using the tuts movies.json, imported into melisearch, it took almost 400MB

update some updates with only a single int field, it took almost 500MB space. this is amazing.

in elasticsearch, import the movies.json only took 35M space. and after bulk updates to a single int filed, it took 42M.
that's 7MB delta.

then we can reclaim the space with the segments merge op:

POST /movies/_forcemerge
POST /movies/_refresh
GET _cat/segments/movies/?v

after this, the space come back again.

in meilisearch, how can we do this?

that's to say, on a high load machine, we have updates to documents all the time with a high qps,

the meilisearch space will increased very soon. while the real data does not (like we are update some counters)

I know meilisearch uses lmdb. and lmdb does allow to reuse space. but, if it can reuse the claimed space, while an simple update to a int field cause the space continued to increase ? not reuse the old space ? because the update is a delete op + a full fields index op, so it should can reuse the space. but in my tests, it does not