# Fix bug

I think I get it: normalization is based on sample while values are based on the full corpus because they're taken from the cache.


# For herarchical block map:

- update term counts when filtering (comes from corpus.ter_docs)
  - even better if we could update to selected domain

- show total topic usage

- remove hierarchical block code

- show some kind of +/- intensity bar for positive and negative characteristic blocks

- for a domain show temporal profile of topics?

- zoom on selected

- check for col_title (crashes visualisations if unset)

- for a domain show network

- color with nested measures

# For block tables
- tables for 'ext' also show terms
