Sunday, August 23, 2009

OpenERP vs Adempiere, an open letter to Adempiere experts

Hi, Thanks to my former French employer, I had the opportunity to study and compare quite deeply the main open source ERP's two years ago. I know some of them might be biased by huge financial interests and some others were simply developed without analyzing too much at the existing alternatives, or because alternatives were less apparent at that time (Tiny had closed SVN access and no mature web client yet for instance). But I know, and had even recent confirmations that at least the Adempiere community is made of a lot of independent and very skilled folks. That's why I'm very interested in knowing why they stick to Adempiere why I would myself rather stick with OpenERP. Here is an open letter to them:

UPDATE: wow, about at the exact same hour when I got interested in Adempiere again, Red1, Adempiere Malaysian rock star who was at the origin of the Compiere fork back to 2006, just released a fantastic book about the open source ERP's story. The book focus on the Compiere, Openbravo and even Ofbiz story, largely omitting OpenERP (cited only once). Still there are really really good critics about Compiere and Openbravo VC backed models; I share them at 100% and feel sorry for not having been able to be more explicit about Openbravo back in 2008. In my opinion what made Adempiere success largely applies to OpenERP (a powerful community), I even see some architectural decisive advantages in OpenERP, that's why I'm writing that post. So for folks interested about the genese of OpenSource ERP's, aside from my own study from 2007/2008 here, I really recommend you take the time read the Book from Red1, it's not only about ERP's, but also about open source thermodynamics of complex systems in general:

Hi Adempiere Community.


By the end of 2007, my former employer, asked me to do an extensive 6 months survey about emerging Open Source ERP (to be found here, sorry French was required ). I quickly identified Adempiere in the top 6. But I should say its integration by has been quickly discarded at that time because:
  • Smile was not so interested in pure community driven open source. In the opinion of Smile leaders, that model hardly ever worked at Smile, probably because they wanted to scale a lot making it difficult to rely solely on individual talents rather than being backed by a strong and reliable editor. I probably see myself more weight in those talents within the oss process because I think there have been an oss speculative bubble about to burst recently and for this successful VC baked oss projects (Alfresco?) remains the exception rather than the rule. Now compare this to talented leaders that led for instance: Linux, JBoss, Compiere before 2006, Rails, Ruby, JRuby, Python, OpenERP... This is yet an other story, but that's why strong editor backed yet community oriented oss projects like OpenERP (formerly TinyERP) and """promising""" Openbravo where favored at the time while unilaterally editor driven oss like Compiere or ERP5 were dropped at the end of the list.
  • by 2007, the path to a strong web interface was still uncertain for Adempiere given the investment to be made. On the contrary, Smile was really web grounded already.


Now I should say I have been totally disappointed by Openbravo since then. Despite the successive $6M + $12M fund raised by Openbravo, it's still very limited in term of usable built'in features and effective refactoring since their original obsolete architecture, not to speak about the lack of effective community (the vast majority of their forums is frequented by Java/oss newbies, apparently seduced by the massively funded commercial hype, but providing no valuable code/feature/help feedback). They might still have some cache to invest, let's see if their new architecture moves (Hibernate Data Access Layer, modules, new GUI which are all really at the proof of concept stage for now) brings them more community/business momentum, but I should say I'm not really optimistic about that, especially considering the competitive OpenERP global outbreak. On the contrary, OpenERP was for sure a great surprise both for us and our customers (see for instance one tesimony here ), allowing a sustainable business, even if it should still be considered as maturing (quickly).


2 years after that survey and more than 4 full SMB's OpenERP integrations later, I recently quit to start my own FOSS consulting company based in Brazil this time ( ). While working hard on the Brazilian OpenERP localization, I see that Adempiere is historically more heavily rooted in Brazil (especially because the community divorced from Compiere and because just like in lot's of developing countries, the usual SMB's proprietary ERP's from SAP, Sage and M$ didn't really attack this market yet, giving more room to oss outsiders). This led me to re-investigate Adempiere. My findings is that Adempiere is finally much stronger than Openbravo and possibly than Compiere. Indeed, Adempiere seems what more advanced than Openbravo in the process of escaping raw PL/SQL at the profit of a higher abstraction, Java code in your case. Moreover, Adempiere definitely has a strong community and it seems that with AdempiereGWT, a viable web based interface is finally emerging.


Now I'm sorry, I still see plenty of advantages in OpenERP over Adempiere. But some of you here will know Adempiere much better than me so that's why I ask you to defend Adempiere here so that I make a more accurate judgment. Don't take this post as an aggression, just like for OpenERP, I'm really aware of the passion behind the Adempiere project and I really hope you make the most of it, emulation is always good, and I'm simply curious about what I might have missed in it.

I think that it's just too hard to really compare the raw business features beteween OpenERP and Adempiere. I think they are roughly equivalent, at some place X will have a broader scope than Y, but Y will do the thing with less bug or in a more integrated manner, so it's really just to complex to say (while it's easy to say that both along with Compiere have a broader scope than Openbravo ERP for instance). I tend to think that OpenERP has a broader scope, especially in the service area while Adempiere might have less bugs eventually at its functional core.

So, instead of judging pure business functionality, I prefer evaluating more impartial criteria that will probably determine mid and long term business coverage. Would you agree that community dynamism and technology govern this in the long run?

