Thanks for testing! By the way, I tried profiling the ruta-core test run to see why version 3 was so much faster than version 2. I couldn't find an easy way to "loop" all the tests (I'm surprised the Eclipse Junit support doesn't include something supporting this!). So my observations are probably not very accurate. What I did see was lots (double) time spent in things like "file exists", and others relating to class loading. I'm thinking the root cause of the improvement is that Version 3 only tries to load 1 class per type, whereas version 2 deals with both XXX and XXX_Type classes.
So this performance delta is probably not "real" in terms of longer running code, once getting past initialization... (But there probably is some performance improvement, I hope! More below, inline. On 1/27/2017 5:38 AM, Peter Klügl wrote: > installed and tested update site in Eclipse Neon.2 - OK > > > The update site is actually > > http://home.apache.org/~schor/uima-release-candidates/uv3-alpha-rc4/eclipse-update-site/uimaj/ Correct. - sorry! > Mind the uimaj at the end > > I thought to remember that there was some mail when the migration from > people to home happened that we should not host release or staged > artefacts there but use the https://dist.apache.org/repos/dist/dev space. I'll take a look, but I don't remember anything specific, other than it was a suggestion. I don't particularly like using SVN (never gets deleted) storage for multiple release candidates - seems a waste... I'll look around for emails re: this topic. -Marshall > Peter > > > Am 27.01.2017 um 09:58 schrieb Peter Klügl: >> First part of review >> >> >> compared svn-tag with source-release - OK >> - the usuale diff >> mvn clean install svn-tag - OK >> mvn clean install source-release - OK >> built modified ruta with staged artefacts - OK >> - there are several problems which need to be fixed on ruta's side >> >> >> Peter >> >> >> Am 23.01.2017 um 22:05 schrieb Marshall Schor: >>> Hi, >>> >>> Here's RC4. The changes from RC3 are: >>> - fixing hashcode impl in one of the new semi-built-ins >>> - making the 3 new built-in types not built-in, but "semi" built-in, for >>> better compatibility with V2 (including uimaC++) >>> - changing the version from 3.0.0 to 3.0.0-alpha >>> - misc updates >>> ----------- >>> >>> If you happen to use JCas in your pipelines, and have JCas classes other >>> than >>> the standard built-in ones, these classes will need to be regenerated or >>> migrated. If you use the Maven JCasGen plugin to generate the JCas classes >>> as >>> part of your project's build, your good - nothing special need be done. >>> >>> There is a migration tool, and there are scripts to run it, as well as >>> Eclipse >>> launch configurations to run it. The instructions for running this are in >>> a new >>> documentation "book", part of the binary distribution (html and pdf) - it's >>> called the Uima V3 User's Guide. >>> >>> The guide also has an overview section which lists the new features of this >>> release. Many of these need more test cases written for them; I'm happy to >>> mentor any volunteers who want to improve the testing . >>> >>> The list of changes in Jira: >>> >>> https://issues.apache.org/jira/issues/?jql=project%20%3D%20UIMA%20AND%20fixVersion%20%3D%203.0.0SDK-alpha >>> >>> The source and binary zip/tars and the Eclipse update site are staged to >>> >>> http://home.apache.org/~schor/uima-release-candidates/uv3-alpha-rc4/ >>> <http://home.apache.org/%7Eschor/uima-release-candidates/uv3-alpha-rc4/> >>> >>> The Maven artifacts are here: >>> https://repository.apache.org/content/repositories/orgapacheuima-1119/ >>> >>> The SVN tags are here: >>> https://svn.apache.org/repos/asf/uima/uimaj/tags/uimaj-3.0.0-alpha/ >>> >>> and for the Eclipse Update Site: >>> >>> http://home.apache.org/~schor/uima-release-candidates/uv3-alpha-rc4/eclipse-update-site/ >>> <http://home.apache.org/%7Eschor/uima-release-candidates/uv3-alpha-rc4/eclipse-update-site/> >>> >>> Please vote on release: >>> >>> [ ] +1 OK to release >>> [ ] 0 Don't care >>> [ ] -1 Not OK to release, because ... >>> >>> Thanks. >>> >>> -Marshall >>> >
