Brad Lanam

FYI: The new epoll() code does not work correctly. https://core.tcl.tk/tcl/tktview?name=1a6a36d901

Looks much better. I ran the busy-wait test, use of dbus package test. I don't have a specific test built for the disconnected socket, but I pulled up an older...

It was the intention of myself and Andy Goth to merge his work and mine to ensure that there would be a single C API. I think this is the...

Yes, it should save a little time, but I don't think the end result is much different from the Tcl code, excepting the overhead of the set call (I could...

No, that is not correct. No, it is not a stock version of `foreach key ....`. Even after bytecoding, I don't think you are going to get any huge speed...

Right. I think once bytecoded, there will some time savings, but nothing significant. The fetch of the value object from the hash table (which has all the trace checking, etc.)...

Well, as DKF hinted, We can: - change arrays to have a global epoch counter each time the array has an insert/delete operation. - not execute any traces on the...

It appears that `array get` does not fire any read traces. I will take a look and see what it is doing.

No, I was incorrect -- it fires the trace on the individual element. A read trace on the entire array does not fire (don't know if that ever does).

`array get` is also using `Tcl_GetVar2`. `dict for` has a speed of 5447.601 microseconds in my same benchmark. Can you simply convert to using dictionaries? You could also keep an...