TOPs, $F, and Timeshift

[Joseph_S]

Hi, first I just want to say that I'm a new poster here, but this forum comes up a lot when I look up issues with houdini and has a big help. Hopefully you all have some advice for my current question!

Recently I've constructed several HDAs that use the timeshift node to compare different frames in a sequence. I work on very large datasets, and use TOPs to parallelize and synchronize complex tasks. Here's my issue:

When you use a Rop node in pdg and reference a sop network that uses the $F expression and then timeshift, that graph works as expected outside of a pdg context -- file nodes with $F evaluate to the correct file for the work item's frame, and timeshift changes the file appropriately.

However, wrapping the same setup in a HDA, using the hda processor node, and passing the same file path via an input, a pdg attribute, or a hda parameter doesn't work. The expression gets evaluated before it gets sent to the hda, so there's no range to timeshift in, whether it's a file parameter or a string parameter. I know a couple of ways around this already -- you can use regex on the hda side to correct the path to an expression, prepare the necessary frames as multiple inputs on the TOP graph side, or escape the path in the topnet (ie myframe\$F.bgeo.sc rather than myframe$f.bgeo.sc) -- these all have drawbacks (brittleness, complexity, and error prone/tougher to hand off to users).

 

In my current work, I've chosen to escape the path, but I was wondering if any users here had encountered similar issues and had any advice or insight on how to handle parameter evaluation between TOPs and hdas.

 

Edit: Attached an example graph and screenshot. In grey, the source. In green, files that are successfully timeshifted (via rop geo top node, and hda processor with \$F.) In red, unsuccessful timeshift (hda with a regular $F parameter) 

F_question.zip

Edited October 31, 2020 by Joseph_S