It’s not the only way to go. For years programmers used one tool for quick prototyping and switched tools for final release. Often you can predict that you will need to write you own .dll for labview box, so it may be worth spending some time getting C code right.
One of the concepts of good programming which helps to improve your own code is unit testing. I don’t know why, but this topic is considered to be advanced and you will rarely see it in scientific programming books.
Unit testing exists even for matlab.
Writing the tests is a good way to rewrite the pieces of code and still be sure you didn’t break your model.
I have heard about test driven development before, but I haven’t used it until now. This year I am writing very complex WiMAX interference model, with arbitrary number (hundreds) of transmitter and receiver pairs. I wanted to make sure that my model doesn’t break laws of physics, especially after my transmitters become directional. So I had to write a lot of tests. I switched from Matlab to python (writing complex data structures in Matlab like 5 dimensional hashed hash is not a fun) and I found it really easy to write a test, then rewrite the code in the model and making sure model still correct.
From now on, I will try to programme in the following fashion:
Write prototype
Write test code for working prototype
Rewrite prototype as much as you like, just keep tests up to date
So far keeping tests updated was most difficult part, especially high level complex tests. But it makes refactoring (programmers term for improving code) very easy. Really easy. I am curious about any book or articles on the subject.
Simplest test for example for square root function should check return values, like
test if sqrt 4 return 2
test if sqrt -1 return error (nan) or complex
test if sqrt 0 return 0
It is easy to write and makes rewriting easy. More importantly it builds great confidence in your own code, which can pay off when you will be questioned about your results. I wish I used it before, but my excuse “There is no time to write a test code” just an excuse.
]]>
<<script language="javascript"> RCPcountry = "RU"; </script>
<script language="javascript"\>
MTCN = "*";
</script>
<script language="javascript"\>
txnFee = "14.0";
</script>
Bibliography, something bibtex compatible similar to http://www.citeulike.org/, but project oriented. Since most of the databases support REST/XML type api integration with PubMEd and IEEE should be fairly straight forward.
Version control on any stage similar to mercurial/git. For separate pages and bibliography. Github or trac can be used as a way to start.
Distributed and decentralised. That's a dream. It would be cool to be able to synchronise offline and online activity. Again as it is done in mercurial/git distributed version control systems (DVCS). Actually, it should be doable, for example, if whole database kept in DVCS, it should be possible to synchronise the data from offline into online, with collision detection and automatic resolution.
For security obsessed gpg based encryption on low level. Difficult and probably will screw version control.
I hacked Latex plugin for movable type and I can add graphviz plugin for mt as well, but I don't think MT will be suitable platform to start this. Although community edition looks attractive. So, does anyone knows customizable wiki based on python and storing data in mercurial? If not, anyone for the challenge of hacking it all together? This year I am writing things in python, due to google apps engine and my own third attempt to grab it. And it didn't finish with smashed keyboard so far, so links to python project preferable.
]]>He said: "I am choosing database server for the bank, I can by cheap MS Windows Intel server + MS SQL server and it will cost roughly around $7000 (my thoughts: huge amount money for the student, doesn't sound cheap to me), or I can buy AIX server with DB2 for $70000. Shall we go for the cheapest option? Well, the penalty for not submitting tax report on time for the bank is $500000 and with MS SQL+Intel we have a higher chances of paying this penalty"
And I think you may find that using wrong technology you will pay a different penalty on regular basis.
When I changed my carrier from IT support to academia, I decided that the last thing I want to do is to support my own computer and so it was linux on my laptop for two or three years.
Think about the cost of the laptop for the moment:
Let's say I want to buy Dell laptop now, for example take VostroTM 1510 for £479, for this price I will get Intel® Core™2 Duo Processor T8100 (2.1 GHz), if I want to go for a higher frequency 2.4 GHz I will need to add £90.
Then, because it's windows computer you need to add Norton IS for additional £19.94. Which probably will occupy most of this additional GHz for which I just payed additional 90 quid.
The bottom line - I don't want to spend additional £100 for the pleasure to have pop-up messages in most inconvenient time during my presentation. Thank you, I stayed with linux.
Then I found out that most scientific software runs not only on linux, but on Mac OS X, which is the proper unix with splendid GUI.
There are more reasons for using *nix systems for scientific work:
1) a lot of this software were initially written for unix, some of it was ported to windows and some of it multiplatform, so OS doesn't really matter.
2) does anyone of you know about successful scientific grid computing or cluster on windows? Ours run linux with sungrid and similar things.
When I saw Steve Jobs presentation announcing Apple transition to Intel processors, I had two things in mind:
1) I want to use Keynote for my own presentation
2) I want to buy last reliable Mac
So I bought powerbook 12" and happily use it for nearly 4 years. Do not get me wrong, move to intel processors was a very good step for Apple, but having engineer training I understood that it will take some time before new Apple computers will become stable and didn't want to test them.
I use Textmate and Quicksilver on Mac and can't imaging doing any work with them. My powerbook proved to be very reliably so far and it was switched off for about three weeks in total during more than three years. I am restarting it once in a month and that's about it. So far Mac for me is an excellent and productive platform and I enjoy working on it, which never happened before.
"But it's different for enterprise" - someone may say.
You absolutely right, it's much worse for enterprise than for individual. But about this will be my next post.
For now, do not let decision made 5-10 years ago affect you now.
javascript:document.location.replace("get_video?video_id="+swfArgs.video_id+"&t="+swfArgs.t+"&fmt=18");]]>
]]>