: You know ... i keep hearing about flamenco, one of these days i should
: really take it for test drive and see what's under the hood.
i still haven't looked under the hood, but i did puruse the owners
manual...
http://flamenco.berkeley.edu/data.html
...flamenco requires that all of your facets and all of your data be
imported at teh same time. and while it's not neccessary for you to
specify that an documents mpas to each level of a hierarchyical facet term
(ie: you can say this doc maps to "Boston" you don't have to say it maps
to "US>MA>Boston") you do have to specify that Boston is a child of MA
which is a child of US when loading the data -- it doesn't look like there
is anyway to redefine the hierarchy on the fly, so it ammounts to the same
amount of precomputed information as the approach Yonik described earlier.
>From what I can tell in the online Demos, Flamenco also doesn't seem to
attempt anything special regarding weighting differnet depths of the
hierarchy based on the current set of results -- for each hierarchical
facet, the nodes at the highest level that have matches are displayed.
(NOTE: i'm not sure if that's a limitation of the system, or an attempt at
making the UI consistent) ... ths is the area where i think hierarchical
facets are particulararly tricky, programaticly guess what the most
usefull subset of the hierarchy to display would be based on the
constraints the user has already applied.
>From what I can tell: functionality could be added to Solr to do
everything Flamenco does right now and still support updates -- all that's
really needed is something to maintain the Facet configurations (order of
fields, pretty lables, and Hierarchy of terms) in memory. Hell, if
someone was so inclined, they could write a "flamenco importer" to parse
the .tsv files Flamenco currently uses.
-Hoss
