Talk:Topological sort/Extracted top item: Difference between revisions

(J has top level now)
Line 5:
::Yep, it is a variation on a theme, but during development, it could be that all items could not be compiled but the items for one particular top level might be compilable. Extracting and compiling just the items for a particular top level allows work to continue. It looks as if the J output is OK. --[[User:Paddy3118|Paddy3118]] 20:38, 13 October 2010 (UTC)
::: When I first saw this, it brought to mind code optimizations where a relaxation of complex ordering rules could reduce processing time, yet still provide results good enough to satisfy core requirements in problem context. I'll be watching with interest to see where specializations for this task crop up. --[[User:Short Circuit|Michael Mol]] 20:46, 13 October 2010 (UTC)
:::: Hypothetically, we could attempt to implement a specification where we attempt to put "close to equal numbers of files" at each level. The concept of sorting against the dependency graph allows this kind of flexibility in some cases, and the example dependency structure is one of those cases. However, solutions to packing problems often have painful resource requirements for large data sets and/or cases where they do not produce the best results. --[[User:Rdm|Rdm]] 18:49, 15 October 2010 (UTC)
 
:: I spoke too soon. The J example needs to also extract the top levels. --[[User:Paddy3118|Paddy3118]] 20:40, 13 October 2010 (UTC)
Line 25 ⟶ 26:
I could change this to read determine ''and show'' the top levels if this would be more clear. Determining the top levels is a useful function in practice as top levels can be 'forgotten' in documentation if they are not primary, but may still remain useful. --[[User:Paddy3118|Paddy3118]] 21:49, 14 October 2010 (UTC)
 
:Ok, I have added "top level support" for J. However, I could not see any way to make this result relevant to the compile order, and it was quite simple to implement without reference to my dependency structure (the <code>depends</code> value), so I implemented it as an independent word. --[[User:Rdm|Rdm]] 18:40, 15 October 2010 (UTC)
6,962

edits