Oh boy was I confused last night. Got some rest. Remembered that Ai Austria sent me a great example of dividers in the Will Wallace XML, remembered that I have six gillion lines of code for parsing credits in BulkEdit.

So now, how to do it with XPath.

For the programmers, the XPath extensions gives you the ability to participate in the iterating loop without having to manage the loop yourself. The XPath iterator calls your code and passes you the node that has passed the high level selection criteria. Then you can do what you want. This is where I perform the famous Ai Austria concatenation, double blanks elimination, priority to CreditedAs and optional BY inclusion in the search.

Problem is, the node is passed raw, out of context. There are no parents, siblings or anything else. You just get the Outer and Inner HTML, and the ability to process attributes.

So I think what I might have to do is add to the top level search, so that dividers are also passed to the iterator, so I get the divider nodes in the correct sequence along with the actor and credit tags, where I can keep track of groups and episodes. That should do it.

Again, this is only for the stripped down CLTBoss/CLTInfo style XML. More compact for anyone who wants "just the facts ma'am, just the facts" and does not need the other profile info, or the bother of parsing the credit dividers.

For the Invelos style output, it is easy: XPath says - "yup, you got credits in this profile", so I just grab the whole profile XML and glom it onto the putput queue. No parsing needed.

I want exactly the same Xpath code to supply both, so that when a problem occurs, I can fix it in one place.

Quoting mediadogg:

Quote:
Cleaning up dozens of loose ends. Each two I clean up creates one more.

That sounds uncomfortably familiar. But you're not really a developer if you never have had that happen to you, I think.

Quote:
One for Gunnar. The CLTBoss credits output, as requested, will be close to Invelos format.

Appreciated! 

Quoting mediadogg:

Quote:
Quick questions, will save me research: Can dividers be nested? I.e., can an Episode have Groups? Groups within Groups?

Yes and yes.

Quote:
Has anybody seem a "Team" divider?

Not sure what you mean.

Quote:
Another thing: I am looking at a particular actor who is often CreditedAs "Writer". Does this mean someone has likely to have incorrectly submitted the credit?

Sounds fishy. Example?

By the way, could I get a couple of profile IDs with nested dividers - as complex as possible, for testing?

Someday I need to go back and make those changes to other plugins (or maybe it is in there - I'll get (more) confused if I look at other code at this point).

Quoting GSyren:

Quote:
I guess that Team dividers is something that never left the planning stage. I'm sure that it was never implemented.

For an example of mixed Episode and Group dividers you could try 3-259190-274396 ('Allo 'Allo: Series 1 & 2).

I don't think I have ever seen Group dividers nested in more than 2 levels. I'll look for some example, but it'll have to wait until tomorrow. The sandman calleth ... 

Many thanks. Anything more than 1 level is enough. I implemented simple stacks for both Cast and Crew, which should handle any arbitrary depth or mix.

Yes, great. Although I am still not confident I understand it. And the forum posts I've read indicate that I am not the only one. And even if the architecture is clear, the XML is entirely dependent on users manually inserting begin / end dividers, and there are a lot of end dividers missing.

I think you've got it, but ...

There may be old profiles from before multiple levels were supported. Those can have groups that are supposed to be at the same level, but no longer are because the groups lack end dividers. I don't think there's much that can be done about that, except someone correcting those profiles.

Also, end dividers are not used for episodes, only for groups. In the widow for adding dividers they are explicitly called "Group End".

Oh man, is the Invelos database a mess!

The first UPC, locality 0 yields credits with nested dividers only because the submitter did not end the groups.

You would not expect that "Officers", "Crew", and "Others" would logically be nested.

I randomly picked one of the actors from that one, and did a CLT and bumped into this profile 4030521181533.5, which has exactly the same Captions (Officers, Crew, Others) assigned to Episode dividers. I thought I had a bug until I looked at the original profile. Clearly the submitter mistakenly used Episode dividers for what should have been groups.

I wonder if either my tool or yours should give the option to "override" such mistakes and examine the results to see if it makes sense to schedule a profile update?

- substitute Episode for Group or vice-versa
- supply missing end dividers

Does this call for a "Divider clean-up tool"?