Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2015-07-12

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:18 mizu_no_oto joined #darcs
01:47 ilbot3 joined #darcs
01:47 Topic for #darcs is now http://darcs.net/ | logs: http://irclog.perlgeek.de/darcs/ | darcs 2.10.0 is out http://darcs.net/Releases/2.10
02:49 byorgey joined #darcs
02:49 byorgey joined #darcs
03:18 byorgey joined #darcs
03:18 byorgey joined #darcs
03:23 byorgey joined #darcs
04:49 Pastaf joined #darcs
08:31 alexei_ joined #darcs
09:42 alexei___ joined #darcs
12:23 maerwald uh, rootDir is not used in darcsden anymore, but how what variable does it use now for the serving location?
12:23 maerwald I feel like this could be a bug
12:25 maerwald yep, it is a bug
12:25 maerwald at least a documentation bug
13:03 maerwald sm: what the hell is "import Paths_darcsden"
13:03 maerwald is that a generated thing?
13:08 maerwald rage-clean :P
13:17 biglama joined #darcs
13:18 biglama hi guys
13:18 biglama i've just discovered hub.darcs.net and was wondering if there was a quota of private repo per user
13:20 maerwald sm: http://hub.darcs.net/maerwald/darcsden/patch/20150712131455-0bd70 and http://hub.darcs.net/maerwald/darcsden/patch/20150712131704-0bd70
13:21 maerwald publicDir handling was sorta weird, now it is consistent with the rest of the configuration format
13:28 maerwald ganesh should have a look too since I don't understand why he added that additional getDataFileName thing
13:33 maerwald anyway, that Maybe handling wasn't really the way to go, since we have the default configuration for that and the default configuration is merged with the user configuration. So publicDir cannot really be Nothing if it is part of the default configuration
13:56 mizu_no_oto joined #darcs
14:43 mizu_no_oto joined #darcs
15:10 mizu_no_oto joined #darcs
16:23 notdan maerwald: rootDir was not actually used anywhere IIRC; now it's just homeDir
16:23 notdan it is most likely a documentation bug tho!
16:24 notdan maerwald: Paths_darcsden is generated by cabal. the whole publicDir business is due to packaging various files (css, js) into the cabal package as data files
16:24 notdan maerwald: so if publicDir is nothing it means that darcsden should look into the location provided by cabal-install
16:24 notdan But yeah,maybe it's better to get rid of Maybe
16:25 maerwald notdan: yes and publicDir cannot be empty now because it is in the default (hardcoded) configuration
16:26 maerwald that was the whole idea of the config concept
16:26 maerwald except the user writes funny stuff in his config file^^
16:26 notdan Well it is not in the default configuration right now
16:26 maerwald it is with my patch
16:27 maerwald http://hub.darcs.net/maerwald/darcsden/patch/20150712131455-0bd70
16:27 notdan Yeah
16:27 maerwald src/DarcsDen/Settings/Production.hs :: line 53
16:27 notdan but you basically get rid cabal's data-files thing
16:27 notdan which was the point of Ganesh's patches
16:27 notdan that you just do `cabal install darcsden` and don't care about those css/js files
16:28 notdan and with your patch you would still have to keep those files in the current directory (by default; or somewhere else if you specify it in the configuration)
16:29 maerwald I don't think you can get away without any configuration anyway if you don't run it from the current dir
16:29 maerwald you would need to extend that logic for other variables as well
16:31 maerwald at minimum homeDir as well, otherwise ssh key stuff will not work
16:31 maerwald and then I am not sure what we gain by that additional logic
16:31 maerwald when everything is already configurable
16:31 notdan What do you mean? You don't have to extend that logic.