About Community dynamism, I would say that both are again roughly equivalent. Both are FOSS led by very passionate people (both are farm grown, just discovered this!), and also backed by active third party integrators. I think that Adempiere is historically more grounded, but OpenERP community has been growing quicker those last 2 years.
At least Google Insight shows that, but I have other indicators showing also (forum topic counts, partners network, sucess stories...):
(IMHO you can consider Openbravo really overrated here because of all the money they smoke making commercial hype, when you count real third party integrations, figures are very different)

So the next factor is technology/architecture in my opinion. I think that:
- the web is the platform. An other client that the web client might be used too, but a very good web client is required to look appealing to the market
- technology should maximize code reuse bewteen third parties (=> modules repository) in order to boost synergies (eg systemic cost savings)
- technology should boost productivity
- technology should should decrease entry barrier (please no J2EE bloat for the bloat, comes a time where the community should contribute efficiently rather than spend time in compiling standards in committees, just like J2EE unfortunately ended. Bloat that is affordable for fortune 500 ERP's is not affordable for SMB's ERP's. It's not a university where you should make the system happy, real world effciciency is required.)technology should enable very broad customizations. We live in a globalized world where the survivors are the one who maintain an original business model, the others die or get merged. This means that past time where one black box ERP could rule them all is over. Even SAP approach, "we think about lots of standard business, pre-build the thousands of tables and then you need to hire a $ expert to customize it because it's so boring and if you want to develop outside behind the tables it would just cost you $ instead", is being beaten by the open API and open module system approach). Today's ERP should be highly customizable at SMB's costs.


Concerning the technology/architecture topic, I see lots of advantages of OpenERP over Adempiere. But that might just be because I don't know Adempiere enough, so please correct me and develop why you think Adempiere is better for you, I hold no OpenERP share and I'm very open.

1) Is it possible in Adempiere that a module EXTEND any core form (eg view) at any point, replacing, adding, removing form element, rather than redefining/copying the whole view or patching the original view directly in the source code which I consider less maintainable and harder to deploy? I saw few very frameworks featuring OOP at the view level...

2) Can modules extend views and business objects in cascade? In OpenEERP it's very common a customer run 30 modules, 10 being custom, and most of the 10 being only like 40 lines at most. This sytem, many small modules and large bundles of modules really makes the most of code reuse, just like OSGI for you Java folks.

3) Can a module extends a CallOut process (or Controller/Business Object layer if you prefer) and call 'super', like for instance: do the normal thing you should do when I click here and then bring me back to my custom code which do exactly and only what I need. Might be a dumb question, but that still not possible in Openbravo. From my day to day OpenERP work, that would mean way more time, patching/duplicating large piece of codes rather than acting with a surgery scalpel.

4) Is anything you do in an Adempiere client, auto-magically natively exposed as a webservice you can programmaticaly call from any system without extra coding on the ERP side? Featuring access rule and object lifecycle validations? Because in OpenERP, this makes integration so much easier. For instance I wrote an oss Ruby on Rails connector where your Model+Controller in Rails can be seamlessly imported form OpenERP while the View layer fits in Rails, and the connector is less than 500 Ruby LOC, no extra code on OpenERP. Openbravo added that recently but they only support 50% of what OpenERP does out of the box, because they only support Cread Read Update Delete and not business methods at the moment. Where does Adempiere stands?

5) I guess that Adempiere also features a "relative ID" system where you can import any XML fixtures and update existing records with their foreign keys, without any need to think about real database ID. But is it really the case on Adempiere? (For instance is is also a definitive boost with Ruby On Rails). Speaking about this I wish everybody (including OpenERP) moves to YAML as this is mostly hand written and need to be human readable and YAML beats XML here.

6) In term of scalability. We all know that Python is an order of mangitude slower than Java. But investigation shows that OpenERP SQL is quite optimized while 80% of request time is spent in the Postgres SGBD layer, so no point in adopting a faster language for less than 10% of the time (the other being latency over a good network) at the cost of code expressiveness. Also notice that raw PL/SQL interpretation is even way slower than Python... But OpenERP optimizes SQL because it leverages both an ORM level 1 cache system and also a very smart SQL cache where results are cached in columns (so read is only one select) while the cached is flushed in a transactional manner by OOP cache invalidation listeners on business objects. All this makes it very easy to have SQL optimization. Something you would achieve polluting business code a lot if you code in PL/SQL like say with Openbravo. Moreover OpenERP has native support for MPTT (Modified Pre-order Tree Traversal, see ) making it only one request to retrieve complex aggregates in tree hierarchies like accounts and stock locations. Does Adempiere feature any of those SQL optimisation, how?

7) I think that the OpenERP Pylons based web-client (Turbogears dependency dropped 4 months ago) is more mature than Adempiere emerging web clients (Adempiere Zk and GWT). I suspect OpenERP web client is used in production in more companies (inlcuding Tiny Saas offer) and is already fully functional and fast. Tell me if I'm wrong. That said, I'm enthusiast about AdempiereGWT but had trouble following the last updates about it.

8) OpenERP is component oriented, making it so easy to compose your screens using various sub-views of all your business object tree. Is Adempiere like this or is it the Openbravo way (page oriented) where you see a record and then linked objects are always shown in other tabs rather than in the same view?

9) Finally, I know that both OpenERP and Adempiere (and even Compiere) feature a powerfull workflow engine which still lacks in Openbravo (what old code the hell did they forked?). None are a standard (but BPEL might be over-bloated for foss), so I see no difference here.

