I eagerly look forward to this work. Since you are far more familiar
with the intent of the code, you can be consequently more aggressive
in replacing full tree walks with walks that visit tnodes or vnodes
only once, with GREAT benefit to the nimbleness of Leo with my data.
I gave instructions for constructing a kind of degenerate case outline
where such improvements are heavily exercised; but would it help
motivate you if I posted something closer to a real world example?
- Stephen
On Apr 16, 8:12 am, "Edward K. Ream" <[EMAIL PROTECTED]> wrote:
> The unified-node "eureka" has turned out unexpectedly. Indeed, it is
> now of mostly theoretical interest:
>
> 1. It shows the proper way to understand the interplay between
> positions and nodes.
>
> 2. It shows that tnodes should be considered subsidiary to vnodes. In
> particular, the tnode-based iters probably should be retired. That
> would be simply a matter of adding a 'unique' keyword argument to all
> the iters. The default should be True, imo. Also, all references to
> tnode ivars should be eliminated where possible (converted to position
> getters/setters) or converted to tnode getters/setters. These "raw"
> references to the tnode ivars is the "bad style" I referred to
> yesterday.
>
> 3. It shows that the list-based implementation of nodes (using parents
> and children arrays) can be done regardless of whether dual or unified
> nodes are used. Furthermore, my work on the unified-node branch shows
> that fully populated position stacks simplify Leo's fundamental code
> considerably, again, regardless of how nodes are organized.
>
> 4. It shows that the issue of whether clones are truly independent or
> not is of little practical significance. That's the reason that the
> decision is so difficult to make :-) That being so, we shall certainly
> stick with the present vnode/tnode based scheme, at least for now. In
> particular, this question has very little relevance to the graph
> world.
>
> 5. It shows that issues of file format are largely independent of how
> nodes are organized. Moving to sax-based read code has made this
> clear.
>
> What an interesting turn of events.
>
> I said yesterday that I shall soon move on to more pressing matters,
> but imo it is imperative that I fix the obvious problems in the
> unified-node branch. Leaving code lying around in an unfinished state
> is just unbearable. And ultimately, it is a big waste of time. So
> I'll spend however long it takes so that all unit tests pass in the
> unified-node world. I do hope it won't take more than a few days...
>
> Edward
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"leo-editor" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/leo-editor?hl=en
-~----------~----~----~----~------~----~------~--~---
