Only pass subgraph into graph_op

[MartinKl]

For complex corpora a lot of workflows could be simplified if there was a mechanism in Annatto to only pass a subgraph into a graph op, but continue with the entire graph after it.

Right now, modules either use discrete low-level operations on the graph or go through the more high-level AQL interface. But sometimes it's desirable to perform low-level operations on a subgraph that can easily be defined in terms of higher-level means, such as an AQL query.

Assume the case of parallel corpora. The might have an internal structure like this (example is English and German):

corpus_____________________
|                          \
subcorpus EN     subcorpus DE
|                           |
doc-1                   doc-1

Documents with the same number represent the same underlying document, but in its specific language (according to the subcorpus). Now, one might want to modify all annotation namespaces in the English documents to be "en" and in the German subcorpus to be "de". Modifying namespaces is a low-level operation. But only by higher-level means, we could (currently) select the nodes whose annotation namespaces should be changed, such as querying for all nodes that are part of DE via node @* node_name=/.+DE/.

[MartinKl]

I am afraid to do this efficiently, graphANNIS would have to provide some kind of view feature for graphs, so that the subgraph is not a separate object in memory which requires a lot of copying.

[MartinKl]

The easiest way would be to rely on own implementations of edge containers. E. g., analogous to UnionEdgeContainer which only holds a reference to the actual storages, one could implement sth like an IntersectionEdgeContainer, which contains edges that occur in all defining edge containers.