Reputation: 13
What does eval
mean in the following code?
proc poissontraffic { src dst interval starttime } {
global ns_ node_
set udp($src) [new Agent/UDP]
eval $ns_ attach-agent \$node_($src) \$udp($src)
set null($dst) [new Agent/Null]
eval $ns_ attach-agent \$node_($dst) \$null($dst)
set expl($src) [new Application/Traffic/Exponential]
eval \$expl($src) set packetSize_ 70
eval \$expl($src) set burst_time_ 0
# idle_time + pkt_tx_time = interval
eval \$expl($src) set idle_time_ [expr $interval*1000.0-70.0*8/250]ms
eval \$expl($src) set rate_ 250k
eval \$expl($src) attach-agent \$udp($src)
eval $ns_ connect \$udp($src) \$null($dst)
$ns_ at $starttime "$expl($src) start"
}
Upvotes: 0
Views: 688
Reputation: 137787
The eval
command concatenates its arguments and evaluates the resulting string as a Tcl script. The concatenation is done by stripping spaces from either end of each argument and then joining them together with a single space between.
In the given code, for example,
eval $ns_ connect \$udp($src) \$null($dst)
effectively does two rounds of substitutions, once to fill in $ns_
, $src
and $dst
, and again to read from $udp(...)
, $null(...)
, and run a command.
You can probably replace it with this:
$ns_ connect $udp($src) $null($dst)
That'll only be a problem if $ns_
is a multi-word value. Since this is fundamentally OTcl where object names are usually well-behaved, it probably isn't and all the eval
s will be adding is confusion, slowness and insecurity.
Upvotes: 1