10) I won't go in a language war, but in my Opinion the Java language is being slowly abandoned ( ) at the profit of less verbose langages, some being dynamic (Ruby, Python, Groovy...), while some statically typed (Scala, Clojure, D...). Notice that on the contrary, the Java Platform and JVM (portable and JIT bytecode execution which is independent from the Java language) still have a very bright future. While having been a Java fan boy by 2003, I find Python less verbose and more effective to code the business layer. Is isn't a very big win as you might argue a statically typed language makes the platform core stronger in Adempiere, which I would agree with. Adempiere might also slowly move to JRuby/Jython/Groovy JVM backed dynamic languages for its business layer while OpenERP might have a smooth transition to Jython enabling Java/scala for its very core, so no real big issue here, exept that personally I think I'm more effective today, as an integrator, writing fewer lines Python business code in OpenERP rather than in Java.

Now I easily admit some issues with OpenERP, mainly:
  • adopt a large test suite to avoid bugs/regressions,
  • set better community management process to deal with required code refactoring, API changes, release planning,
  • better ensure modules play well together to minimize runtime/integration time bad surprises,
  • consolidate financial periods ending processes and period financial reports. But so far, I think this is being solved right now and OpenERP is currently already very usable for many companies.


Still I'm very curious, what you Adempiere folks will answer to those points? Are you aware of those pros of the OpenERP platform (I recognize that knowing several ERP's deeply is hard)? Or is it me that misses Adempiere key points? What are those Adempiere pros over OpenERP in your opinion?

Sunday, May 24, 2009

Python runtimes for OpenERP, Jython, Unladen Swallow, where do we stand?


little time ago, I decided to showcase the world a proof of concept of what I assume the most mature/promising oss ERP so far - OpenERP - running on Jython, the Java based Python interpreter. I made a branch, adapting the OpenERP database connection use zXJDBC+JDBC Java Postgres drivers, escaped code that were not running on Jython yet (C libs wrappers). That attempt has been related here:

But before all, just to make it clear: I'm posting here an update of the Python runtime situation for OpenERP. That doesn't mean I consider this is a very important issue for OpenERP, it's just that there are reasons to hope for a convergence toward the Java platform over the year, nothing more. On the contrary, my only very hot topic for OpenERP now is code quality, testing and business code refactoring/cleaning. Still, that might be for an other post, I'll now only post an update about the Java convergence and Python runtime perf status.
Now, 3 months after my first post, were do we stand? There are mostly plenty of great news as I'll expose.

But again, why Jython?
1) Speed in the long run
I prefer to make it clear: for one, OpenERP v5 is now fast enough for a lots of situations. I can't see any performance issue to deploy it in large companies and it's actually being done in several already (it might not always compete in term of native features or quality in large companies yet however). And second, yes I know that in most enterprise systems, the bottleneck rather lies in the database load and in the client/server bandwith usage/latency. Now this is especially true when the language layer is fast, for instance with Java. In Python, and OpenERP especially, the language layer might not be the bottleneck, it's still common to see 30% of CPU usage eaten by Python. So any significant improvement here will also help.

And SQL requests of OpenERP are already quite optimized in v5. For instance they try as much as possible to handle n records in an O(1) or O(log(n)) time. They also have database transactional cache thanks to their fields.function + store=True/Hash of invalidation trigger feature. They also model SQL hierarchy using MPTT. So at then end having a faster Python runtime, whatever it is will certainly help.

2) Enterprise support
Speed is one thing, but an other thing that Jython would bring to large corporations is: easy enterprise integration. In Jython, any Java class/lib can be seamlessly invoked. And Java, thanks to its design by comity mantra has at least all sort of mature features the enterprise world like to rely on: good SOAP integration (we recently had some pain with the Python Soappy implementation on the contrary), standard ESB's, JCR, JDBC, JMX, BI tools, ETL's...

Google Unladen Swallow is likely to provide the best short and mid term performance:
Still, OpenERP is probably to enjoy the best short and mid term performance thanks to Google commitment with their 'Unladen Swallow' branch of CPython. In an interview, the creators of Unladen Swallow explained that Google has a lot of "legacy" code where the high level plumbing is done in Python while the low level algorithms are done in C++ and invoked via Swig wrappers. So they said they would certainly have look in Jython for long term performance, but they also need short term performance for standard CPython and that's why they launched the Unladen Swallow project.

Unladen Swallow, wants to bring a simple JIT (Just In Time) compiler to Python and do other standard optimizations. They claim they could rich 5x perf increase by the end of 2009. Those improvements will be merged back in the standard CPython distros. So far they already have a solid 20% perf improvement. They might even try to remove the Python GIL efficiently.

In the long run, even if they relied upon the LLVM, that's however unlikely they build a Virtual Machine as sophisticated as the Java one (often considered as the best VM, especially now that they added support to dynamic languages with the InvokeDynamic JSR). So if Jython then focus on performance as they announced at PyCon 2009, Jython might win the perf war in the long run. Anyway the enterprise integration I was talking about would only be achieved on Jython.

OpenERP on Jython might be a reality by Q1 of 2010.
Yes, several very good news allow me to hope this could be done. Let's enumerate them:
most of the C libs wrappers will be avoided, Tiny, the editor of OpenERP took that commitment:
- Turbogears for the web-client layer -> done! The full blown Turbogears is no more required, on the web-client trunk branch (already very stable), Tiny completely removed the Turbogears dependency and only depends on CherryPy (pur Python) now. At the discretion of few Jython bugs, OpenERP web-client should now run unchanced on Jython.
- mx.DateTime: Tiny took the commitment to remove it in the very next months!
- the minidom XML lib: Tiny took the commitment to replace it by the Etree lib which is implemented on Jython. This as already been started in community and editor branches:
- libxml usage will be removed too.
- I investigated a bit around ModJy, the Jython standard Java servlet wrapper, and it's should be quite easy to put the OpenERP server on ModJy using a WSGI wrapper.

