elasticsearch-sudachi icon indicating copy to clipboard operation
elasticsearch-sudachi copied to clipboard

Published 20 hours ago • verified publisher icon WorksApplications

Automatic reloading of the user dic

Open HIRANO-Satoshi opened this issue 5 years ago • 5 comments

There is a nice plugin which syncs config files such as user dic among ES instances.

Is it possible to add an automatic reloading feature when the user dic is updated?

Without such functionality we need to open/close all indexes that use the dic on all ES instances, or restart all ES instances. That is a burden.

HIRANO-Satoshi avatar Feb 13 '20 12:02 HIRANO-Satoshi

Transferred to elasticsearch plugin repo. WIll try to implement this in some capacity. Dictionary reload will not be supported for binary dictionaries of plain sudachi (but may be supported for auto-compiled csv-based dictionaries).

eiennohito avatar Dec 24 '21 01:12 eiennohito

After a consideration there is a problem: what to do with old documents which were analyzed by a different dictionary. When reloading a dictionary, some documents can produce different token streams and will not be searchable anymore.

Possible actions:

  1. Do nothing (easiest implementation)
  2. Reindex all documents (this will be pretty difficult to implement atomically)

Any ideas on possible behavior?

eiennohito avatar Aug 01 '22 01:08 eiennohito