freenode/#clim - IRC Chatlog

3:17:49 beach Good morning everyone!

3:54:32 loke Hello beach!

3:57:40 beach What's up?

4:00:04 loke Nothing much. Been thinking about how to deal with output records recordingtrnasformations

4:00:26 loke I managed to do it, but it was so ugly that I didn't even want to commit it as a work-in-progress (I don't want any record of me actually writing that code)

4:00:45 beach Heh.

4:00:49 loke https://gist.github.com/lokedhs/792f6c8b2a500f772be7a05032cafb22

4:00:53 loke There's the code.

4:01:26 loke Note how I use a dynamic variable to pass the override transformation to be used by the text rendering

4:01:41 beach Oh, I see, yes.

4:02:23 loke What I really need is to ament the MEDIUM-DRAW-TEST function to accept a transformation

4:02:35 loke I'm considering doing that right now.

4:02:55 beach I assume you meant TEXT, right?

4:03:00 loke Yes

4:03:03 loke Indeed I did

4:24:46 loke beach: Hmm...

4:24:56 loke I'm looking at the class COMPLETE-MEDIUM-STATE

4:25:14 loke It mixes in a few other states, such as INK, CLIP and LINE-STYLE

4:25:16 beach Never heard of it.

4:25:42 loke at first I thought that transformation should belong there too, but no... that wouldn't work.

4:26:03 loke I can't think of any solution other than changing the interface of MEDIUM-DRAW-TEXT to accept a transformation

4:26:44 beach We should consider changes like that. It is not as though we have to support some legacy software that does arbitrary stuff.

4:27:12 beach I mean "consider", i.e. we should think about it and not be tied too much to the specification.

4:37:02 loke beach: Indeed.

4:37:45 loke Someone, at some point, implemented proper transformations for the postscript engine, but it was rolled back because the “spec”

4:38:15 loke It honestly makes no sense to me... I'll link you:

4:38:54 loke https://github.com/lokedhs/McCLIM/blob/freetype2/Backends/PostScript/graphics.lisp#L562

4:40:11 beach Looking at it now...

4:41:15 beach Yes, I see.

4:45:47 loke My opinion is that the spec is incomplete. Broken, if you want to be blunt, in this case.

4:46:21 loke ...which is why I intedn to extend the specification of MEDIUM-DRAW-TEXT to take a transformation into account.

4:46:31 beach Or, let's say it took into account the realities of the time.

4:46:40 loke I guess at the time when this code was written, only postscript had the ability to transform text.

4:46:44 loke Right.

4:46:56 loke See my previous message :-)

4:47:49 loke This was around y2k, and I think then, most font rendering was still using bitmaps.

4:48:02 loke This has compeltely changed some two decades later.

4:48:10 beach Yep.

4:49:18 loke beach: Is there a difference between (SHEET-NATIVE-TRANSFORMATION (MEDIUM-SHEET <medium>)) and (MEDIUM-TRANSFORMATION <medium>) ?

4:49:51 beach I think so, yes.

4:50:21 loke beach: What are those differences? So far, I only ever paid any attention to the latter.

4:50:37 beach Let's see if I remember.

4:51:22 beach The latter is used to transform from coordinates given to (draw-...) to sheet coordinates.

4:51:23 beach Right?

4:51:47 beach The former has to do with the transformation between sheets.

4:52:01 loke beach: Yes. That' smy understanind (and making that assumption makes my text drawing work :-) )

4:53:33 beach Let me dig up the definitions.

4:56:31 beach *sigh*

4:56:35 beach Give me a few minutes.

4:59:38 loke Thanks, and sorry :-)

5:00:28 beach No, it's fine. I need to brush up on this stuff.

5:01:03 beach The native transformation transforms between the sheet coordinate system and the coordinate system of the (not necessarily direct) mirror of the sheet.

5:01:13 beach You should not have to worry about that one.

5:23:48 loke beach: Good, thanks!

6:11:22 d4ryus1 ** NICK d4ryus

7:20:26 loke It was remarkably easy to make the change to MEDIUM-DRAW-TEXT, once I figured out what to do.

7:22:15 jackdaniel that usually is the case

7:22:35 jackdaniel good morning

7:24:21 loke We have the same issue with draw-ellipse by the way... I think.

7:25:10 beach Hello jackdaniel.

7:30:55 jackdaniel gadget writing screencast: attempt no 3 :-) wish me luck

7:31:40 jackdaniel the first time I've got interrupted by a phone, the second time I've confused myself with some code passage and spent >15m on that

7:31:46 loke jackdaniel: Good luck!

7:35:47 loke Single take, no edits?

