http://www.adprecision.net

Advertising and classifieds solutions for publishers, portals, advertisers etc. Saas build on unidata.

What Tony says is good advice.

Rgds
Symeon

When we brought Oracle on board, it was only after we were fully in bed with them that the costs were at least 10 times those of our UniVerse platform; 10 times the servers, 10 times the disk requirements, 10 times the administration requirements, 10 times the cost in license fees and maintenance!!!

UniVerse is still the mule that pulls our applications down the road. We have bolted on interfaces with Oracle, CADD tools, project scheduling software, and now 3D CADD with real time tie-ins to our material requirements planning app running in UniVerse. All these tools are bound together by the core apps that were developed over 20 years ago. And the mule keeps on pulling.

Maybe OSS needs a model like “buy it or develop for it”. You could get credits by helping develop, debug, document, or manage. Or you could just buy the software.

http://python.about.com/b/2006/11/11/python-is-free-who-pays-to-support-it.htm

Best converter for snippet to vsi converter can be downloaded from http://snippet2vsi.codeplex.com

The focus should be encouragement to drop the DIY mentality and use what are considered mainstream repositories, resources, and tools; not only to aid the old hands, but give MV better exposure to new blood. Host code & collaborate at Bitbuck, Github, etc., and post questions/answers at Stack Overflow (SO).

There is a bit of MV activity at SO with questions tagged as “multivalue-database”. There are also questions tagged with Universe, Unidata and U2. Unfortunately, the tag “multivalue” itself seems to have already been associated with an aspect of some SQL databases.

– Kevin

The rhinocerus website has no content of it’s own. It’s a mirror cite for Usenet newsgroups. comp.databases.pick, which I haunt, is one of the groups. CDP is just one of many Pick-based forums. I think it’s healthy to have many forums for focused discussion (and CDP for anything else).

I think it would be as healthy to have a number of community collaboration sites, but I suspect if we had ten sites to appeal to ten tastes there might be ten people in total from this community using the resources, one for each site. Of course I jest, but we are that factionalized, and that disinclined to use such resources. And thus as a marketplace, here we are.

Thanks for your comments guys!

And Rhinocerus:
http://www.rhinocerus.net/forum/databases-pick/

etc. ad infinitum

If we can’t even agree on a central repository, how does anyone expect to collaborate? And when searching the internet for information, it doesn’t help that we use some of the most common words in the English language to name our products: PICK, UniVerse, RECALL, INFORM, Cache.

While there are many aspects of FOSS itself that I am not a huge fan, I agree completely with your top three reasons for the lack of its success in the MV community.

As for collaboration, the DIY mentality that you identify as part of the problem with the adoption of FOSS in the MV world can be extended to partially explain the failed efforts of MV centric code sites. I appreciate the good intentions and great effort that Glen B. and others put into PickSource, MVDevCentral, etc., but those sites themselves were prime examples of that “roll-your-own” mentality; they were destined to never reach beyond an existing MV market.

While SourceForge was all the rage back in PickSource.com days, “modern social developers” are collaborating via sites such as GitHub, BitBucket, Google Code, and a few others. Of course, using such sites would require that MV people started using a VCS/DVCS that they didn’t write themselves, if they’re using any at all.

- Kevin

Besides that, I’ve been busy looking for work, getting work, learning Python, and learning D3. D3 is so not(UniVerse). I’ve also been trying to carve out time to learn Cache’.

Maybe we need to use for a central repository is what most everyone uses everyday, like Facebook, where you could put programs in Notes, and have people comment on them. Keep it simple.

Our Seattle Area Pick Users Group (SAPUG) group is getting motivated to help educate the world about multivalue, and we are talking about posting links to classes (like Dawn’s), and to do some Multivalue University stuff ourselves to get people excited and educated. There are so many brilliant people around, it is a shame not to use them as teachers of the coolest language in the world. In the past couple of months I tried to learn Python by myself, and got to where I could cut and paste to create programs, and sort of figure things out, but I have since signed up for a beginning Python class, and have learned more by leaps and bounds then I did in self-study. There is something about a teacher telling you what is important, and teaching in a logical sequence that beats reading books about programming all to pieces.

We should do the same thing with multivalue.

– Louie

p.s. In my first Python class session I realized that Excel was a multivalue database. Just enter data in a cell, then press the Alt+Enter keys in a cell, enter more data and you have an example of a multivalued cell. Whether or not you can easily extract that data is another question, but it demonstrates that people know about multivalues already — they just don’t know they know.

