I'd be curious to know the conclusion on this, it sounds like a scalability issue in the parser. I've seen similar behavior in our application, but I've not had time to isolate it further. We're still stuck on Groovy 2 because we're unable to compile on Groovy 3 because I haven't been able to give the parser enough RAM or CPU time. A few hundred MB is OK to compile in Groovy 2 but even with multi-GB heap sizes result in out of memory exception in Groovy 3 and minutes long compile times.
Jason Winnebeck Sensitivity: Internal From: Stephen Mallette <spmalle...@gmail.com> Sent: Friday, October 16, 2020 2:58 PM To: users@groovy.apache.org Subject: groovysh and long input CAUTION: This email originated from outside of the organization. Do not click links or open attachments unless you recognize the sender and know the content is safe. Hello, I happened to be testing some issues related to this pull request I submitted the other day: https://github.com/apache/groovy/pull/1405<https://nam02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fapache%2Fgroovy%2Fpull%2F1405&data=04%7C01%7Cjason.winnebeck%40windstream.com%7C297adea240524d1b68f708d8720583ef%7C2567b4c1b0ed40f5aee358d7c5f3e2b2%7C0%7C1%7C637384715293498345%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000&sdata=%2B4jW23Rl4FZj60uHofA0hq%2FeX4a2yc9Nu1yUltxFQ3U%3D&reserved=0> In testing between groovy 2.5.x and 3.x I've noticed a considerable performance difference when pasting long multi-lined scripts. 2.5.x is quite quick to accept the input whereas 3.x seems to slow more and more as additional lines are consumed. While I did more complex tests trying to get to the bottom of the problem, it's fairly easy to recreate with a simple addition script: 1 + 10 + 1 + 10 + 1 + 10 + 1 + 10 + 1 + 10 + ... repeated a bunch of times and then ended with 1 copy that out of your text editor and just paste it into both versions of groovsh and see the difference. I've found 500 lines demonstrates it well but you could probably get away with less and still see the difference of evaluation at play. I tried to do some profiling to try to isolate the problem myself, but things got a little too deep into antlr4 for me and I got lost. Does anyone have any ideas as to what might be different for 3.x?
