Re: Tclkit vs Tclkit Lite VFS performance

To follow up again ...

> It is indeed.  When I change this:
> foreach {n s d c} [eval [linsert $tag end *]] break
> to this
> foreach {n s d c} [{*}$tag *] break
>
> ... then the timings become normal again [...] So *something* in  
> eval is shimmering, even though its argument is a pure list.

I'm going to blame this on Tcl until told otherwise.  Have uploaded a  
new vqtcl.tgz with a workaround in mkclvfs.tcl which avoids the  
problem.  Here's what I see now with mingw/8.5:

     $ tclkit-cli.exe ../../vfsperf.tcl
     kitten C:/msys/home/jcw/kitten.kit
     sdx C:/msys/home/jcw/sdx
     tclkit 8.4.13 : C:/msys/home/jcw/tclkitsh.exe
       lite 8.5a6 : C:/kitgen/8.5/base-std/tclkit-cli

     tclkit {} : 48439.0
       lite {} : 29041.8
     tclkit pkgreq : 49740.1
       lite pkgreq : 54942.2

     tclkit stat : 106.66
     tclkit stat : 196.42
     tclkit mtime : 90.402
     tclkit mtime : 216.237
     tclkit glob : 1279.03
     tclkit glob : 460.64
     tclkit open : 140.06
     tclkit open : 434.82
     tclkit read : 345.93
     tclkit read : 578.88
     tclkit read-large : 4571.79
     tclkit read-large : 7580.36

Am going to leave the remaining slowdown as is for now - that  
optimization really needs to be done in the Vlerq extension, not as a  
hack which adds more Tcl-based caching.

-jcw