Re: [Refdb-devel] xnote links

[Damien J. D. <D.J...@cs...>]

Gidday gidday

Sorry if I'm a bit obscure. I'll try and clarify what I meant.

By non-directional links (at a user interface level) I meant only that 
if I link note A to note B then note B would automatically be linked to 
note A.
A directional link would be that if I link note A to note B then when I 
query note B I don't get back any reference to note A.

At an implemention level non-directionality means that rather than 
having note A "link to" note B and, conversely, having note B "linked 
from" note A, you would have note A "link to and linked from" note B. 
This would involve an extra table.

Now, you can have non-directional links at a user interface level but 
not at the implementation level. If this was the approach, you would 
need two distinct query processes, finding all those notes linking to 
your current note and finding all the notes linked from your current 
note, and then exclude any duplicates. Which is fine, if you're okay 
with a bit of potential duplication in your DB and/or manual constraint 
enforcement. I guess I'm just being a pedantic relational purist here 
(wherein the relational DB enforces this kind of constraint rather than 
the app).

It also depends on what behaviour we want - I really don't care which we 
get at the user level and it only seems that directional links are 
easier to implement without adding in, either extra database redundancy 
management code as discussed in the previous paragraph, or a new table.

Anyway, I shouldn't argue with you on the implementation details! I'm 
just happy that you're considering putting the feature in!

Peace
Damien

Markus Hoenicka wrote:
> Hi Damien,
> 
> probably I'm a bit dense here, but I think that links via the t_xnote
> table are bidirectional by nature. It is just a matter of the proper
> query. We've got the same issue with the t_refdb, t_author, and
> t_xauthor tables. I can start with a name and find all
> references in t_refdb that use that name as an author. I can also
> start with a reference ID and retrieve all authors of this
> reference. Speaking in terms of extended notes, "links to" and "is
> linked from" should be pretty much the same story. Or am I missing
> something here?
> 
> regards,
> Markus
> 
> Damien Jade Duff writes:
>  > I think I know what you mean - you'd allow presumably a NOTE string in 
>  > the xnote_type field of the t_xnote table. Though I think that would 
>  > only allow for unidirectional links unless you were happy to let your 
>  > backend try to keep two links updated (I believe this is sort of frowned 
>  > upon in the relational DB world) or to search two indices, note_id and 
>  > xref_id for a particular note_id (maybe okay but seems clumsy).
>  > 
>  > I could handle non-directional links outside of RefDB by automatically 
>  > generating dummy references to group notes with but it's kind of messy, 
>  > particularly as there's no real reason to group notes in groups bigger 
>  > than 2 (since you could then group notes by linking them all to another, 
>  > group note, anyway).
>  > 
>  > On the other hand, while non-directional links would be nice, 
>  > unidirectional links would be just fine as far as I'm concerned. It 
>  > would be up to the DB user to specify if he/she wanted a link going both 
>  > ways.
>  > 
>  > As far as a GUI tree view on the data is concerned, well it may recurse 
>  > sometimes, but that's just a consequence of choosing a tree to view a 
>  > potentially cyclic directed (or undirected) graph.
>  > 
>  > Feel free to talk me out of it!
>  > 
>  > Peace
>  > Damien
>  > 
>  > -------------------------------------------------------------------------
>  > Take Surveys. Earn Cash. Influence the Future of IT
>  > Join SourceForge.net's Techsay panel and you'll get the chance to share your
>  > opinions on IT & business topics through brief surveys -- and earn cash
>  > http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
>  > _______________________________________________
>  > Refdb-devel mailing list
>  > Ref...@li...
>  > https://lists.sourceforge.net/lists/listinfo/refdb-devel
>  > 
>