libera/#commonlisp - IRC Chatlog

11:10:18 paulapatience jackdaniel: So, it only transits data through its own streams in some cases. On SBCL and CMUCL on Unix, stream arguments are passed directly to the underlying implementation. On other implementations it passes them through UIOP's stream functions.

11:10:32 paulapatience But pathnames are passed to the underlying implementation.

11:10:54 paulapatience I think the reason for the streams is that not all implementations support that.

11:11:57 paulapatience E.g., I think passing octet streams to some implementations is unsupported, but theoretically with vomit and slurp it would be, because they get converted to files that are then ok.

11:12:03 paulapatience But I'm not entirely sure yet.

11:23:20 jackdaniel in any case I've always been sceptical regarding uiop's run-program/launch-program design; I've studied it a little, but that was a few years ago

11:32:46 jackdaniel I wish there was a portable way of specifying by the programmer what is executed under certain optimization levels

11:34:41 jackdaniel s/executed/compiled/

11:36:34 jackdaniel but that's more a golden fish wish than a must have, so.. :)

11:40:41 paulapatience The optimization level used by run-program? Or were those two disjoint ideas?

11:44:07 edgar-rft` ** NICK edgar-rft

11:46:11 edgar-rft I'm afraid that run-program can't optimize the programs it runs

11:46:51 jackdaniel disjoint, just a loose thought inspired by what I'm doing at the moment

11:50:08 edgar-rft but it would be cool to optimize programs just by running them from Common Lisp :-)

11:52:18 jackdaniel I've meant something like (optimize-case ((and (< safety 1) (> speed 3)) nil) ((> safety 2) (check-type x float)) (otherwise (float x))) ; lousy example, I know

11:53:04 jackdaniel technially something what compilers already do have for their own inner workings (in ecl it is called compilation policies)

11:53:44 jackdaniel and they are named there btw, but hopefully you know what I mean

12:11:38 paulapatience Th

12:11:57 paulapatience I've seen the policy-cond library that does something like that

12:12:04 paulapatience It's used in magicl

12:12:26 paulapatience (The linear algebra library)

12:12:38 jackdaniel ACTION ported magicl to ecl at some point

12:12:58 paulapatience Oh cool

12:13:05 jackdaniel as part of a gig from rigetti computing

12:13:43 paulapatience I've been keeping an eye on that for when I need to do linear algebra, but the necessity has not arisen

16:14:16 alcor` ** NICK alcor

20:23:41 NotThatRPG paulapatience: Thanks for filing those issues.

20:37:42 paulapatience NotThatRPG: My pleasure.

20:40:25 NotThatRPG It's really helpful; wish I could fix these things more rapidly...

20:41:14 paulapatience I mean, I could submit a PR for both of those if that would help. Is there anything specific to do beyond submitting the PR? Like tests and such?

21:27:10 HER_ ** NICK HER