Speed up of superset

[oktaal]

For example by indexing lemma for superset query. Ideally "instantaneously"

[bbonf]

I'm experimenting with BaseX locally, and it seems like
/alpino_ds/node[..//node[@lemma="staan" and @pt="ww"]] uses no index while
//node[@lemma="staan" and @pt="ww"]/ancestor::alpino_ds does use the attribute index

[bbonf]

I'm seeing x10 improvement (locally) on the entire Europarl:

Running query: /treebank/alpino_ds/node[..//node[@lemma="wet"] and ..//node[@lemma="staan"]]
results: 1195
time:  17.174638271331787 seconds
Running query: //node[@lemma="wet"]/ancestor::alpino_ds/node[..//node[@lemma="staan"]]
results: 1195
time:  1.6503210067749023 seconds

@tijmenbaarda I was thinking another thing that could have gone wrong here is that by default we save the cached results inside the django source folder, which iirc is NFS mounted and probably isn't meant for fast read/write...