plotname.cir

Enhancement-345: naming a plot no longer walks the plot list
* plot_alloc() and plot_add() pick a unique plot name by counting a shared,
* monotone plot_num up until <abbrev><plot_num> is not the typename of any plot
* in plot_list. That membership test was a walk of the WHOLE list with a
* case-insensitive compare, so naming a plot cost O(plots) -- and a sweep
* creates a plot per point, which made naming them quadratic in the sweep
* length. After Enhancement-343 removed the other quadratic term, profiling a
* 64000-point sweep put 89% of the run right here, in plot_alloc -> cieq.
*
* E-345 keeps a hash index of the typenames currently in plot_list, so the
* membership test is O(1). ONLY the test changed: the search still starts at the
* same shared plot_num and still counts up by one, so the names handed out are
* exactly what they were -- including the reuse of a number that `destroy all`
* frees, which a "remember every name ever issued" cache would have changed.
V1 in 0 dc 1
R1 in out 1k
C1 out 0 1p
.control
* the shared-plot_num sequence: note tran1 then op2, and ac2 not ac1
op
tran 1n 5n
op
ac dec 2 1 100
op
echo NAMES $plots

* destroying frees the numbers again, so the next plots reuse them
destroy all
op
op
echo REUSED $plots

echo SURVIVED
.endc
.end