List Commands Library for 50g

[snrowe]

(09-11-2017 05:14 PM)DavidM Wrote:  

(09-11-2017 01:53 PM)David Hayden Wrote:  Brilliant!

This is indeed a great option for avoiding the list/GC issues. Now that I've had a chance to run some tests with it, I've realized that there's still a potential "gotcha", though. In particular, it's not necessarily as memory-friendly as I was originally thinking.

The reason is that you will end up with (at least) two copies of the original list in memory for almost all scenarios while you are processing the list contents (assuming the list wasn't already saved to HOME). The first copy is the one saved in TEMP by the CKx... step of each command. Then the second is in HOME (by definition). Any altered elements will also be allocated in TEMP. So memory constraints can still play into this for large lists.

Avoiding that first copy in TEMP is possible, but it gets messy. Keeping it out of TEMP means it can't already be there when the CKx&Dispatch step is performed. Saving it to HOME prior to CK&Dispatch requires validating argument types before the "official" argument checking occurs, and has to be done carefully to avoid any errors being raised.

The payoff is great in the situation where lots of exploded list items exist on the stack and a GC occurs. Deciding (programmatically) how and where to apply that technique is not always straightforward, though. I'm still trying to figure out the best approach.

Thoughts are appreciated!

Hi David,

I was wondering about your large lists with your functions. I had this problem with the radio program you helped with. All the senarios I could try, ended out with a temporary copy if I desired to modify a large list.

Ended out using directories with to save large amounts of data. This way the directory could be modified with little memory opposed to creating a temp copy of a large list. If you get the list functions to so they don't require a temp copy while modifying it I'd be very interested.

Thanks,
Scott