Taher,Right, it was not a problem of space but licence. With the Gradle solution we no longer have licence issues with external jars, another good point to this solution :)
Jacques Le 22/06/2016 à 10:46, Taher Alkhateeb a écrit :
Hi Jacques, How about we actually download all JDBC drivers automatically regardless of the selected database. Their size is very small and it is worth making the build script smaller no? And also anyway some people use different versions of the database and have to change the driver version, in this case it would be a much cleaner solution for them to go to the build script and change the version (simple string) than to actually download the jar file or modify the logic for the download tasks. With that being said, thank you all for your input, I will draft the first patch based on your feedback very soon. I will try to consolidate your opinions as much as possible but may differ. So please take it lightly as everything can be changed upon your request, I'm just trying to balance all opinions and come up with something nice and clean. Again, I really appreciate the support! I will update you soon. On Wed, Jun 22, 2016 at 11:38 AM, Jacques Le Roux < [email protected]> wrote:Le 22/06/2016 à 10:05, Julien NICOLAS a écrit :On 22/06/2016 09:53, Julien NICOLAS wrote:On 21/06/2016 22:09, Taher Alkhateeb wrote:- download-PG-JDBCIf it's possible, keep this one :)Ok, I don't see the information of Michael that with Graddle, we don't need a task for that because of the Graddle dependency functionality. So my mistake, forgot it :)I think you are right about this one and I believe we should keep all Ant JDBC driver download targets. Except if Gradle is able to infer that it would be possible in certain circumstances an user could need one of those drivers. This would need to parse entityengine.xml and I highly doubt this can be done w/o some human intervention. On the other hand if similar Gradle tasks are introduced, then of course I'd not see any reasons to not drop them. Same for all Ant targets actually. Hope I'm clear enough :) JacquesSecond Question ----------------------- it seems many of the load tasks are too specific. So I suggest to only implement loadDemo and the rest are executed manually by users, for example: ./gradlew 'ofbiz --load-data reader=seed, seed-initial, ext' instead of load-extseed.I think load-seed is important as well, so if you can keep the load-seed task, it could be fine. Thanks! Julien.If you would like to add the other load data tasks, please specify which ones. Appreciate your early responses. Taher Alkhateeb On Tue, Jun 21, 2016 at 1:02 PM, Taher Alkhateeb < [email protected]wrote: Hi Everyone, Thank you all for your support and kinds words. This is truly a wonderful atmosphere and I am lucky, honoured, and privileged to work with you all on this project. My patch is almost done, but definitely there is a lot of work to be done which includes the following: - I have one failing test out of 889 that I need to dig through, maybe you guy can help - I want to change / delete / add some tasks - Documentation needs to be updated in multiple areas - Testing, testing, testing, testing, testing, testing, testing, testing, testing So the plan of action is as follows: - I will continue the discussion on this thread for a few questions that I need an answer for. - I will issue a JIRA to hold the patch and everything else Please consider helping, this is something that definitely needs a team, more than one brain! If you are working on something not urgent, please consider dropping it for a while and jump along for help. I will post another email soon with the JIRA details and list of questions I need answer for. Again, thank you, you guys rock, I love OFBiz and this community! Regards, Taher Alkhateeb On Tue, Jun 21, 2016 at 12:49 PM, Nicolas Malin < [email protected]> wrote: I'm in over for these technical aspects but the motivation andenthusiasm for many PMC and commiter tells me that seems a good way. So now I will learn gradle ;) and I'm in favor to realize this change directly on trunk Thks Taher to your engine energy on this subject ! Nicolas Le 21/06/2016 10:43, Jacques Le Roux a écrit : As Gavin mentioned, Gradle can run Ant so no worries using only Gradlehttps://docs.gradle.org/current/userguide/ant.html Jacques Le 21/06/2016 à 09:59, Michael Brohl a écrit : I have no strong opinion for/against Gradle (I simply have noexperience with it) but I agree that it should be either Ant or Gradle. Running two build tools in parallel would make it too complex an gain nothing. I'm in favor for learning new things so Gradle sounds fine for me :-) Regards, Michael Am 21.06.16 um 08:11 schrieb Taher Alkhateeb: Hi Deepak,Ant would be removed completely for the following reasons: - First to resolve the ASF issue about the libraries mentioned by Sharan below without expending effort on both build systems. - Ant is an obstacle to refactoring the framework. If we keep both systems side