oracle / truffleruby

A high performance implementation of the Ruby programming language, built on GraalVM.
https://www.graalvm.org/ruby/
Other
3k stars 183 forks source link

Documentation, TruffleRuby, JRuby and oldschool GUI elements - a meta-issue (but please do feel free to close it at any moment in time) #2844

Closed rubyFeedback closed 1 year ago

rubyFeedback commented 1 year ago

Hey there truffleruby folks and everyone else who may read this.

I am not a programmer by trade; my background is actually molecular biology, but I did write quite a bit of code over the years, e. g. via the "bioinformatics" bridge. Even then I consider myself more a biologist than a hacker/programmer. Anyway, that is just a bit of background.

I discovered ruby many years ago and it's a really great language (if we ignore some oddities).

One of my main projects is bioroebe: https://rubygems.org/gems/bioroebe. It is not hugely elegant or epic, more a hobby project where I try to cover all "reasonable" use cases for bioinformatics and molecular biology, if that is possible. For instance, translating all open reading frames in a DNA or RNA sequence into a aminoacid sequence, on all six reading frames (three on one DNA strand, three on the other; three because one codon is read as a contig of three nucleotides, at the ribosome).

One goal or objective that I have for this project is that I want to offer GUI elements for the user. The natural choice is www, that is, to make these available in, say, .cgi pages, perhaps .html, sinatra, rails - wherever we can have a browser, things should work. But, aside from this, I also like classical GUIs. I spent most time with ruby-gtk3 actually. And I wrote quite a lot of widgets here.

ruby-gtk3 does not work well on windows.

A few years ago I started to learn (or rather dabble in) Java. I made quite a lot of progress in the last 2 years in this regard. I am still kind of a newbie, but getting better.

About that time as well I discovered GraalVM and kind of switched to it. So I don't even use openjdk anymore - graalvm is the way for me to go:

openjdk 19.0.1 2022-10-18
OpenJDK Runtime Environment GraalVM CE 22.3.0 (build 19.0.1+10-jvmci-22.3-b08)
OpenJDK 64-Bit Server VM GraalVM CE 22.3.0 (build 19.0.1+10-jvmci-22.3-b08, mixed mode, sharing)

GraalVM works very well. I tested it on windows too and it works.

I even got statically compiled binaries to work on linux, via GraalVM. \o/

That one is REALLY great - and fast. It does not yet work for me on windows, but hopefully it'll be possible one day.

Anyway. Since I want to be as platform independent as possible, my code should work on windows too, in particular the GUI stuff.

So, since I wrote some ruby-gtk3 applications, I recently decided to port them to java SWING too, via jruby.

Today I "finished" (kind of) the alignment widget, that is, where different sequences can be aligned and then the consensus sequence is calculated and displayed.

The ruby-gtk3 widget looks like this:

https://i.imgur.com/SC02Al1.png

The jruby java swing variant I wrote just now looks like that:

https://i.imgur.com/oHtDQCA.png

So, don't worry that these don't look that pretty. I just wanted to get the functionality to work; I will improve this visually at a later point in time.

For me the big win here is that this will most certainly work on windows too. I already tested this yesterday via jruby for another test widget. Hopefully at a later time I can extend it. Anyone knows libui and libui-ng? And Andy's glimmer GUI? My dream is to build some kind of IDE for bioinformatics-related activities, and have it "translate" to different toolkits, including the www. I also intend to make use of JavaFX and SWT - Andy uses SWT a lot, you can look at his glimmer-specific GUI for these parts (https://andymaleh.blogspot.com/ and perhaps nebula too https://github.com/AndyObtiva/glimmer-cw-nebula). Anyway.

You may now wonder "but what does this have to do with truffleruby".

Well. One idea I have is, evidently, to translate the ruby-specific code into a compiled variant via graalvm+truffleruby one day. I already tested various commandline applications in graalvm (plain java) and these work, both on windows and on linux. So that part is ok-ish.

I could not get swing-applications to work on graalvm yet, but other folks reported that it is possible, e. g. we have to use some inflection (see this one 2 years ago https://www.praj.in/posts/2021/compiling-swing-apps-ahead-of-time/ he got it to work, but I don't know how he did it, it's above my current abilities to understand). I am still a noob so I will wait until others smoothen the path for me to follow. :)

Alright - that's the lengthy introduction and background. With that use case explained, I would now like to make a few items. Perhaps you guys could consider some of these use cases in the long run, say, 2023, 2024, 2025 - just that one day this could perhaps work, or be improved. And as stated above please feel free to close the issue at any moment in time - I totally understand keeping the issue tracker clean. See how Jeremy works in regards to sequel.

(1) JRuby versus Truffleruby dichotomy/split:

truffleruby has some kind of history in regards to jruby. I don't want to go back in history, and there is cooperation between the projects, also in regards to ruby specification; and change takes time, I get it. Consensus has sometimes be reached. That takes time too.

BUT!

From a user's perspective, ideally I'd love to have "just one ruby-java thingy". Note that I mean that we should RETAIN the flexibility. So jruby's standalone flexibility is a GOOD thing. Truffleruby+graalvm promising compilation and integration into a larger framework is ALSO a good thing. I do not mean this to LOSE an advantage. The advantages should be retained.

What I am saying is that, hopefully, one day, rather than 500 different ruby implementations, we could have something to "use just this one only" and it would have all the advantages and none of the disadvantages.

The ruby specification is also useful for other projects, single-hobbyist ones such as in natalie. While the author of natalie says he just wants to have fun, having a correct and working ruby implementation via tests + specs is a good thing. It may allow more integration with other projects too. See also matz' work in regards to mruby (unfortunately I don't know C and C++ so I am more in the java niche/ecosystem now).

(2) Truffleruby + GraalVM + compilation. I have had problems here. For me it does not work well. I ran out of memory and then the resulting binary did not work well. This may be graalvm's fault, but at that point I'd just prefer for things to work. Failure frustrates me a lot. I tend to move to other things when I run into too much failure - not a good mindset but it keeps my frustration level smaller when I work on things that I know I can get to work.

Part of this issues is in regards to documentation and examples. I'd love for there to be more examples, in particular via GUIs. Tutorials and such things. Or, in my case, I'd love tutorials that would start from hello world, and then expand the use case to more complicated stuff, including use cases on the www but also via GUIs (all of swing, JavaFX, java-SWT - I'd love to have specific examples. These examples can be somewhat simple, like a button, an input field - but I would like to know how I can integrate this into graalvm and the resulting binary so that it works. For some reason, whatever I am doing right now, it does not work well. GraalVM did improve quite a bit in the last +1 year or so, so some of the older problems I had go away, but the issue of documentation is still kind of a problem, IMO. I'll check every few months, so things may improve slowl, but that was one pain point I had here so far).

Anyway. Me writing more jruby + SWING will take some days, weeks, so that'll take some time - hopefully at the end of 2023 things will have improved a lot here. It would be kind of cool if we could write GUIs in ruby (sort of), with java a a "back-end", and, at the least on the linux platform, even be able to compile these things statically. That would be really great - the speed benefit I could see with graalvm is AWESOME. I love ruby, but speed-wise graalvm is like a ferrari.

eregon commented 1 year ago

I'll turn this into a discussion, it seems more appropriate for this, and I'll reply there.