Seite - 22 - in Programming for Computations – Python - A Gentle Introduction to Numerical Simulations with Python

22 1 TheFirstFewSteps you try to run theprogramto seewhatPython’s response is. Thenyouknowwhat the problem is and understandwhat the errormessage is about. Thiswill greatly helpyouwhenyouget a similar errormessageorwarning later. Very often, youwill experience that there are errors in the program you have written. This is normal, but frustrating in the beginning. You thenhave tofind the problem, try tofix it, and then run the programagain. Typically, youfixone error just to experience that another error is waiting around the corner. However, after some time you start to avoid themost common beginner’s errors, and things run more smoothly. Theprocessoffindingandfixing errors, calleddebugging, is very important to learn. Therearedifferentwaysofdoing it too. Aspecialprogram(debugger)maybeused tohelpyoucheck (anddo)different things in the programyouneed tofix. A simpler procedure, that often brings you a longway, is to print information to the screen fromdifferent places in the pro- gram. First of all, this is something you shoulddo (several times) duringprogram development anyway, so that things get checked as yougoalong. However, if the final programstill endsupwith errormessages, youmaysave a copyof it, anddo sometestingon thecopy.Useful testingmay thenbe to remove,e.g., the latterhalf of theprogram(byinsertingcommentsigns#),and insertprintcommandsatclever places to see what is the case. When the first half looks ok, insert parts ofwhat was removedand repeat theprocesswith thenewcode.Usingsimplenumbersand doing this inparallelwithhandcalculationsonapieceofpaper (forcomparison) is oftenaverygood idea. Pythonalsooffersmeans todetect andhandleerrorsby theprogramitself! The programmermust then foresee (whenwriting the code) that there is a potential for error at someparticular point. If, for example, someuser of the program is asked (bytherunningprogram)toprovideanumber,andintendstogivethenumber5,but insteadwrites thewordfive, theprogramcouldrunintotrouble.Atry-exception constructionmaybeusedbytheprogrammertocheckforsucherrorsandactappro- priately(seeChapter6.2foranexample),avoidingaprogramcrash. Thisprocedure of trying an action and then recovering from trouble, if necessary, is referred to as exceptionhandlingand is themodernwayofdealingwitherrors inaprogram. When a programfinally runswithout errormessages, it might be tempting to think thatAh..., I am finished!. Butno! Thencomesprogram testing, youneed to verify that theprogramdoes thecomputationsasplanned.This is almost anart and may takemore time than todevelop theprogram,but theprogramisuselessunless youhavemuch evidence showing that the computations are correct. Also, having aset of (automatic) tests saveshugeamountsof timewhenyou furtherdevelop the program. Verificationversusvalidation Verification is important, but validation is equally important. It is great if your programcandothecalculationsaccordingtotheplan,but is it therightplan?Put otherwise, you need to check that the computations run correctly according to the formulayouhavechosen/derived. This isverification: doingthe thingsright. Thereafter, youmust also checkwhether the formula you have chosen/derived is the right formula for the caseyouare investigating. This is validation: doing the right things. In the present book, it is beyond scope to question howwell themathematicalmodelsdescribeagivenphenomenoninnatureorengineering,