To your points on the first matter:

PS was phpNuke and MVDC was GForge=Sourceforge (history), and since these are industry standards, inability to navigate must be a “localized” problem. Sure, those environments have their problems, but tens of thousands of people “get it” and use them every day. I was really just citing Glen’s noble efforts as examples – I can’t think of any other MV initiatives that use industry standard CMS.

That aside, I think the larger issue is that the MV market has enough people to quietly kick the tires on a community site, but not enough people to care enough to post notes to a forum or to admins about UI issues, and to see issues through to usability. Every effort in this market is a personal initiative by one or two people working in a vacuum. So there is little collaboration on environments. And when projects hosted at these sites get zero collaboration to enhance the projects hosted at these sites, there is just no motivation to keep maintaining the resources for admins or authors who post code.

As to your notes about checkout, etc: sure those are valid considerations. It’s easy to toss a single program around, but there are no standard protocols for passing around multiple programs and dict items. I wish we could elevate the discussion to the improvement of our distribution and collaboration methods, but with little interest in FOSS itself, is it worth it? Maybe this is a chicken and egg scenario, but no efforts to “fix the egg” have led to any chickens, and few chickens are laying any eggs, so it’s tough to justify any community efforts to solve this problem of MV-integrated CVS, etc.

To the second matter:

As with any utility, MVWWW wasn’t for everyone, but again that was just one example. None of the code at PickWiki gets enhancements and I’ve never received updates to my own freeware either. People are discouraged from posting FOSS for open consumption, not so much because there’s nothing coming back, though that’s significant, but usually because authors almost never get any feedback at all. Tell the author there’s a bug or that it doesn’t work in a specific environment. Tell us it’s great or that it sucks. Anything. Commentary is collaboration. Usage is collaboration. If it doesn’t look like anyone is using published FOSS there’s no motivation for even publishing it.

As another example of lack of interest for FOSS in this market, rather than copying code from author sites to PickWiki, in some cases I posted links from PickWiki to the authors. Along with some links I commented that it would be a loss if those sites closed down because we would collectively lose the wonderful resources. No one took initiative to do anything about that. Sites did close down. No one cared. You see, it’s not the environment, at the core of our issue here is a total lack of interest in the concept of FOSS.

I’ll agree with your summary, but don’t think that’s what’s stopping this market from collaboration. I hope we get an upswell of people who say I’m wrong. (Wouldn’t be the first time that’s happened. )

I didn’t understand the motivation behind the MVWWW project. I know I can create html in java and other languages, why not BASIC too, right? In my own experience, we had a similar project (tool) and every time the web guys wanted to change the html they had to get a PICK programmer to make the change. So in my own bias, MVWWW did not deserve my attention.

Aside from those issues, think about the dynamics of collaborating on a specific project. In other environments, I use some sort of source code “checkout” tool, or perhaps the IDE itself, to get the source code from the central repository. I modify, build, and test my changes within the IDE’s debugger, and commit them to the project. Similar steps for a PICK based project are clunky at best; a person has to be pretty motivated to go through the pain and agony of collaboration.

This really is a “social interface” problem. Building tools and infrastructures that “work” but don’t model the natural way that PICK programmers would otherwise collaborate are a waste of time and energy.

I also have a project for MV on SourceForge – ‘UniBasic Code Coverage’. I previously wrote this for my employer, but got permission to rewrite it and release as open-source and have just started looking at it again.

As to who, I am putting my hand up. The last few weeks I’ve been taking notes on what would be required to make an open-source repository for MV databases work.

For anyone who is reading this, feel free to contact me if you wish to collaborate. Important to note is that you must be willing to actually work on, or support the administration of, open-source projects in the MV stack.

BTW, I believe D3 v9 now has COMO, or at least it’s in the queue because I requested it.

Also, D3 does have files similar to type19. You don’t create a DIR file in D3, you just ensure that you can read/write from a host OS path, then setup a q-pointer like this:

01 q
03 /tmp/somepath

or

01 q
03 c:\abc\def

This is called OSFI and I’ve written about it extensively, split up between this blog, the TigerLogic forum, and comp.databases.pick.

You can also open directories directly from basic:

OPEN “D:\MYPATH” TO FV…

or use paths from TCL:

SORT /home/me/foo WITH A1 NOT “X” …

Enjoy!

Yes, the techniques described here are used for AccuTerm GUI, in DesignBais, and in other products. I’ve used this with mv.NET, UniObjects, Java and PHP interfaces, FlashCONNECT, and others. I’ve even used these techniques to add GUI options to a RPL application, where RPL is a Proc-like language alternative to BASIC for D3. Regardless of the languages used, these are common coding practices among those of us who do this regularly, but many people do not yet recognize the patterns and might not know how to start doing this from scratch with their own apps.

