Tuesday, September 23, 2008

Going Bad with UI Prototypes

This whole month went busy at office, it was challenging work, made a number of mistakes and I am happy to say, by the end of the day I learnt out of it.
I was tyring to get the module working with little knowledge of what the requirment was, this led me to lot of re-work and most of the time was spent discussing with the team what API should be given and how things should be done. Now I think it would have been better if we could have understood what we were really building.
I have been developing the front end, most of the time it invloved only UI logic, the only thing that I had to do was, make service calls, get the data and display it, the requirment was based on a document that provided what the UI should provide, that was easy...I started designing the the UI, got a bunch of APIs from the service team, and started filling in, what I did not understand was the whole idea behind the requierment.
As days went on, issues started to rise, and I had to sit with the service team and ask ourselves why we were doing a particular thing , this ultimatley led into rework and hours of time were spent on discussion.
I learnt my lessions, what ever you are doing, understand the business requirment for why you are doing something, without basing your implementation on a mere UI prototype.
Well, guess that's the disadvantage with UI prototypes; you try to implement what ever that is in the prototype without ever thinking why your are doing it.
I would not even want to design a screen for the sake of the UI prototype, without understanding what each and every field does or it's intended purpose.

Sunday, September 7, 2008

Dynamic Language Runtime (DLR), IronRuby and Advantages

What's being popular out there right now is something called IronRuby, which is an implementation of the Ruby language in .Net, which is built indirectly on top of the CLR.
So let’s talk a little about how IronRuby fits into the .Net world, something that you should know at this point is that Microsoft is building a common platform for dynamic languages called the Dynamic Language Runtime or DLR for the short, which runs as a service on top of the CLR.

DLR provides common functionalities like MSIL generation, GC, so that dynamic language implementers can focus on the implementing there dynamic language on top of this without worrying about the “gooing” part to the CLR.
All that the language implementer needs to do at runtime is to parse the language, tokenize it and create a tree structure containing the expressions and statements and pass it on to the DLR.
At the next step the DLR is going to generate MSIL from these trees, and the MSIL would execute under the CLR. Some pretty interesting things to note here is that when the DLR does not know how an operation should be actually executed, it looks at the currently executing dynamic language and ask it how the operation should be done, then the language responds with another tree on how the operation should be done, if the language does not know how the operation should be executed than the DLR takes a default action into evaluating the expression.

Something interesting to note here is that, these trees that the DLR gets back from asking questions on how to do certain thing in a language is actually cached, so that the next time the DLR encounters something like that, it looks at the cache for an answer, and does not go through asking the language for help.

The DLR is still being built and up to now developers involved in the DLR , says there are lot more work to be done.
The whole point of the DLR and the language actually complies into MSIL and gets executed under the CLR is interoperability between these dynamic languages built on top of the DLR and other .Net languages. This would actually mean that you can consume a library that was written on one of these languages inside C# or vise versa, i.e. consuming a C# library from these dynamic languages.

This brings us to the advantages of this hectic effort, something that most developers face is covering unit testing of there respective module, one the problems that they face is the time limit when writing these in something like C# or Java.
Now, with dynamic languages being able to interop, you can write your unit testing in a language like IronRuby and consume your .Net APIs. This is much easier cos' writing your test cases in a dynamic language is must faster, and increase your productivity.

Another area where this can help is in the area of domain specific languages (DSL), where your domain specific grammar can be written in a language like IronRuby or IronPython and where your actual code generation can happen in a statically type language like C#.
This effort is also worth while in places where we use XMl as “processing” language, like in build scripts, where your build "logic" can be written in a language like IronRuby, after all writing logic in a programming language is much easier then writing in XML, I never liked XML for these purposes anyway..

There are some dynamic languages that are being implemented now, like IronRuby and IronPython, and there initial implementation can be downloaded and played around with. I choose IronRuby as the language and trying it out a little bit now a days, you can try it out too, but before that, try learning the real power in dynamic languages and try playing around with Ruby before you go ahead with IronRuby.

Wednesday, September 3, 2008

Don't Try To Handle What You Cannot Handle

One thing that came up last week was that the sorting on one of the paged grid did not work, ok…so we were hunting the bug down, but lucky something popped up in the log file and saved the day.
Clicking on the sort headers, makes a service call to retrieve the sorted data, and the UI code was written in such a way that the service call is trapped inside a try-catch block, and if any exception occurred, the exception is logged and empty data returned.

And unfortunately the service query had a bug on it. So which brings me to the point, you should catch exceptions only when you can handle it, for example, you might have anticipated that a DuplicateNameException would be thrown, well go ahead, catch it and show a nice message to the user, but if you cant, don’t put it inside a catch block and bypass the exception just for the sake of not crashing the UI.
If you really want to log it, fine, log it then remember to re-throw the exception although re-throwing as its own performance cost.

And to be amazed, the error was there al the time and no one noticed it, which brings to the justifiation the if you dont handel exceptions that you cant handle, you would have seen the issue so much earlier.