web search done right
Commits on main
concurrent reduce and map. Mapreduce doesn't use async anymore
7db1d92 Mikkel Denker committed 4 years ago
mapreduce simple wire protocol to handle cases where job or result is greater than buf size
919c3c5 Mikkel Denker committed 4 years ago
less locking+cloning in webgraph. Default trait for webgraph store
2db70f2 Mikkel Denker committed 4 years ago
significant speedup in webgraph by inserting into memory and only flushing to disk when needed
abcada0 Mikkel Denker committed 4 years ago
fixed case where mapreduce worker panic did not register as failure at manager if 0 is a valid answer to the job
22f74e2 Mikkel Denker committed 4 years ago
Custom tokenizer that stems multiple languages and fixed snippet generation as a result of the changed tokenizer
c8b51db Mikkel Denker committed 4 years ago
Tokenizer can produce multiple tokens for each term. Should search for stemmed and non-stemmed version.
11953e8 Mikkel Denker committed 4 years ago