directory-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "John E. Conlon" <>
Subject Re: Class + Class Helper Pattern
Date Tue, 13 Feb 2007 19:19:27 GMT
Emmanuel Lecharny wrote:
> Hi Ole,
> I will give you my opinion about unit tests and method protections. Do 
> not consider it anything else than my personnal opinion, based on my 
> own feeling and 25 years of coding experience (not always happy).
> On 2/13/07, *Ole Ersoy* < 
> <>> wrote:
>     Hey Guys,
>     <snip> 
>     You'll notice that I used a
>     Class + ClassHelper pattern.
>     So I can have all the methods public on the Helper,
>     thus I could easily test them all.
>     I asked others one the JUnit list if there are any
>     security risks to having all public methods.
>     They said not really. 
> if you ask people who never add any accident driving without a seat 
> belt, they will tell you that the risk is very limited. private 
> methods is a kind of seat belt you like to have when you hit a wall.
If God (See below) wanted all methods public he would have not inspired 
his prophet Gosling to make some private. 

Seriously the more private our classes the less external coupling, the 
less coupling the faster we can adapt to change.  We have a large code 
base right now.  The more it grows without introducing the idea of 
decoupling the harder it will be to change it in the future.  Adding 
more classes for testing and introducing the idea of all public methods 
is moving in the opposite direction then IMHO we need to go.

> IMHO, trying to cover all the code with test (ie, 100% test coverage) 
> is useless, because it's simply impossible to guarante you will get a 
> 100% coverage (
> The more that we can do is to do our best.
> Now, that being said, about helper classes :
> 1) a method is private because we don't want to expose it to the 
> public. When you write such methods, as you can't test them easily, 
> you have to be very carefull. The parameters should be tested, and 
> 'assert' is a good way to do so. At some point, ypou can also test 
> those methods simply by testing the public method which use it.
> 2) having 2 classes instead of one drive you to double the number or 
> bugs. This is mathematic.
> 3) Sometime, tests are simply wrong. You may have to double check the 
> tests, and fix them if needed. If you have a classhelper  in between, 
> you may have to check two classes instead of one (the test plus the 
> helper class). This is adding some burden.
> 4) Having two classes to maintain is really problematic. The more 
> classes you have, the more work you will have to do. All this work 
> won't be put into the single class, so one can think that this 
> helperclass drives the developper to loose focus on what is really 
> important.
> 5) If the developper has choose to declare a method as private, 
> public, protected, or with a default protection, it is because he 
> sometime has good reason. This is somehow the basis of OO design. I 
> don't want to loose this, I don't want to go back to C or, worse, Basic.
> 6) From a user friendliness point of view, I really like to see at 
> first glance which are the public methods of a class (they are green 
> in Eclipse).
> Ok, now, this is not only my opinion. It's also somthing I will push 
> strongly if voted. I must tell that I will veto such a decision (to 
> have all methods public) if asked. Of course, if i'm the only one to 
> veto such a decision on the project, I will dismiss, naturally. I'm 
> not threaten anyone, I'm just saying that I don't think we should 
> change the way we are coding. I try to be pragmatic : we are writing 
> an Ldap Server, not a giant unit test system around an LdapServer. We 
> are already have somthing like 3 000 tests, and we add a few every 
> single day, I'm confortable with it.
> Trying to reach perfection is insane. Even God can't do everything 
> perfect : he created mankind :) (ok, for people like me who don't 
> believe in God, think about the Chaos Theory... )
Emmanuel your viewpoints do not appear to originate from a chaotic 
source, nor I think, if we followed them would lead to a chaotic source. 


View raw message