The best is what is being done at the database/ORM level: OpenERP is moving to SQLAlchemy!
My work on the Jython branch has been to workaround the psycopg2 native driver and replace it by zXJDBC, the Jython wrapper for JDBC + Python DPAPI2. Well the good news is that OpenERP is going away from psycopg2. They will indeed move all the SQL generation logic to the standard and awesome SQLAlchemy ORM. And SQLAlchemy is Jython compatible out of the box. That's great to see them coming back to standards (SQLAlchemy was not mature when OpenERP was already in business).
Tiny already began that work on a dedicated branch:
and I encourage the community to help them do it right. Also notice that they already used SQLAlchmey for SQL generation in their emerging BI OLAP cube.

SQLAlchemy is not only a good news for Jython support. It also means that OpenERP domain (SQL filtering) logic is going to become more subtle and powerful. Currently, that part of OpenERP was not as the level of the best ORM's around such as SQLAlchemy or Hibernate. Even ActiveRecord (Rails) was doing a smarter joint job. Still, among the mature ERP's it was probably the best available, years ahead Compiere or Openbravo with their millions of pure pl/SQL legacy code (Openbravo recently supported Hibernate in their 2.50 platform, but their business code remains millions pl/SQL code lines lurking in XML CDATA statements and will still take years to be migrated to Java if ever it's to be achieved).

Oh and yes, That SQLAlchmey transition will support database independance as there are SQLAlchemy adapters for most of the market SGBD's. Especially, MySQL and Oracle will be supported. That's great because it will strengthen the OpenERP community. In some organizations Oracle is not an option (and it can arguably provide marginally superior perf), while MySQL has a much larger commuity penetration than PostgreSQL.

Unlike most large software projects, OpenERP technical quality improves slowly but keeps improving
With all that refactoring being done, OpenERP belongs to those rare kind of compex enterprise projects where the code improves over the time rather than those going desperately to chaos as businessmen add layers of crap over crap as the commercial deals are made. Also beyond the ORM transition and the webclient refactoring (removed Turbogears+changed for Mako templates), Tiny also rewrote their reporting engine an just allowed the Mako engine (a powerful standard).

My vision of software engineering is indeed pretty much a thermodynamic vision: code quality has to improve over the time much like a living system should not generate entropy in its scope. Design mistakes can very easily be done along the road and once tons of business code is build on non optimal models, then you are all screwed up: any energy you will invest in trying to correct a part of the system will actually result in a larger energy waste, pretty much like if you would like to refresh your home by letting your fridge door open: all you will get is actually get it warmer.

Why this claim? Because in business driven software, it's always easier in the short term (the only business scope as the current economic crisis teaches us), to make money supporting your existing customer base. Meaning that once mistakes are done, it's always too expensive - and thus never attempted - to re-think the core abstractions. Instead, those companies tend to make several abstractions cohabit, prefer invest in building chaotic non abstracted code instead of re-factoring the concepts that could be abstracted in the core platform. And once you have millions of lines of chaos code, you are never going to factor them back into clean/intelligible/maintainable concepts. That's pretty much how the money has been wasted in today's proprietary ERP's and even a in few claimed open ones. So they might have lot's of working features appealing to collapsing traditionnal industries, they are facing exploding maintenance costs and can't adapt further to the new emerging business as will the few viable oss ERP's do.

So overall, by early 2010, I think we should be able to see an OpenERP distributed also as a war archive you'll deploy in one click on your enterprise webserver such as Glassfish. This might shift OpenERP adoption from small SME's to large organizations which is always good because it will fuel more heavy engineering inside the project. And at the end, even the small businesses will benefit the resulting quality/usability improvement. Still I don't expect extra perf from Jython before Q2 2010 at least.

OK, enough said about Jython and Python runtimes. All that enthusiasm should also not shadow my very urgent press to Tiny to increase their quality (less bugs, less regressions). An efficient way to achieve this will be by taking the test first approach way more seriously than they have been doing so far. They just took a commitment to this too at their May Community meeting so we will see. Given what they did in the recent past (transition to international, transition to an open distributed forge, English documentation, transition to an editor business model, performance issues fixes...), I hope they will manage that challenge too.

Tuesday, March 03, 2009

magento-openerp-synchro module: we moved to Launchpad!

Sure Google Code Hosting is great,

But we are now seeing too much interest in that connector poping all over the world to continue the development in a centralized way. Also, we have been contacted by Jordi Esteve, the primary extern OpenERP contributor (after Launchpad stats) who wanted to jump in and join efforts, so he also asked us to move the contribution process forward.

That's why we have the pleasure to announce you that we migrated our development trunk from Google Code SVN to Bazaar on Launchpad, meaning we just use the advised OpenERP community advised distributed development platform. Also notice that we took care of migrating the previous SVN commits.

That was also the occasion to split the development in two branches:
  • a 4.2 compatible maintenance Branch. That one is for the conservative folks who can't afford migrating to OpenERP v5
  • a 5.x branch where all the new interesting stuff is expected to happen:
In both cases, the OpenERP addon module name is now called magento_openerp_synchro.
The last SVN version (revision #27) has been ported as bzr revision #3581.1.9 on v5 branch and rev #148 in 4.2 branch. Later on both branches got a fix for issue
v5 branch got even more attention and code clean up.

