I have code that is leaking, I been trying to figure out what I am doing wrong. I created a test example and I am about to reproduce in tclsh

set holder {}

for {set i 0} {$i < 1144195} {incr i} {
set attr [dict create]
set all_keys {"X" "P" "A" "B" "T" "Y" "E" "W" "B"}
foreach key $all_keys {
dict set attr $key [expr int(rand()*20)]
}
lappend holder $attr
}

This eats ~2GB of RAM, using Tcl 8.5. If I then do set holder {} , after which in my head would release the holder object and all of its things, instead I see Tcl still using that ram and not letting go. What am I doing wrong here? I have similar example I did in Python and the moment I set holder in Python to None, everything is freed, but in Tcl I cant see to actually get my memory back.

I am on mac btw, not sure if this would do anything.

Thanks,
Anatoly

Anatoly Torchinskiy schrieb am Donnerstag, 30. November 2023 um 04:33:37 UTC+1:
> I have code that is leaking, I been trying to figure out what I am doing wrong. I created a test example and I am about to reproduce in tclsh
>
> set holder {}
>
> for {set i 0} {$i < 1144195} {incr i} {
> set attr [dict create]
> set all_keys {"X" "P" "A" "B" "T" "Y" "E" "W" "B"}
> foreach key $all_keys {
> dict set attr $key [expr int(rand()*20)]
> }
> lappend holder $attr
> }
>
>
> This eats ~2GB of RAM, using Tcl 8.5. If I then do set holder {} , after which in my head would release the holder object and all of its things, instead I see Tcl still using that ram and not letting go. What am I doing wrong here? I have similar example I did in Python and the moment I set holder in Python to None, everything is freed, but in Tcl I cant see to actually get my memory back.
>
> I am on mac btw, not sure if this would do anything.
>
> Thanks,
> Anatoly
Afaik tcl is maintaining an own list of used memory and will reuse it.
Just put your code in a proc and run it. The holder var is then local and will be deleted at the end of the proc.
The memory will grow in the first call and then stay nearly the same in further calls.

HTH
rene

On Thursday, November 30, 2023 at 4:33:37 AM UTC+1, Anatoly Torchinskiy wrote:
> I have code that is leaking, I been trying to figure out what I am doing wrong. I created a test example and I am about to reproduce in tclsh

Do you have tclreadline in use? If so, the history could be the culprit holding its own copy of Tcl objects.
You could cross-check the refcount of $holder with ::tcl::unsupported::representation before resetting it to empty.

On Thursday, November 30, 2023 at 2:02:03 PM UTC-8, heinrichmartin wrote:
> On Thursday, November 30, 2023 at 4:33:37 AM UTC+1, Anatoly Torchinskiy wrote:
> > I have code that is leaking, I been trying to figure out what I am doing wrong. I created a test example and I am about to reproduce in tclsh
> Do you have tclreadline in use? If so, the history could be the culprit holding its own copy of Tcl objects.
> You could cross-check the refcount of $holder with ::tcl::unsupported::representation before resetting it to empty.

You are correct putting it in a proc:

proc test {} {
set holder {}

for {set i 0} {$i < 1144195} {incr i} {
set attr [dict create]
set all_keys {"X" "P" "A" "B" "T" "Y" "E" "W" "B"}
foreach key $all_keys {
# dict set attr $key [expr int(rand()*20)]
dict set attr $key "a"
}
lappend holder $attr
set attr {}
}
}

and then calling test multiple times, doesnt just add memory, but I also never reclaim memory. This makes it seem like memory will just always grow and it will never get reclaimed which is a bit weird. Yes that memory could be re-used but in my situation I expect to have this object as transient, so I would like that memory to actually be gone.