I'm certainly fine with any of that. It would be good to get some historical perspective and buy-in from say Hoss or Yonik (CC'ed). Do you guys have any opinion on this? The introduction of the overwrite parameter isn't apparent in CHANGES.txt. It seems to be https://issues.apache.org/jira/browse/SOLR-60 assigned to no version; but that issue is more of a refactoring of two other parameters overwriteCommitted & overwritePending. So in summary there doesn't seem to be any historical issues in JIRA with conversation that can help inform why the feature ought to exist.
On Tue, Sep 18, 2018 at 8:17 AM Jan Høydahl <[email protected]> wrote: > Seems like the overwrite=false flag is purely for better performance > whenever you KNOW that you are indexing docs that have not been indexed > before. So it would be easy to deprecate the feature and let it always be > true if that plays better with updateLog? > > -- > Jan Høydahl, search solution architect > Cominvent AS - www.cominvent.com > > 18. sep. 2018 kl. 14:02 skrev David Smiley <[email protected]>: > > Do you mean write new code for adding a feature for _version_ to behave > like overwrite=false? I suppose anything's possible with new code, though > I'm not sure if it fits semantically since the result of overwrite=false is > the possibility of a duplicated document -- something otherwise impossible. > > > I don't really like the very existence of overwrite=false and don't want > us to take on the burdens of supporting it in dubious cases, like having an > updateLog. > > On Tue, Sep 18, 2018 at 5:09 AM Jan Høydahl <[email protected]> wrote: > >> Can't the overwrite=false logic be replaced with some logic using the >> _version_ support, which would also support updateLog? I know you can >> provide a specific version in the update and it will not update if the >> index already contains a newer version, but I'm not sure if there is a >> _version_==null kind of feature? >> >> -- >> Jan Høydahl, search solution architect >> Cominvent AS - www.cominvent.com >> >> 18. sep. 2018 kl. 06:20 skrev David Smiley <[email protected]>: >> >> Is <add overwrite=false> supported when there is an UpdateLog? Perhaps >> only when you are darned sure the doc is in fact uniuqe? Maybe we should >> throw an exception if you even try at all with an UpdateLog? >> >> Context: I'm working with Moshe, a contributor, on doc updates for >> nested docs. It's all very much WIP with TODOs but nonetheless >> ConvertedLegacyTest failed due to some oddity. It turns out this ancient >> test deliberately adds docs with overwrite=false that already exist by the >> same ID. Youch! This test very much pre-dated the UpdateLog, but at some >> point the UpdateLog ended up in the default config thus this test ended up >> using it. It *appears* to be accidental luck that the test hasn't been >> failing -- i.e. I doubt this situation above was deliberately made to >> work. We're fiddling with some related code and now it fails. >> >> I'm inclined to think the test is broken by virtue of it using a config >> with an UpdateLog -- something it wasn't originally designed for. >> -- >> Lucene/Solr Search Committer, Consultant, Developer, Author, Speaker >> LinkedIn: http://linkedin.com/in/davidwsmiley | Book: >> http://www.solrenterprisesearchserver.com >> >> >> -- > Lucene/Solr Search Committer, Consultant, Developer, Author, Speaker > LinkedIn: http://linkedin.com/in/davidwsmiley | Book: > http://www.solrenterprisesearchserver.com > > > -- Lucene/Solr Search Committer, Consultant, Developer, Author, Speaker LinkedIn: http://linkedin.com/in/davidwsmiley | Book: http://www.solrenterprisesearchserver.com