So you are very much welcome to jump in and contribute to whatever branch you need. Still, before adding any serious extra feature, we would like to insist that some refactoring should be undertaken BEFORE bloating the code.

  1. Too much code is lying inside the OpenERP wizard layers. That was a bad design choice making it hard for additionnal third party modules to extend the connector behavior and add custom features that way OpenERP allows it (thanks to its extensive OOP design). Instead, that code should be moved INSIDE THE OBJECT LAYER. That shouldn't be to hard, it's only about extracting methods, putting them in the appropriated objects (sale order, product...) and calling those methods from the wizards instead. For instance we deployed a derived version of that connector for a customer along with the sale_supplier_direct_delivery module and we had to hardcode the connector to make it account for the direct delivery eventuallity. With the new design things should work more seamlessly. See the following tracker:
  2. The sale order push feature (from Magento) has not beeing integrated properly yet and is still polluting the code. It has been a wonderfull contribution by Charles Galpin, but we had not time to integrate it the way we wanted to and it has not been tested extensively. Instead, we would like to remove the push code from Magento and instead makes Magento call OpenERP and tell him to pull the given sale order reusing it's standar sale order import code. Finally that design should be able to deal with possible network or OpenERP failure and flag failing push inside Magento for later processing. Not at top priority, but something you should be aware of. See following tracker:
  3. Finally, we insist that we din't take care yet of exposing ALL OpenERP XML/RPC server webservice the standard Magento way. One of the consequences is that currently importing sale orders is not secure unless you block external connections by IP at say an Apache level. A much better way would be to properly expose our custom extra webservice the Magento way, see the following tracker:

Meanwhile we welcome Jordi Esteve as a core contributor + project member. We know that things have been a bit slow with Smile and the connector recently (we are flooded by OpenERP non Magento demand), but we really hope to help moving foward as much as we can. We really hope all those inputs will streamline the installation process + natively supported features while minimizing the required development skills. Enjoy!

Sunday, March 01, 2009

New planet for Magento - OpenERP synchro module

Hi, I'm Raphaël Valyi, one of the main author of the synchro module between OpenERP and Magento. Some 10 months ago, after completing an extensive study about open source ERP's, I convinced Smile, my employer, to support that synchro module allowing Magento and OpenERP to work together. Some two months later, thanks to Smile and the tireless work of Sylvain Pamart, we got that connector in pretty good shape already.

Over the time, we improved it and tried to integrate community contributions as much as possible.

Since then, our OpenERP schedule at Smile got totally loaded with customer contracts letting us little time for the community even if Smile is still commited to support the effort in theory (my boss agrees on a 2 man days per month effort so far, as long as we have no higher priority, and that's the catch) . So unfortunately, it means that the connector is always lagging behind direct commercial activities.

Anyway, as I do personally see interest in pushing that connector forward, I'm taking on my spare time to give you updates about the connector in this blog, under the magento-openerp-synchro tag, so it get exported to our brand new Connector Yahoo Pipes powered planet mashup:

Ideally, I'll let any contributor (including post in that planet when they contribute in the future. So the corporate work supported by will hence probably be published both here and on the corporate blog, that is

So I hope this planet will be a fair place to give updates about the connector while accommodating corporate, personal and community contributions. Feel also free to comment in the respective planet blogs.

Finally note that this planet is managed as a Yahoo Pipes aggregator wrapped inside a Google Gadget eventually. You can contact me to have your blog (+specific tag) included in that planet if you wish.

Raphaël Valyi

Sunday, February 01, 2009

OpenERP running on Java (Jython) - ROUND 1

Hi folks,

Looking for an open source ERP that doesn't suck? Well there is at least one: OpenERP. Like it or not it's Python based. For most of the things this is a perfect fit, but let's face it, Python has less enterprise acceptance and the lower layers of that ERP could also have been powered by a powerfull virtual machine. And since building a new ERP can't really take less than 3 years, the situation will hardly evolve any time soon, so we better deal with it: I bet my money OpenERP will now take the market by storm, I don't see anything stopping it to take the lead, at least for the next 5 years.

So I wanted to show the world that it was possible to back OpenERP by the Java platform, much like it's now being admitted that Java is the best bed for Ruby on Rails applications thanks to JRuby. I built a proof of concept. Now I hope some get the message and join efforts to make it happen. I'll hardly finish that journey in my spare time, but hopefully, some people start to release how important this is and how close we are and help covering the extra miles to get it in production running on Java. Of course, OpenERP doesn't need to run Java to be the best ERP already, it is just that it could be even better. And if you want to help out, I'll be there too.

what about the claimed Java based ERP's?

I've been myself a Java programmer since around 2003, even created my first open source project on Sourceforge back in 2004 (EmSim, later on I used those skills in a variety of advanced projects when working at Amadeus and But, like it or not, after deep investigation, my conclusion is that the best open source ERP was coded in Python. I've looked deep inside Java based or claimed ERP's such as Ofbiz, Neogia, JFire, Compiere, Openbravo, Adempiere but those were too limited in my opinion. Basically, the trouble is that those 'Java' based ERP don't rely on true object oriented modeling since the ground up, some of them being even more SQL based then really Java based. I could explain in deep details why, but this is beyond the scope of this post.

