Post #9by Janus » 11.04.2020, 16:23
Scratch that.
After getting a good look at what has been done, I am rewriting it for my own fork.
The loading routines now use fs::recursive_directory_iterator which makes a very effective random loading order function.
This is the sort of thing I normally remove when I find it, along with any other stateless stuff.
Random loading order is a source of chaos, which makes tracing/troubleshooting very hard.
To be fair though, most of my work is low level, where this sort of high level fluff is just wasted cycles.
It also guarantees you have no way of really knowing what is happening, as it is happening.
Mostly I work in the Mhz range, not Ghz, with KB or MB, not GB, 1 core, not ??? many.
So I am used to thinking in terms of how long is this going ot take, and what else is happening while it does.
I also freely admit to a dislike/mistrust of this sort of 'Trust in the compiler' thinking.
For me, the simpler you can make something, the better.
To me, this sort of abstraction just shoves the complications under the rug.
I have nothing against object oriented programming (OOP) at all.
I use it, more than I like admitting, but it has its place, and filesystem stuff is nominally one of those.
But stateless is never an efficient answer, and seldom traceable.
I will be writing my own alphabetical loading routine.
Once I have it, I will post a build in my forking thread, along with code.
If anyone wants to use it, they are welcome to.
I also renew my warning, I am not a real C/C++ programmer.
So whatever code I post will not be pretty, just functional.
Janus.
