John,

I am not sure I completely understand your problem.

I just recently finished a project in which we go from xpp to word in which we had to output things in the same order as they were vertically positioned on the page (2 column job with frozen pickups and sometimes frozen main blocks next to each other). I was using the divxml output for that and found all the necessary information in there.

divxml outputs the same information as you can find as system variables as compose.
It will give you the starting x,y coordinates of the pickup stream, the x, y coordinates of each block inside the pickup and the y position of each line. And it does the same for the main text blocks and line.
Using that information it should be perfectly possible to determine the exact position of each thing on the page.

I am guessing that your problem is:

John Hoffman said:

We try looking down in the the <group> and <line> attributes, but there are a lot of those tags in there, and we can't rely on taking the values from line[1]/

But isn't it possible to look inside the <line> and look at the <t> elements (which you have to do anyway as they contain another set of x and y coordinates) and find the first one that contains actual text?

Can you explain why this is a problem to you (maybe with a small excerpt from the divxml result)?  

John,

I am not sure I completely understand your problem.

I just recently finished a project in which we go from xpp to word in which we had to output things in the same order as they were vertically positioned on the page (2 column job with frozen pickups and sometimes frozen main blocks next to each other). I was using the divxml output for that and found all the necessary information in there.

divxml outputs the same information as you can find as system variables as compose.
It will give you the starting x,y coordinates of the pickup stream, the x, y coordinates of each block inside the pickup and the y position of each line. And it does the same for the main text blocks and line.
Using that information it should be perfectly possible to determine the exact position of each thing on the page.

I am guessing that your problem is:

John Hoffman said:

We try looking down in the the <group> and <line> attributes, but there are a lot of those tags in there, and we can't rely on taking the values from line[1]/

But isn't it possible to look inside the <line> and look at the <t> elements (which you have to do anyway as they contain another set of x and y coordinates) and find the first one that contains actual text?

Can you explain why this is a problem to you (maybe with a small excerpt from the divxml result)?  

Other than looking to the x/y location of the first real "text" in the main block versus the pickup to see which is "first", I think you're probably safe in assuming that if the x/y of the main block and pickup stream match then the pickup is first (and has effectively pushed the lines in the main block down below the pickup).

Your kind of doing something outside of the "scope" of XML Export (divxml), which was designed to give enough information to duplicate pages, i.e. the "placement" and "characteristics" of the data. I don't think we considered anything specific about the "order" (e.g. of main text versus pickup text) in the design. There really isn't inherently any "order" of the two streams - they are somewhat independent (although they affect each other) and they are just "placed". And obviously which is "first" on a page (according to your definition) can vary.

Jonathan Dagresta
SDL XPP Engineering

I understand, so there is some risk in this that needs to be flushed out. 

The primary intent of placing the text in document order is to support the conversion to MSWORD/HTML with formatting characteristics obtainable from the divxml markup.

Cheers, 

John Hoffman