So my conclusion was that even if I needed to learn Python, I would get the work done faster with OpenERP than with any of those existing Java based ERP's. My experiments proved me that I was right. With OpenERP, extending the relational model, the forms and making it fit to my needs was faster than with any of the other tried ERP's, even while being a Python noob. One year later, with a few successful implementations behind me, I'm only recommending it more than ever.

But wouldn't it be great to port OpenERP on the Java platform?

The benefits of porting OpenERP on Java are beyond the scope of this post. Still, here is a short list:

  1. Noticeable speed up. Even if the Python layer peaks up at 40% of the CPU load under high concurrent loads, getting twice as fast would be noticeable. If Jython get as optimized as JRuby which is 2 to 5 times faster than C Ruby, we will go there.
  2. Real multithreading and less memory use under huge loads (CPython scales using process and sharing nothing accross runtimes because it has a Giant Lock in its interpreter, much like PHP or Ruby). The expected benefit there is much like what benchmarks prove for heavily loaded JRuby on Rails applications.
  3. One click installation on lots of hardwares (imagine how easy it will be to sale to large companies when you will say it's only a .war you pu on your Tomcat server).
  4. Penetration in large companies where Java tends to rule more than Python
  5. Easier and solid cross database support thanks to the unified JDBC API.
  6. Larger community for OpenERP. OpenERP is certainly the open source ERP with the largest active open source community. But just imagine how large it could get when you managed to connect with the Java community, telling them: hey, we are somewhat Java based too. If you know Java you'll be able to leverage those skills to get the job done in OpenERP. Again, JRuby proved here that connecting communities was possible.
  7. Possibly larger fundings and exposure for OpenERP. Openbravo certainly had the large fundings it had by playing on its Java image which as a broad acceptance in the enterprise world.
  8. And the best of all: cross language implementation: by leveraging Jython, you would be able to call the best existing Java lib (there are some good ones like JDBC, JCR, JMX, JMS, ESB's, SOAP...), but also externalize some code pieces where speed is critical to Java, meaning an easier way to reach C speed (a warmed up Java6 runs at C speed and sometimes faster) than writing C bindings.
  9. Jython will probably ends up implementing the MOP (Meto Object Protocol) that makes it possible to share OBJECTS between JVM based languages such as JRuby, Javascript, Scala, Clojures, Groovy... Meaning that more people could adapt OpenERP to meet their needs using the language they already master. That means you could code OpenERP modules in Ruby for instance and that's not a small thing, while preserving backward compatibility. Of course, published modules would have to keep beeing closely controlled, but at least one shot veticalization modules could offer more options.
  10. Because it's fun and we can do it.

What does it take?
Well, Jython is really alive again. Notice that Microsoft hired the orginial Jython creator to create Iron Python on the .Net platform. Next, by 2008, Sun Microsystem hired the two Jython leads, Ted Leung and Frank Wierzbicki to make it Python a first class citizen on the Java platform again. Espcially, JRuby recently proved that the JVM was able to run dynamic languages faster than their native C interpreter that don't come up with such a sophisticated virtual machine. So in theory the future is bright. A 2.5 Python compliant Jython is expected by February and the current trunk largely reflecting it already.

Still, the trouble, comes with native libraries. Since Python tends to be somewhat slow, Pythonistas tend to back lot's of Python libraries by C extensions to provide extra speed. The trouble is that the Jython interpreter, sandboxed in its virtual machine environment can't invoke the same C extension, mostly because it doesn't work the same at the lowest levels.

In some places, OpenERP uses such libraries. Some of them are psycopg2, mx.DateTime, libxml2, libxslt. Overall there aren't too many fortunately so that's why this shouldn't be too hard to have a full blown OpenERP running on the JVM. The common strategy here is to set up a wrapper over existing Java libraries in place of those C extensions. Also notice that in the near future, Jython might also support CTypes, a standard way of building C extensions for Python. This is largely because Jython is benefiting here from the fantastic JRuby work of Wayne Meissner around JFFI (Java Foreign Fuction Interface). Other synergies exist, like the Da Vinci Machine, or Invoke Dynamic bytecode instruction that would help the JVM speeding up dynamic languages while optimizing memory usage.

Current status:

After some two days of heavy hacking, I'm getting:

  • OpenERP server starting powered by the last Jython Java based interpreter, backed by the regular JDBC Postgres driver over a standard Java JNDI database connection pooling, instead of Pyscopg2.
  • It answers most of the webservices calls, be it from the clients or from the webservice API. The largest thing that isn't working here is datetime operation that are using the mx.DataTime library that isn't yet fully supported.
  • Also, views don't work yet (the fields_view_get method). This is due the current XML and XPath limited support. Everything need is available on the Java plateform, but it should be wrapped properly to fake the Python API OpenERP is expecting there.

Next stages:

  • I guess that faking the required xpath and xml API would be great. That shouldn't be that hard given that OpenERP only use a few things in those API. We should only wrapp them over Java libraries such as Xalan.
  • As for the mx.DateTime trouble, the OpenERP team said they are ready to move away from it. Indeed mx.DateTime has a bug with dates before 1970. That's why the Tryton fork moved away form it, proving the thing is possible (they only use it marginaly for parsing, but alternative can be found). Advised by Jim Baker from Jython, it appears that the standard python datetime, datetime.timedelta and the library might be used instead. So aside from helping getting rid of mx.DateTime, we probably won't have much work here.
  • Once all that is working, the Java based OpenERP server will already be in pretty good shape to be used in production.
  • Of course, ideally we would package the whole thing within a standard J2EE war package and serve it with a standard servlet container such as Tomcat or Glassfish. The idea there is to write a servlet connector that will use standard Java servlet instead of the python SimpleXMLRPCServer. Then we would route HTTP request to a pool of Jython runtimes and the appropriate OpenERP server layers, much like others do with Django on Jython or the Glassfish JRuby on Rails gem supported by Sun Microsystem.
  • Once performance starts to proves better than CPython (Jython and JRuby guys are expecting this to happen by 2009 already) and once large companies show interrest in deploying OpenERP over their existing Java stack, may be the OpenERP team starts to support it officially and provide alternative Java based layers to increase performance even more or provide extra features (like SOAP, JMS async messaging, ESB like Mule or ServiceMix, a REST layer, JMX remote monitoring, OpenTerracotta clustering...)

