Miscellaneous Computer programming

Tenor Viol

Lurking halfway up a hill Cumbria
Messages
7,447
Location
Penrith, Cumbria, UK
Level
Ca. G7 cello, G6 sax
It is worth remembering this basic fact about computer programs: in order to test a program completely, you have to test every path through the code. Every time there is a choice: A or B, then there are two paths... The number of paths in a program is therefore 2 to the power of the number of decisions. That very quickly becomes a very large number.
 
@tenorviol Meaning that there are an almost infinite number of possible outcomes or decisions in music? What do you call decisions in music? In composing or improvising? Or were you referring to the AI stuff?
 
Last edited by a moderator:
Meaning that there are an almost infinite number of possible outcomes or decisions? What do you call decisions in music? In composing or improvising?
No, it's nowhere near infinite. It is also remarkably easy, trouble is a lot of programmers can't be bothered and assume the faults will be found in testing. Then they aren't and the systems crash live,
 
No, it's nowhere near infinite. It is also remarkably easy, trouble is a lot of programmers can't be bothered and assume the faults will be found in testing. Then they aren't and the systems crash live,
Agreed, especially when some idiot only uses 'perfect' data that matches the requirements for testing and not real data...
 
Agreed, especially when some idiot only uses 'perfect' data that matches the requirements for testing and not real data...
This is also why you must always ensure testing is done by a service independent of the programmers. Obviously the programmers must do their own end to end testing as part of their quality checks, but the testers must be absolutely independent of programmers to ensure there are no influences of vested interests in geting something signed off.
 
I'm not sure why this thread is on the cafe, but way back in the dinosaur days of IT, I once gave a 'guest course' on program testing at a local university. It was many years ago but one of the basic distinctions at that time was between 'white-box' and 'black-box' testing. 'White-box testing basically entailed a full-path analysis and testing of each path through a program/system.

Black-box analysis and testing was entirely different..Black-box analysis simply ignored the 'specifications' and the internals' of any program/system and tested the responses of the 'black-box' program/system to any real-world situations that could occur (and for which the program/system perhaps wasn't fully designed to cope with).

These days, I have no idea how IT-techs cope with an ever-changing IT-landscapes. In my IT-dinosaur days, we had some automated test tools. I suspect that test tools available nowadays are way more advanced!
 
A tad more exponential than that

BBC yesterday:

"Google's Sycamore quantum processor was able to perform a specific task in 200 seconds that would take the world's best supercomputers 10,000 years to complete."

That's about a 1,500,000,000 times faster but IBM claim they could do it in 2.5 days. on conventional equipment which makes the quantum machine a measly 1,000 times faster.

As I understand it they are not general purpose machines and are only suitable of solving a certain class of problems which includes cracking "uncrackable" encryption algorithms.
 
It's not the testing, so much as the test plan's accuracy to the specifications. And there is a big potential for deviation. Much testing can and is automated. But while this may prove functionality, it may not prove capacity or performance. Testing these is often done by extrapolation as full sized test systems are simply too expensive.

So one must track changes and monitor systems in real life and be able to reverse the changes if there's a problem which sneaked through.
 

Similar threads... or are they? Maybe not but they could be worth reading anyway 😀

Social

New! Social Feeds

Look out for the navigation icon:

Follow CafeSaxophone

Latest Song of the Month

Trending

Forum statistics

Topics
32,076
Messages
578,681
Members
8,136
Latest member
peterdehelegrote
Back
Top Bottom