the following:
relalg
for the common base, optionally factoring out data structures (variables) and expressions to a separate, parent modelqueryplan
(or qplan
) for the non-incremental plan, which is the traditional approachiplan
for the incremental planAs we are moving to Catalyst, we should get back to this issue and get the names right on this occassion :-).
Moving to Catalyst resolved this.
cc @szdavid92
We had a discussion with @imbur today and concluded that the current naming is quite confusing as the
search
model (built by theRelalgBuilder
class) does not define an actual search plan. We propose the following names:relalg
(formerly search) for non-incremental plans,rete
(formerly rete) for incremental plans,queryplan
(formerly relalg) for the common model ofrelalg
andrete
models (defined in thequeryplan.xcore
file).@jmarton what do you think?