Hey, I just want to point out that this seems like a big gap. “Just invert the values negatively” isn’t a great answer when you’re sorting complex objects with custom comparitors. What, I’m just supposed to invert my gt and lt methods in an expensive wrapper class? What if I’m trying to use the objects I put in the heap for other purposes? Other algorithms use heaps. What if I’m writing an A*-like heuristic function (actually, I have a lot of these, hundreds and hundreds of lines of code of tens of implementations of variations on these for game AI) and need to maximize a heuristic? I have to take my heuristic val and invert it before putting it in the heap, resulting in a ton of overly complex (and computationally wasteful) conversion code, just to get the value back out and invert it again before doing all the other value comparisons that need to be done. I’m surprised more people aren’t complaining, I guess not that many people write complex heuristic searches in python?
I got sick of the overly complex and obtuse negation in my value functions (over 2x the amount of code is necessary because of missing max-heap, I need a heuristic func that produces the actual values for other things to use, and then a separate heuristic func that chains the output from the value func and inverts each property individually for the heap-based search), and started switching to use heapq_max · PyPI instead for much, much cleaner code for my (many) use cases where the search heuristic and value heuristic are the same, but it is so old (or inefficiently compiled?) that the performance is about 100% worse than the min-heap in heapq. I’ve tried heap-class · PyPI as well but its performance is even worse than that, for both min and max heaps.
I’m using python because I dont want to write C (nor figure out how to get my C compiled as fast as heapq is, as well as figuring out all the interop stuff…)
Please, just give us high performance max-heap methods from heapq 
it seems so strange for such an important core data structure from core algorithms to be completely missing.