Here's some of the more interesting things that happened with TOPs / PDG since the last production build.
Full list is here [www.sidefx.com]

Custom Cache Handler API [www.sidefx.com] added

New feature for Partitioners: Split By Attribute [www.sidefx.com]

Python Script [www.sidefx.com] and Python Processor [www.sidefx.com] and friends now have auto-complete for their top-level variables.

When merging imported files Geometry Import [www.sidefx.com] can now optionally create an attribute to easily identify which file the pieces came from. Also added min/max bounds attributes to work items when importing by piece.

Added expected outputs and cache control UI to Command Send TOP [www.sidefx.com], as well as option to copy spare parms onto work items as attributes. Those values can then be used in the job script.

Python Script [www.sidefx.com] can now run during Generate.

Improvements to caching behavior:

Work items in PDG nodes that support caching will now always effect the cache state of downstream work items. For example, if a work item in a node with the cache mode parameter set to Automatic ends up writing a new file to disk, all downstream work items that also use caching will have their results invalidated. Those work items will also need to recook unless they have their cache setting set to Read Files.

There is also a new HOUDINI_PDG_CACHE_DEBUG environment variable that can be set to a value between 0 and 3 in order to get debug output about why PDG is or is not loading a file from cache. A value of 0 means the output is disabled, and a value of 3 is the most detailed output.

Multiple improvements to JSON Input [www.sidefx.com]

Task Graph Table now has access to the same RMB context menu as the TOP graph dots.

More authentication options added to Tractor Scheduler [www.sidefx.com]