If the goal is to allow for a very large shared strings hashtable, perhaps a lighter weight option like MapDB [1][2] is a better fit? It's JAR is only 700K, and it looks quite reasonably featured (has an implementation of HashMap) and flexible in its configuration for trading memory/performance/scalability, including mmapped files and some level of on-disk encryption [3].
[1] https://jankotek.gitbooks.io/mapdb/content/ [2] https://github.com/jankotek/mapdb/ [3] https://jankotek.gitbooks.io/mapdb/content/format/ On Fri, Jan 26, 2018 at 1:23 PM Dave Fisher <dave2w...@comcast.net> wrote: > Hi - > > We need to be really careful not to make OOXML deployments larger. > > Why is H2, a database engine, being considered? > > Regards, > Dave > > Sent from my iPhone > > > On Jan 26, 2018, at 11:55 AM, Alain FAGOT BÉAREZ <abea...@for-scala.it> > wrote: > > > > You might have heard of Hypersonic SQL, some 10 years ago. H2 is the > second version thereof. > > > > Gesendet mit BlueMail > > > > > > -------- Originale Nachricht -------- > > Von: Greg Woolsey <greg.wool...@gmail.com> > > Gesendet: Fri Jan 26 16:42:44 GMT-03:00 2018 > > An: POI Developers List <dev@poi.apache.org> > > Betreff: Re: ***UNCHECKED*** RE: adding dependencies on h2 and mockito > > > > Total dependency size is important to my deployment, and probably others. > > I don't use SXSSF at all, and would not need/want the dependency (which > > I've never heard of in 20 years of database and Java development, which > is > > strange to me, but irrelevant). My preference is to make it optional, > even > > though it's more work to code. Default would be the current behavior, > > which works for almost everyone, apparently, and an option would be to > > enable this behavior and manage the package availability externally. > > > > I suppose one could manually exclude the package as well, if SXSSF isn't > > used at all, since Java wouldn't try to load the classes unless a class > > referencing them was loaded, but that behavior is always subject to > change > > and should not be relied upon. Plus I wouldn't want to impose that on > > existing users who don't need/want it. > > > >> On Fri, Jan 26, 2018 at 4:47 AM pj.fanning <fannin...@yahoo.com> wrote: > >> > >> I could make h2 a `provided` dependency in our poi-ooxml pom. > >> The use of h2 is opt-in in the new code in my PR but I'll need to > refactor > >> the code to allow our code not to throw ClassNotFoundException if the h2 > >> classes are not on the runtime classpath. This is do-able but my > concern is > >> that this is difficult to automate tests for (checking the code works > when > >> the h2 jar is available and when it is not). > >> > >> https://mvnrepository.com/artifact/com.h2database/h2 is very common > >> dependency, so my preference would be to have the explicit dependency > from > >> poi-ooxml on h2 - but I'll go with whatever the consensus is. > >> > >> > >> > >> -- > >> Sent from: > http://apache-poi.1045710.n5.nabble.com/POI-Dev-f2312866.html > >> > >> --------------------------------------------------------------------- > >> To unsubscribe, e-mail: dev-unsubscr...@poi.apache.org > >> For additional commands, e-mail: dev-h...@poi.apache.org > >> > >> > > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@poi.apache.org > For additional commands, e-mail: dev-h...@poi.apache.org > >
