freenode/#sicl - IRC Chatlog
Search
16:23:45
heisig
I might still tweak it a little. I realized that while some functions (FIND, POSITION) should always be inlined, others (SORT, SEARCH) should not.
16:25:27
heisig
- In terms of functionality, I only wrote a few demo functions (FIND, FILL) to check whether my technique is viable.
16:27:27
heisig
- The exception are functions like SORT, and the manipulation of sequences (MAKE-SEQUENCE-LIKE, ADJUST-SEQUENCE) from the paper by Xof.
16:31:07
heisig
Oh, since I want to write a paper about sealable-metaobjects, I will definitely finish the sequence code before the next ELS.
16:33:52
beach
So I should just go ahead and delete the existing code. I don't think I need the native sequence functions urgently. I can import them from the host for now.
16:35:54
heisig
Everything I wrote is in the Code/Sequence/Generic directory. The ASD file is Code/sicl-sequence.asd .
16:44:55
heisig
I was already joking to some people that SICL's FIND is approaching one publication per line of code :)
16:45:52
beach
Seriously, that's exactly the kind of code I would like to see in SICL: small, to the point, efficient.
16:46:18
beach
So it is worth pondering different solutions for a while, and also to document things in papers.
16:50:39
heisig
I can already warn you: To make things fast, I will need a way to query the compiler's knowledge about the types of the arguments of each call to a sealed generic function.
16:51:08
heisig
I can use compiler macro's and Trucler, but then the results might be underwhelming.
16:53:02
beach
Nothing in SICL is going to be fast from the start. And that's by design. I want to find out where I need to put in the most effort first, before I attempt to guess something that might not turn out to be true.