by side we gain nothing, actually we lose value because the builds become more complex. For example, we will not be able to intrduce the unit tests, and we will have two build outputs, and we will have two ways of running the framework (java -jar ofbiz.jar and gradlew ofbiz) and we will have other incompatibility issues. With that being said, we will not make the switch before a thorough and full testing. That is why we ask everyone who is willing to please help us out to make this transition smooth by testing and providing feedback and comments. Taher Alkhateeb On Tuesday, 21 June 2016, Deepak Dixit < [email protected] wrote: +1 for Gradle.Are we going to remove ant from framework completely or planning to keep both ant and gradle? Thanks & Regards -- Deepak Dixit www.hotwaxsystems.com On Mon, Jun 20, 2016 at 6:20 PM, Sharan Foga < [email protected] <javascript:;>> wrote: Hi EveryoneThis is the second of two emails to inform the community about what has been happening around how we are planning to handle external dependencies in the trunk. Two weeks ago the community discussed and agreed to the use of Gradle to help us put together a unit test framework. While trying to get this set up while Ant remained as our build tool became very difficult.This was because our Ant scripts:- are massive and contain a lot of code - are complex - are very brittle and make it very hard to change things - have no dependency management - need everything to be declared We realised very quickly that the re-factoring issues and limitations we are facing are because of our build tool – Ant. Ant is verbose so it needs everything to be declared. We did a brief assessment of Maven and found it better than Ant but not a good fit for OFBiz because it has strict requirements for the convention-over-configuration rules to work. Instead we decided to take a closer look at Gradle. So why Gradle? As Taher was already looking at Gradle for unit testing, we decided to look at what we would need to do to totally replace Ant with Gradle. We received some great support and feedback from David, who is already using Gradle with Moqui. After some preliminary tests we found that Gradle has some very good features such as: - a much shorter code base (e.g. one single file of around 250 lines ofcode replaces all the build.xml files and thousands of lines ofcode) - Programming is DSL based and links in well with Groovy (e.g. the script is short because despite heavy custom requirements for OFBiz, two small functions took care of the complex directory structure) - It handles all the external jar files by downloading any dependenciesdirectly via internet- Jars can be upgraded by simply changing a string - It has matured a lot and has a high level of support in tools,IDEs, books, documentation - It also has a lot of plugins which means that it works with pretty much all build systems, supports multiple programming languages, and many other features (e.g. OSGi) We understand that it can help us make OFBiz more modular and also settingup a framework for managing addons would be a lot easier.So what's been done? Taher has been working very hard on a patch for the trunk that completely replaces Ant with Gradle. (Huge thanks to David for providing some examplescripts to help us get started!) The patch is now ready to beapplied to the trunk and includes the following: - java -jar ofbiz.jar is now replaced with -> gradlew 'ofbiz --whatever-options-here' - In addition to gradlew 'ofbiz' we also have gradlew 'ofbizDebug --whatever'. What does that mean? It means we can run debug on ALL ofbiz commands, not just start - If we decide to change the source directory structure in components say from /src to /src/main, it would literally be a change of 5 charactersin the build script- We can immediately move all jar files if we want to a unified location in /lib for example - We can delete most of the jars and declare them as dependencies saving space and resources - We can automate the creation of the .classpath file so when we update libraries no need to do this manually (under development) - We can ignore components that are not define in the xml files for loading (under development) - We can introduce unit tests with about 10 minutes of work We are finding that the flexibility and control we are getting with Gradleis truly amazing. We know that Gradle will be a major change tothe projectbut we think that it will significantly improve the project byremoving a lot of build complexity and take care of that essential dependency management that we need to comply with. Our next steps will be to apply the patch to the trunk and then continue the re-factoring work. We will need to organise some knowledge transfer sothat all our committers understand what the changes are and howthey wouldneed to work in the future.The PMC are very, very excited about having Gradle as part of future of OFBiz and we hope that the community will think so too. As always, feedbackwelcome.Thanks Sharan