Steps to follow to test the current Jython powered version:


  1. you should know how to start a regular OpenERP v5 server
  2. you should have an existing OpenERP v5 database to test again
  3. you should have Java installed (1.6 advised)

First you have to grab a recent Jython interpreter. Make sure you have a recent version of Java installed (1.6 advised) and then grab Jython:
get a fresh version of Jython, and build it:

$ svn co jython-dev
$ cd jython-dev
$ ant

now, put the jython-dev/dist/bin/jython command in your path.
now if you type

$ jython

that should bring you a Python 2.5, java based commande line interpreter.

Now grab my public branch of the OpenERP server:

$ bzr branch lp:~rvalyi/openobject-server/trunk-java-jython

As for the addons, you can use your regular OpenERP v5 addons, or grab them from here:

$ bzr branch lp:openobject-addons

Now open the trunk-java-jython/bin/tools/ and make sure you properly set up the database connection params and you addons path location.
May be it's better to leave the db_name to False so it will load all the features lazily when requested.

you can already see what happens if you:

$ cd trunk-java-jython/bin
$ jython

You'll probably have errors because we didn't properly set up some required libraries.
So it's time to copy paste a few libraries from our regular Python path to our Jython path:
you can copy/paste the following directories:
pychart, pytz, reportlab
/user/lib/python2.5/site-packages/ (or whatever you Python sys.path)
to [...]/jython-dev/dist/Lib

There is one more catch though. Currently, there is bug in Jython preventing the reportlab lib to load fully. This is because font listings in the pdfbase/ are just too large for the JVM spec to fit in a single Java method. The Jython folks are aware of that bug but it's low priority: see
Meanwhile, helped by the fellow Charles Headius Nutter from the JRuby Sun team who faced similar issues implementing JRuby, I patched the file to make reportlab works in Jython. It's only a wrapping of large list instantiations and doesn't remove any feature. I submitted it to the reportlab developers and hope to see it included waiting for a Jython fix.
So in any case, for now, you should just move the file provided in the trunk-java-jython branch and copy it in place of the reportlab/pdfbase/ file of your Jython reportlab installation.

Now you should be able to start you OpenERP server on Jython. Look at the following logs on my machine. Then I connected using XML/RPC with my GTK client (in that development version I'm printing all the requests, you could remove those print of course):

rvalyi@rvalyi-laptop:~/DEV/openobject_trunk/openobject-server-jython/bin$ jython
[lun. févr. 02 2009 02:41:22] INFO:server:version - 5.0.0
[lun. févr. 02 2009 02:41:22] INFO:server:addons_path - /home/rvalyi/DEV/openobject_trunk/openobject-addons
[lun. févr. 02 2009 02:41:22] INFO:server:database hostname - localhost
[lun. févr. 02 2009 02:41:22] INFO:server:database port - 5432
[lun. févr. 02 2009 02:41:22] INFO:server:database user - openerp
[lun. févr. 02 2009 02:41:22] INFO:objects:initialising distributed objects services
WARNING; Python Imaging not installed, you can use only .JPG pictures !
[lun. févr. 02 2009 02:41:32] INFO:web-services:starting XML-RPC services, port 8069
[lun. févr. 02 2009 02:41:32] INFO:web-services:starting NET-RPC service, port 8070
[lun. févr. 02 2009 02:41:32] INFO:web-services:the server is running, waiting for connections...
[lun. févr. 02 2009 02:41:40] INFO:dbpool:Connecting to template1
executing the following DB query:
select datname from pg_database where datdba=(select usesysid from pg_user where usename='openerp') and datname not in ('template0', 'template1', 'postgres') order by datname
query passed!
[lun. févr. 02 2009 02:41:40] INFO:dbpool:Closing all connections to template1
['jython', 'openerp']

So the GTK client can connect to the Jython powered server and list the available databases for instance.
But if I try to really connect to the database and go further here I get an error unfortunately:

