"This implementation has received extensive optimization work, resulting in over 110x reduction in runtime since optimization work started. The only obvious way to optimize it further would be to integrate the logic directly into odgi in order to give it direct access to the graph object model in memory, removing the need for file I/O."
This would mean at least partially rewriting component segmentation in C++. Since Python compiles to C, much of the code could be maintained and possibly interoperable at the expense of a more complex compiling environment.
"This implementation has received extensive optimization work, resulting in over 110x reduction in runtime since optimization work started. The only obvious way to optimize it further would be to integrate the logic directly into odgi in order to give it direct access to the graph object model in memory, removing the need for file I/O."
This would mean at least partially rewriting component segmentation in C++. Since Python compiles to C, much of the code could be maintained and possibly interoperable at the expense of a more complex compiling environment.