03:28 <cr1901> Dunno where to put this, but this is the only python project I know that uses metaclasses, so... I actually had a legit use for a metaclass tonight: http://gopher.wdj-consulting.com:70/paste/ef1c0157-78ec-466e-9500-6123be550a01.txt Requesting comments on cursedness

07:19 <whitequark[cis]> that looks very cursed

07:20 <whitequark[cis]> why not make it a toplevel for loop that does setattr on a class object?

15:22 <cr1901> I abandoned the class approach after sleeping on it, but yea that approach would be fine if the setattr loop is deferred to after tasks are loaded by doit (I think the top-level would run too early for the default task loader)

15:40 <cr1901> Actually, I don't think setattr loop works- if you do it before task loading, the both the class object and instance have the create_doit_tasks attribute, and if you do it after task loading, doit skipped over the objects due to lack of create_doit_tasks attribute. Or am I misunderstanding.

22:12 <cr1901> (FWIW, if anyone's following along, I fixed all the things I was unhappy with in my code by using decorators and higher-order functions with control flow skeletons rather than classes)

22:17 <tpw_rules> cr1901: making your own snapshot engine? is it public?

22:18 <cr1901> tpw_rules: Not yet, will be soon. Also it's very limited (zfs full snapshot from one pool to another to initialize, and then inc snapshots thereafter). doit is in there to guard against "accidentally rotating out snapshots more than once per backup"

22:21 <tpw_rules> i had to write my own to handle proxmox VMs and i've been looking to kill it