16:32 notdan The idea is that when you do `cabal install darcsden`, Cabal already downloads the public/ files for you and puts them in some directory with other packages
16:32 notdan so you already have them somewhere on your system
16:32 notdan and you can get the name of that directory using Paths_darcsden
16:33 notdan for example, on my system it is /Users/dan/projects/darcs/darcsden/.cabal-sandbox/share/x86_64-osx-ghc-7.8.2/darcsden-1.1.99/public
16:35 maerwald ok, so we don't want it in the default config then
16:36 maerwald instead I read the default thing from cabal
16:36 maerwald then we can still get rid of Maybe
16:37 Heffalump maerwald: that Maybe type was a mistake, I realised afterwards but never got round to cleaning it up
16:37 Heffalump but the cabal data-dir stuff is important
16:37 maerwald ok, I will clean up the Maybe then
16:37 Heffalump I need it to deploy darcsden with nixos for example
16:37 maerwald ah
16:37 maerwald those FHS breakers
16:37 maerwald :)
16:37 Heffalump and it fits in with the cabal "standard"
16:37 notdan thanks maerwald!
16:38 notdan Heffalump: do you have the .nix file for darcsden btw?
16:38 Heffalump I have one, but it's not necessarily perfect
16:39 * maerwald tries to not rant about nixpkgs
16:39 notdan Yeah, my nixos/nixpkgs experiment was a disaster :S
16:40 maerwald nix is a cool idea (the PM), but NixOS is utter chaos
16:40 notdan Waste an unbelievable amount of time on various things and at the end it didn't work
16:40 notdan ended up running nixos in virtualbox
16:40 maerwald you cannot run a distro project without policies, QA or some sort of consistency concept
16:40 maerwald they have neither of those
16:41 maerwald that means it's getting worse every year
16:41 maerwald ok, I said I will not rant :P
16:41 notdan oh I g2g, but I would like to hear more of your ranting later maerwald :D
17:03 maerwald Heffalump: there are two ways: taint defSettings with the IO type or append the cabal public dir to the defSettings inside getSettings. I am currently doing the latter
17:04 maerwald http://lpaste.net/136406
17:04 maerwald I find it weird to make defSettings have the type "IO String"
17:05 maerwald but maybe it confuses people
17:14 maerwald http://hub.darcs.net/maerwald/darcsden/patch/20150712171047-0bd70
17:15 maerwald and for the readme http://hub.darcs.net/maerwald/darcsden/patch/20150712170948-0bd70
17:52 Heffalump I'd be inclined to make defSettings be IO String
17:52 Heffalump it reflects reality
18:34 stulli joined #darcs
18:55 * Heffalump reappears
18:55 Heffalump hi stulli
18:58 Pastaf joined #darcs
19:05 notdan Heffalump: hi; sorry can you give me 10 more minutes before we start?
19:06 stulli hi Heffalump
19:07 Heffalump sure
19:14 notdan Alright, I am back.
19:14 Heffalump 'k
19:15 notdan So, I saw some semi-related shortcomings this week.
19:15 notdan Tried to finish editing a blog post and remembered that I suck at writing
19:15 notdan but got one published nevertheless
19:15 notdan and I have an idea about a second one
19:16 notdan also tried to make a nix package for darcsden -- failed miserably there
19:16 Heffalump great - I skimmed it and it looked good (the blogpost)
19:16 notdan nix doesn't work that well on OSX, and I didn't managed to get nixos to work properly
19:17 notdan wireless was dropping every 15 minutes of inactivity
19:17 maerwald Heffalump: I'll refactor it in a bit, I am currently ingame
19:17 notdan Regarding the code, some progress was made
19:18 notdan following the release of darcs-2.10.1 (thanks gh!) I refactor out all of my silly changes to darcs so darcsden-local compiles with darcs-2.10 branch fine now
19:18 notdan also uploaded new darcs binaries for osx online
19:18 notdan Oh, btw, I forgot to mention: I set up an instance of darcsden-local online - http://den.updog.xyz/local
19:19 notdan and as usual if you want to test out new changes you can grab them from hub.darcs.net/co-dan/darcsden-local
19:19 Heffalump great - probably won't this week, as my internet connection is a bit ropey
19:20 notdan So I factored out login/password -- it is no longer hardcoded and is read from the configuration file
19:20 Heffalump re darcsden with nixos, I just used the standard cabal2nix approach, but if nixos isn't working for you anyway I wouldn't waste more time on it
19:20 notdan Then there was this whole thing about DenInstance typeclass
19:21 notdan I don't think I have further comments on that
19:21 notdan Well actually I had some questions about that for you
19:22 notdan What are the better names for the DenInstance typeclass and for the EInstance module/datatype
19:22 notdan and what else should we add there?