An important message here is that coding like this is required for any external connectivity, and once done it can be adapted for use with any product. So start now, separate the UI from the rules, and figure out what external tools to use as a separate exercise.

Articles like these are very interesting to compare techniques and to guide developers. If my English was correct, I would write more comments and I would exchange more opinions.

I disagree with “You won’t be coding in DataBasic for the web stuff”. What I’ve learned from creating MANY web interfaces is that we really don’t want the business rules in the client or the middle tier. We want those rules and the database manipulation back on the server where it’s always been. Sure, there are some things we can do off-server to save net transit time. But when it comes to heavy reading, writing, and processing data, I choose not to do this in a cool middle-tier framework that supports CRUD and other fancy features. I keep the UI and middle tier as thin as possible and pass just about everything to the server for processing. The CRUD tools and class generators have their place, but as usual they are not universal solutions.

For multi-tier development sometimes a decision needs to be made about where to put a code segment. For example, do I pass field X and Y to the server to test them, or do I test them in the client or middle-tier? Do I pass X to the server so that it can read and test with Y, or do I pull Y from the server, save the burden of passing in X, then test X and Y in the middle tier? Those sorts of questions are on the fringe. Most complex operations are a no-brainer for the server. We’ve got a great engine and we should use it.

THAT is why I personally choose Pick today.

Skip the text and just notice that I do have nice formulas like this: =NxRead(“server”,”file,id,atb,val,sv”,”options”).

Then notice the screenshot of the Database Explorer I built into Excel, plus toolbar buttons, etc.

All of that was done with a great product called Add-in Express http://www.add-in-express.com/. I can’t say enough good things about the software, people, and community for that line of products (without drooling on my keyboard). Their helper libraries saved me from all the pains documented in this blog – but by the time I got to savor the tasty goodness of success I had to move on to other business, and I never got to publish that offering. Perhaps one of these days I’ll get a chance.

I strongly encourage Microsoft Office developers to get familiar with the Add-in Express offerings. The documentation might leave you wondering where you separate VSTO from their components, when you need either or both. But the answers are out there, at least you have a choice, and success is on the other side. Just looking at the screenshots you can see some great things done with Excel. End-users look at that and think “I expect this of Excel” but developers like us recognize the pain required to make it all happen, and wonder if it’s even possible outside of Redmond. That pain was eliminated with Add-in Express. (End unintentional and unpaid marketing speech.)

I hope I can find some time just to blog about Add-in Express, which I used successfully for both Excel and IE integration, but unfortunately not for anything for sale yet. Good luck!

Great work on this series… wish I had found your blog sooner, it would have saved me a world of hurt!!

I realize that a couple years have past since the last update here, but nevertheless its still pretty relevant stuff, disappointingly very little has been done by Microsoft to make this diabolical process easier.

I have spent the last two weeks on getting my automation add-in right, I now have my solution, even though I am still working out all the kinks in production, but at least I have successfully deployed.

I went the route of VSTO with two assemblies (I noticed you wrote about this option somewhere but hadn’t explored it), this was incidental though, had I seen your blog I might have taken a different route… the idea is that they are both COM Visible, strong name signed and the Assembly with my UDF Class implements IDTExtensibility2 with the difference being that it is registered for COM Interoperability… I have to run regasm /codebase it seems on both dll’s after installing and only then does it work. I have seen some Microsoft bug reports about this, that the types don’t get registered correctly unless you explicitly do this.

I am not well versed yet with all of the semantics, VSTO and COM are new to me, but I appears I can get a reference to the Excel application using IDExtensibility2, it seems to be the same instance as well but my managed VSTO code resides in another AppDomain so I simply marshal the objects I need using domain.CreateInstanceAndUnwrap, and those objects extend MarshalByRefObject and are serializable so interoperability is quite seamless.

People suggest performance is not ideal with this approach, but I disagree, its certainly acceptable (they run within the same process for goodness sake) I find its pretty fast (even close to instant), and if you design it well even better.

I am using the VSTO managed code to preload some large datasets, and they don’t get too much larger than the ones I have, which I indecently fetch using SOAP Web services, native SQL will be even faster, and these DataSets can even be cached now.

I didn’t go the shim route, in theory that does look like a better option from reading your blog, nevertheless I thought I would share my alternative approach that seems to do the job, being slightly simpler to implement and deploy.

Cheers!
Peter

I highly recommend you check with the people at Add-In Express to see if you can do what you ask to extend Excel. Please post here when you get an answer. Good luck!

http://www.add-in-express.com/