Taken from proposals (both V relational DB and V graph DB) - lambda-land/VDBMS GitHub Wiki
Future ideas:
we will represent the invalid values by "-" which separates them from null values and relieves us from the null-padding problem. As future work, we plan to make tuples share their data as long as the shared data is consistent with the v-relation schema. e.g.: tuples: (A1,A2^A,A3^B,pc)
x,y,-,A
x,-,z,B
will be joined in one tuple:
x,y,z,A or B
To dos:
"variant consistency property: every variant of a vquery is consistent with the corresponding variant of the vschema. We'll enforce this property statically."
"we will prove that our alg enforces the variant consistency property."
"we will define correctness properties in terms of variant consistency and variation preservation for queries and other functions over variational data."
"we will prove that these properties are satisfied by our formal semantics and are preserved by any optimizations that we perform."
Notes borrowed and elaborated or need to look into:
"dimensions can have relationships", which is captured by the feature model, i.e., the propositional formula attached to the v-schema.
"choices can occur anywhere and at any level of granularity with a vquery".
"Graph dbs inherently support structural heterogeneity but can still benefit from explicit support for variation in a variety of ways. E.g., variation can be used to track data provenance when integrating data from a variety of sources, or to categorize nodes and relationship that are relevant in different contexts."
"Existing graph dbs already allow annotating nodes and edges by properties, which can be used to simulate some applications of v-graphs. However, v-graphs have two key advantages: 1) v-graphs are more expressive since properties themselves can be made variational and 2) the variation in a v-graph can be referenced in variational queries over its data, which provides a more systematic and efficient way of working with data variation. "
question for Eric: so variations can only occur in properties? it cannot occur in the existence of nodes or edges? before asking Eric look into the variational graph paper!! According to the proposal, the existence of nodes and edges are also variational.
look into graph query langs!
The commuting diagram below describes the variation preservation property and it can be generalized to any variational function or operation over variational data in order to specify and evaluate its correctness.
d --q_v_c--> r
^ ^
| | ._c ._c
| |
d_v----q_v----> r_v
The second motivation for making variation explicit is to make queries over data variations more efficient. There are two primary ways that explicit variation can be exploited to speed up queries. The first is by pre-emptively filtering data based on the variation expressed in the query. The second is by combining many separate queries into a single query that can still be efficiently executed. We plan to apply such optimizations.