19:23 notdan I wanted to  include a hook that would register additional handlers -- this should be useful for another change that I made
19:23 Heffalump DenInstance seems ok to me. EInstance, my best previous suggestion was DenConfig
19:24 notdan http://den.updog.xyz/local/nope/darcsden-local/browse/src/DarcsDen/Instance/Class.hs#18 I've included a procedure that should be run before the web server is started
19:24 notdan it is used in the local backend to pre-compute the list of repositories and cache it in a file
19:24 notdan it actually works surprisingly well
19:25 notdan you have to wait a lot during the start up, but after that it is pretty fast
19:25 Heffalump so is the idea that it's a cache, or that the user explicitly runs it when they want an update?
19:26 Heffalump and what if the user adds a new repo while the server is running? (in the filesystem)
19:26 notdan Well, that's a very good question. I thought that a user can manually re-compute the list of repositories by clicking a button or whatnot
19:27 Heffalump how long does that take?
19:27 notdan and 'cache' is probably the wrong word to use here -- sorry, I just realize that it doesn't make much sense in this context
19:28 alexei___ joined #darcs
19:28 notdan Heffalump: well, the same amount of time. it reindexes everything.
19:28 Heffalump was that a couple of minutes?
19:28 notdan but if user just adds a repository, then it shouldn't be so slow - we can just append a repository to the list
19:29 notdan Heffalump: no, something like 30 seconds
19:29 notdan on a directory with a bunch of darcs repositories and cabal sandboxes
19:30 notdan hm, but I bet if I try to do this on my home dir it will take ages
19:30 notdan let me test it out
19:31 notdan But I think there is nothing much we can do about initial indexing (just using your OS "search" feature on your home directory is also not going to give you a lot of results very fast)
19:32 notdan We can however use different datastructures and algorithms for speeding some operations on the index
19:33 Heffalump so would we reuse the index the second time darcsden-local starts?
19:35 notdan Well, I don't think we can. The old index is only useful for checking for things that are still there -- but we want to gather all of the repositories, including the new ones
19:35 notdan after all, we are not doing a search function
19:36 Heffalump if it's going to take a while to start the server, that implies having a persistent server, rather than launching it when the user wants it
19:37 notdan yeah, you are right :(
19:38 * sm waves.. isn't it possible to show something useful without such costly up-front indexing ?
19:39 sm the gitk/git-gui tools feel really light weight
19:39 Heffalump they don't know about other repos
19:39 Heffalump (AFAIK)
19:40 Heffalump but yes, we need to figure out something
19:40 byorgey joined #darcs
19:40 sm I see, the indexing is required to compare repos (the "branches" and "compare" pages ?)
19:41 Heffalump and to display the front page
19:41 Heffalump which actually probably isn't needed if the user launches darcsden from a specific repo
19:41 Heffalump they should see a page that's for that repo
19:41 sm just to list the repos on the front page, needs the big indexing operation ?
19:42 Heffalump the problem is finding all the repos in the user's home dir
19:42 Heffalump the alternative we've considered is making them manage it manually somehow
19:42 sm yeah that should should definitely be simplified so it's fast or made the non-default view
19:43 Heffalump perhaps the user should just configure an explicit list of search directories
19:43 sm DARCS_REPO_PATH, defaulting to ~:~/src ?
19:44 Heffalump would it be recursive within each entry?
19:44 sm that would find repos like ~/* and ~/src/* only
19:44 sm (not recursive)
19:44 Heffalump would be useful to have a syntax for a recursive search too, I guess
19:44 notdan Not having an index also slows down just the repository page -- apparently it tries to load the branches at that stage as well
19:44 Heffalump notdan: that can probably be fixed though
19:45 notdan yeah.. I don't know why I haven't done that yet. My bad
19:46 Heffalump not sure if it's worth the hassle, but it's also possible to write a test for that kind of thing, by having a dummy backend that errors if you want the list of repos/branches
19:47 notdan it's the patches button that's the culprit http://den.updog.xyz/local/nope/darcsden-local/browse/src/DarcsDen/Pages/PageUtils.hs#173
19:48 notdan Heffalump: it is also possible that if we open darcsden from the command line it sets up the DARCS_REPO_PATH to the current directory (or whatever the directory we are opening)
19:49 Heffalump notdan: what happens if the user runs the cli command from two repos in parallel - will they go to a single darcsden instance or multiple ones?
