freenode/#sicl - IRC Chatlog
Search
12:56:40
heisig
Please do! I am very grateful for your remarks. Especially that time where you pointed me to Olin's elegant trick for merging lists.
12:57:20
heisig
And I think there are several other brilliant tricks hidden in some Scheme implementations that are useful for SICL.
13:00:30
heisig
Encoding the merge state (whether the last element was from list A or list B) by having two recursive functions, and just CDRing along them.
13:01:07
heisig
So when more than one consecutive element is taken from the same list, there is no need to emit a write to memory.
14:39:56
jcowan
Oh, a counterexample to "always write so as to minimize prescriptive spasms of revulsion from the reader":
14:40:18
jcowan
Consider "The patient should always be informed in advance how much ____ treatment will cost."
14:41:14
jcowan
Each of "their", "his", or "his or her" are guaranteed to produce such objections in a substantial group of people.
14:41:48
jcowan
There are similar problems with "(American) Indian" and "Native American"; even the referents don't agree on which term is acceptable and which is unacceptable.
18:09:55
Bike
a problem with the disembodiment/dfun thing is that another thread could change-class an argument to a method, or whatever, at any time, so if you want to be real safe about it you have to do a type check, and that's pretty much what discrimination is
18:50:55
Bike
i did a quick histogram of how call history lengths are distributed. in my clasp there are 921 generic functions. 441 have no call history, 229 have one call history entry, 82 have two, etc. decreasing
18:51:18
Bike
but the ones with long call histories are important ones like cleavir-ir:inputs and shared-initialize