API-suggestions » History » Version 4
Version 3 (C. Emmerich, 02/08/2012 01:31 PM) → Version 4/6 (C. Emmerich, 02/08/2012 07:22 PM)
h1. API-suggestions
* incomplete documentation: Buffer->get(), Buffer->tryGet() - where is the difference
* improve naming for factory-methods, e.g. JointAngles::fromRad can be renamed in JointAnlges::createFromRad
* some questions concerning the get() and publish() logic of CCA::Ports:
** +get()+:
*** Why do I get a pointer to the buffer's content? Wouldnt it be more correct to - at least - get a pointer to a const object? Because otherwise I am allowed to manipulate the buffer's content.
*** Even if i get a pointer to a const object, there is still the problem, that buffer's content can change while I am using it and I will not be informed about that change
** +publish()+"
*** same for the const correctness: why dont I publish a pointer to a const object
** As a user, I would prefer to not work with pointers
*** problems amongst the mentioned ones:
**** it forces me to work with shared_pointers even if it's not necessary at all
**** it can produce overhead
* print-methods for DTOs: plz do not append a newline() command at the end of printing a DTO (see e.g. JointAngles.print()). this is confusing, because the 'user' himself wants determine a when a new line starts, or he wants to print two of them in one line what is not possible with your current printing style
* incomplete documentation: Buffer->get(), Buffer->tryGet() - where is the difference
* improve naming for factory-methods, e.g. JointAngles::fromRad can be renamed in JointAnlges::createFromRad
* some questions concerning the get() and publish() logic of CCA::Ports:
** +get()+:
*** Why do I get a pointer to the buffer's content? Wouldnt it be more correct to - at least - get a pointer to a const object? Because otherwise I am allowed to manipulate the buffer's content.
*** Even if i get a pointer to a const object, there is still the problem, that buffer's content can change while I am using it and I will not be informed about that change
** +publish()+"
*** same for the const correctness: why dont I publish a pointer to a const object
** As a user, I would prefer to not work with pointers
*** problems amongst the mentioned ones:
**** it forces me to work with shared_pointers even if it's not necessary at all
**** it can produce overhead
* print-methods for DTOs: plz do not append a newline() command at the end of printing a DTO (see e.g. JointAngles.print()). this is confusing, because the 'user' himself wants determine a when a new line starts, or he wants to print two of them in one line what is not possible with your current printing style