[jira] [Commented] (FINERACT-1127) Modularize Fineract to allow something like e.g. the Pentaho integration to be built and run separately

["Michael Vorburger (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FINERACT-1127?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17194761#comment-17194761 ] 

Michael Vorburger commented on FINERACT-1127:
---------------------------------------------

[~ptuomola] Yes. Agreed. That would actually be pretty cool, IMHO. It's not quite what I originally had in mind and described above, but as a a first step, I love it. And I do think this could add a lot of real practical of value to this project... because this would, IMHO, be, much, better than that "forked" solution that is currently being promoted, if you've followed various related confusing posts on list. That solution will keep requiring someone to rebase and release a separate project that is de-facto a Fineract fork. This approach would avoid that.

But there are some minor practical technical details we would have to figure out together here, no? For one, there currently is no easy way to have an entirely separate project, imagine say e.g. a [https://github.com/ptuomola/fineract-reporting-pentaho-plugin], be "built against" what {{[PentahoReportingProcessServiceImpl|https://github.com/apache/fineract/pull/1262/files#diff-0708dfc4fbb4b67f43b9b935c8bb1ad8]}} imports, do you see what I mean? We can't use {{build/libs/fineract-provider.jar/.war}} (I trust you'll understand and agree with why). The most pragmatic and simple initial quick solution may be for such a {{fineract-reporting-pentaho-plugin}} project to use just {{../fineract/build/classes/java/main/}}?! According to [https://docs.gradle.org/current/userguide/declaring_dependencies.html], that may be possible using something like \{{dependencies { implementation files("../...") }}} ? It's a bit of a hack, of course, and somewhat brittle, but perhaps good enough to get this off the ground? Later in the future perhaps a "stable plugin API", published as an "official artifact", could be better.

If you want to have a go at this some time, by all means please do! Otherwise I may be motivated to try to do this on some rainy autumn weekend in the coming weeks (or months) - I'll add this fairly high up on my personal Fineract TODO... but I would love for you to get to it, first!

> Modularize Fineract to allow something like e.g. the Pentaho integration to be built and run separately
> -------------------------------------------------------------------------------------------------------
>
>                 Key: FINERACT-1127
>                 URL: https://issues.apache.org/jira/browse/FINERACT-1127
>             Project: Apache Fineract
>          Issue Type: New Feature
>            Reporter: Michael Vorburger
>            Priority: Major
>
> [~francisguchie] thanks for raising [https://github.com/apache/fineract/pull/1262/files] for FINERACT-1094, even though we cannot merge it due to licensing, it helps at least me to see it nicely isolated like that, e.g. for this discussion! 
> Seeing that, and subsequent FINERACT-1125, led me to wonder if an alternative future architecture approach here could be to have a 3rd-party, such as the Mifos Initiative, offer Pentaho integration for Fineract not anymore by forking Fineract and adding code to it (which is always a PITA to maintain, in the long run), but by building and releasing an entirely separate runnable binary artifact (WAR / JAR) for it... this could be very neat even from a runtime perspective - completely separate REST API, and (possibly "heavy"?) report generation?
> _One could even image breaking out running scheduler jobs separately from API serving. Ultimately, this could effectively be the start of breaking Fineract 1.x into a CN-like "microservices" deployment... but I'd envision that to, always, just be one option, a possible alternative - with the current WAR/JAR remaining as is, forever - but it would just "assemble modules" for the "monolithic deployment distribution". Anything along these ideas is further down the line, but starting with making this possible for reporting could be a pragmatic start.... let's focus on that only, in this issue._
> This is more of a still somewhat vague idea at this stage. The next step would consist of spending more time understanding the details of how Fineract's {{ReportingProcessService}} is designed... I personally don't know that much about it, but looking at PR #1162, one can kind of gather that this {{PentahoReportingProcessServiceImpl}} needs to implement {{ReportingProcessService}}? What would it take for the front-end to be able to call another service (or Fineract to HTTP redirect reports to an external service...), and then for such an external service to be able to... do exactly what, actually - what's the lowest common denominator integration touch point, here? From an only very quick glance at the code, it looks like we're basically "passing through" JDBC connection details? So...  what one would need is to be able to build an external Fineract (non-CN) service that can access the same DB? Sharing a minimal amount of {{fineract.infrastructure.core}} code, as a library...
> [~francisguchie], [~ptuomola], [~awasum], [~xurror], [~edcable] FYI.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)