7:35:57 loke (that's what usually do, because I don't know how to edit video :-) )

7:36:41 jackdaniel maybe I should consider recording it in parts, but yes, these were single takes

7:37:16 jackdaniel obs is a good software for recording and it makes easy many things

8:06:50 hjek Hi, I still have the problem of McCLIM taking quite a while to load (on a reasonably new laptop). In this nice video tutorial, https://www.youtube.com/watch?v=kfBmRsPRdGg , the clim-listener loads in a matter of seconds, but when I load it, it takes exactly 5 minutes. I am getting McCLIM from Quicklisp, and I have no additional operating system repository common lisp packages installed that could possibly interfere.

8:07:42 hjek Are there any caching tricks or tweaks that are good for speeding up Quicklisp loading?

8:08:13 loke hjek: What is it doing during these 5 minutes?

8:08:45 hjek @loke: It displays some dots (.)

8:09:14 loke hjek: And it only happens on McCLIM?

8:09:20 jackdaniel hjek: first compilation takes some time

8:09:25 hjek @loke: when I was on Debian, previously, and loaded McCLIM via ASDF from the debian repos, then it would display compilation messages too.

8:09:26 jackdaniel but after that most things are cached

8:09:29 loke jackdaniel: True, but 5 minutes?

8:09:41 jackdaniel so if you restart image you may load it again and it should take seconds

8:09:44 hjek @jackdaniel: Yes, but the issue here is that things are *not* cached

8:10:05 jackdaniel btw, thank you for calling my video tutorial nice :-)

8:10:20 hjek @loke: only happens with McCLIM of what I've seen so far. But I've not used too many CL packages yet

8:10:35 hjek @jackdaniel: It looks really interesting. Particularly the "presenter" stuff.

8:10:40 jackdaniel loke: McCLIM is probably the biggest library in Quicklisp with a lot of dependencies

8:10:53 jackdaniel why fasls are not cached?

8:11:09 jackdaniel you may load mcclim and save the image (and load simply this image)

8:11:17 loke Stull, 5 minutes? It can't take more than 45 seconds for me to load all of mcclim on my old computer.

8:11:21 jackdaniel then it should take even less (if you can't afford having dev cache)

8:11:39 jackdaniel maybe you have faster disk? I don't know.

8:11:47 hjek What configuration do you use to cache things? Should it work automagically or do I need to set stuff?

8:11:47 jackdaniel for instance ironclad can load for minutes on my laptop

8:11:50 loke A standard SSD

8:12:02 jackdaniel asdf manages cache, nothing to be set

8:12:02 loke Ironclad is slow, takes perhaps 30 seconds to load fo rme.

8:12:08 jackdaniel it puts it in ~/.cache/common-lisp

8:12:20 hjek @jackdaniel: I'll check if they're there

8:13:52 hjek there are some mcclim cache files

8:14:07 hjek like, ./asdf-output-translations/home/pelle/quicklisp/dists/quicklisp/software/mccli -20180328-git/Core/clim-basic/encapsulate.o

8:14:26 jackdaniel ah, you use ECL?

8:14:32 hjek yup

8:14:37 jackdaniel then all is clear

8:14:42 jackdaniel I know about this issue

8:15:01 jackdaniel McCLIM is rebuild on each quickload, so there is ASDF cache problem

8:15:14 jackdaniel also ECL compiler is slow (because it goes through gcc)

8:15:21 jackdaniel these two things lead to long time of load

8:15:42 jackdaniel it will be either this slow, or you can use sbcl or ccl (they load McCLIM much faster)

8:15:45 hjek @jackdaniel: is there any way of explicitly telling QL to not do that?

8:15:52 hjek (not rebuild on every load)

8:16:29 jackdaniel hjek: it is a bug in ASDF, not in QL (or even a bug in our mcclim.asd system definition)

8:16:38 jackdaniel I'm aware of this problem, but didn't investigate that so far

8:18:38 hjek Ok, but ASDF is likely the issue? So it might make sense to look at ASDF cache options? (Right now using SBCL is not really an option for me, as the packages provided on sbcl.org don't work here, because of musl... etc etc)

8:20:18 jackdaniel this is not a configuration bug. It is either a problem in mcclim.asd definition (what causes recompilation), or ASDF internals which doesn't recognize, that system is already compiled and cached

8:20:53 jackdaniel if the problem is with mcclim.asd (or any other *.asd in mcclim) fix should be considerably easy

8:21:11 jackdaniel if that are asdf internals, then I wouldn't bether if I were you if you are not familiar with asdf internals

8:21:42 jackdaniel but when you load it all works acceptably smooth?

8:22:10 hjek Yes yes, once it's loaded, it works fine.

8:22:31 hjek Is it easy to save and load images in ECL, or does it not work that way?

8:22:43 hjek *system images

8:23:35 jackdaniel you may build image, it's not the same as saving image

8:23:55 jackdaniel it is easy, but it doesn't work with mcclim (another known issue which i didn't investigte yet0

8:24:11 jackdaniel so many issues, I need to clone myself