Re: [Tcllib-devel] Graph Manipulations

[Andreas K. <and...@ac...>]

Michał Antoniewski wrote:
> Hello.

> 3. Generally, I'm implementing algorithms as normal procedures. Then 
> after writing tests and checking if those procedures work well, they 
> should be placed in tcllib ( just like hierarchy at SVN says). Do you 
> accept this approach?

Having seen the code now I believe I begin to understand what you are doing and 
asking here.

	In essence you are writing each algorithm X as a set of global
	procedures in a file X.tcl, including basic test cases at the bottom of
	file.

	Then, when you are satisfied with the implementation you plan to
	transfer the set of procedures to the struct/graphops.tcl file, and
	rename them so that they are in the proper namespace. The test code you
	further plan to transform into a set of regular test cases which are
	then put into a .test file under struct/graph/tests directory.

Is this understanding of your planning correct ?

The question you should ask yourself here is

	What are the advantages, if any, of this approach over the alternative,
	i.e. of working directly within the struct/graphops.tcl file, and the
	tests files under struct/graph/tests ?

	Conversely, what are the disadvantages of this approach ?

One disadvantage, in my opinion, is that you are doing something in two steps 
(code, and transfer) where one step would be sufficient (code). This is a 
disadvantage in my eyes because the transfer step is simply one more place 
where bugs can be introduced. Does that make sense to you ?

Andreas.