SELECT ir_act_window_group_rel.gid,ir_act_window_group_rel.act_id FROM ir_act_window_group_rel , res_groups WHERE ir_act_window_group_rel.act_id in (1) AND ir_act_window_group_rel.gid = order by offset '0'
query passed!
{'view_id': (1, u''), 'src_model': False, 'view_ids': [], 'context': u'{}', 'view_type': u'tree', 'auto_refresh': 0, 'usage': u'menu', 'res_model': u'', 'domain': u"[('parent_id', '=', False)]", 'name': u'Menu', 'id': 1, 'view_mode': u'tree,form', 'target': u'current', 'views': [(1, u'tree'), (False, u'form')], 'limit': 80, 'type': u'ir.actions.act_window', 'groups_id': []}
('jython', 1, 'admin', 'ir.ui.view', 'read', [1], ['model', 'type'], {'tz': False, 'active_ids': [], 'lang': 'en_US', 'active_id': False})
executing the following DB query:
SELECT "model","type",id FROM "ir_ui_view" WHERE id IN (1) ORDER BY priority
query passed!
[{'id': 1, 'model': u'', 'type': u'tree'}]
('jython', 1, 'admin', '', 'fields_view_get', 1, 'tree', {'tz': False, 'active_ids': [], 'lang': 'en_US', 'active_id': False})
executing the following DB query:
SELECT arch,name,field_parent,id,type,inherit_id FROM ir_ui_view WHERE id='1' and model=''
query passed!
executing the following DB query:
select arch,id from ir_ui_view where inherit_id='1' and model='' order by priority
query passed!
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[01]: Traceback (most recent call last):
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[02]: File "/home/rvalyi/DEV/openobject_trunk/openobject-server-jython/bin/osv/", line 60, in wrapper
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[03]: except orm.except_orm, inst:
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[04]: File "/home/rvalyi/DEV/openobject_trunk/openobject-server-jython/bin/osv/", line 120, in execute
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[05]: cr.commit()
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[06]: File "/home/rvalyi/DEV/openobject_trunk/openobject-server-jython/bin/osv/", line 112, in execute_cr
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[07]: File "/home/rvalyi/DEV/openobject_trunk/openobject-server-jython/bin/osv/", line 1082, in fields_view_get
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[08]: doc = dom.minidom.parseString(encode(result['arch']))
[lun. févr. 02 2009 02:44:30] ERROR:web-services:[09]: NameError: global name 'dom' is not defined

So yeah, not everything is working yet, and especially the view layer don't work yet because we didn't provide appropriate wrappers for the xml/dom and xml/xpath Python modules used by OpenERP.

The same goes withe the web-client (eTiny), it can list the databases but won't show any view properly yet.

Still, more something interesting is that webservices, which only care about the model and controller layers are already working quite well. See yourself, in an standard python or jython console, I can for instance read all the info related to a given product (here with id 1):

>>> sock = xmlrpclib.ServerProxy('http://localhost:8069/xmlrpc/object')
>>> sock.execute("my_test_base", 1, "admin", 'product.product', 'read', [1])
[{'warranty': False, 'property_stock_procurement': [5, 'Procurements'], 'supply_method': 'buy', 'code': False, 'list_price': 38.25, 'expected_margin_rate': 0.0, 'pricelist_purchase': 'Default Purchase Pricelist (0.00) : 25.50\n', 'incoming_qty': 0.0, 'weight_net': False, 'standard_price': 25.5, 'cost_method': 'standard', 'active': True, 'price_extra': 0.0, 'mes_type': 'fixed', 'uom_id': [1, 'PCE'], 'uos_id': False, 'ean13': False, 'default_code': False, 'type': 'service', 'property_account_income': False, 'qty_available': 0.0, 'sales_gap': 0.0, 'id': 1, 'expected_margin': 0.0, 'uos_coeff': 1.0, 'virtual_available': 0.0, 'seller_delay': 1, 'total_cost': 0.0, 'purchase_ok': True, 'date_from': '2009-01-01', 'property_stock_account_output': False, 'track_outgoing': False, 'company_id': [1, 'Espace Loggia'], 'product_tmpl_id': [1, 'Onsite Senior Intervention'], 'state': False, 'loc_rack': False, 'pricelist_sale': 'Public Pricelist (0.00) : 38.25\n', 'uom_po_id': [1, 'PCE'], 'price_margin': 1.0, 'price': 0.0, 'property_stock_inventory': [4, 'Inventory loss'], 'loc_case': False, 'sale_avg_price': 0.0, 'description': False, 'track_incoming': False, 'property_stock_production': [6, 'Production'], 'purchase_avg_price': 0.0, 'weight': False, 'supplier_taxes_id': [], 'volume': False, 'normal_cost': 0.0, 'outgoing_qty': 0.0, 'dimension_type_ids': False, 'date_to': '2009-12-31', 'procure_method': 'make_to_stock', 'sale_num_invoiced': 0.0, 'variants': '', 'partner_ref': 'Onsite Senior Intervention', 'loc_row': False, 'purchase_num_invoiced': 0.0, 'sale_ok': True, 'rental': False, 'packaging': [], 'sale_delay': 7.0, 'name': 'Onsite Senior Intervention', 'total_margin_rate': 0.0, 'description_sale': False, 'property_account_expense': False, 'categ_id': [8, 'All products / Sellable / Services / Onsite Intervention'], 'invoice_state': 'open_paid', 'property_stock_account_input': False, 'track_production': False, 'sale_expected': 0.0, 'lst_price': 38.25, 'taxes_id': [], 'dimension_value_ids': [], 'produce_delay': 1.0, 'seller_ids': [], 'description_purchase': False, 'turnover': 0.0, 'purchase_gap': 0.0, 'product_manager': False, 'total_margin': 0.0}]

OK, enough said for now. I'll try to post updates on the dedicated Launchpad blueprint:
Don't hesitate to contact me if you want to push this work further.

Raphaël Valyi.

Saturday, January 24, 2009

Java 7 build 44 improves jar build speed a lot indeed


from the changelog of jdk7 build 44, one can read that they improved on the speed of the jar task.

Indeed, I can confirm this for building JRuby which is quite a large project:

Before, Java 7 build 43:
>ant-jar complete
Total time: 1m30 seconds

Now, Java 7 build 44:
>ant-jar complete
Total time: 56 seconds

Quite enjoyable indeed!