19:49 Heffalump if a single instance, then DARCS_REPO_PATH won't work
19:50 sm as a user I'd be very happy to just browse my current repo by default, with inter-repo browsing as a more advanced feature
19:50 Heffalump for me inter-repo browsing is actually quite important, but baby steps make sense anyway
19:50 sm ("click up arrow to see your other repos")
19:50 notdan Heffalump: good point :[
19:52 sm maybe each repo can define it's related repos, so darcsden can just use those ?
19:53 notdan but then that information should be update on every `darcs get`
19:54 notdan and what happens when we delete one of the children repositories?
19:54 sm (if I have it right: searching the filesystem to find repos is not too slow, indexing a darcs repo's patches for comparison is a bit slow, and indexing all possible darcs repos for comparison is very slow)
19:54 notdan well I guess in that case we can just remove the entry from the list when we fail to access the children
19:54 notdan oh, actually searching the filesystem to find repos is very slow
19:55 sm really ? ok
19:55 notdan a big bottleneck :(
19:55 sm wouldn't bother with it then
19:55 biglama i've found a wronk link in the FAQ of darcs
19:55 biglama where can I report it ? or better, how can I correct it ?
19:55 notdan but once you have a list of repositories, finding all the children for a specific repository is just linear
19:56 notdan biglama: if it is the FAQ at http://darcs.net/FAQ you can just register an account and edit the page yourself
19:56 notdan or you can send an email about it to darcs-devel@darcs.net
19:59 biglama notdan: done. Thanks !
19:59 sm darcs maintains a list of upstream repos (remembered from past get/pull commands ?), and darcsden maintains a list of child repos (remembered from darcsden fork operations). It would be nice to integrate and improve these mechanisms eh
20:00 Heffalump and if we could just index all repos quickly, the rest would be easy
20:01 Heffalump I'm quite liking the "specify a search path explicitly" route
20:01 sm biglama++
20:01 Heffalump it would mean an annoying setup step, but apart from that cuts out a lot of the other problems
20:01 sm biglama: PS you asked about quotas for private hub repos, there aren't any but should be
20:02 Heffalump and if it would support foo/* and foo/** syntax or similar, the user should get a fair amount of control
20:02 biglama sm: yeah. I guess there will be quotas when there will be too many private repositories
20:03 notdan hm, that sounds good
20:04 Heffalump (check what syntax _darcs/boring uses)
20:04 sm if you really want it to just work with no config, you could offer to scan new directories in the UI..
20:05 Heffalump perhaps always add the current directory to the config, if it's not already covered by it
20:05 sm certainly
20:07 sm "these repos were found in, adjacent to or immediately below the current directory. You can configure DARCS_REPO_PATH to search additional directories for repos"
20:08 Heffalump I'd put that in the darcsden config rather than an environment variable, btw
20:08 sm but really, I don't see discovery of all your repos as being the most important thing darcsden can be doing for you
20:09 notdan hm, I can't find much about the boringfile syntax, but I think it's just regular expressions
20:09 Heffalump agreed, but knowledge of all your repos is very useful for some of the other things it can do for you
20:10 sm alright.. it's sounding pretty slick
20:11 sm notdan: I wouldn't bet on that
20:11 notdan Well, I wanted to pinch in an old idea that I had in mind -- just having a separate directory dedicated to darcs/darcsden repositories; if you want to add a repository just symlink it there
20:11 maerwald Heffalump: http://hub.darcs.net/maerwald/darcsden/patch/20150712200953-0bd70
20:12 sm notdan: I thought the same
20:12 Heffalump maerwald: fine with me
20:12 notdan we can utilize the exising interface ("new repo" button) for creating symlinks
20:12 Heffalump sm/notdan: that's kind of like the explicit index
20:12 Heffalump i.e. an explicit list the user manages
20:12 notdan yeah
20:13 sm yeah.. or many of them (all repos found in the current directory are the related set)
20:13 Heffalump the UI needs to distinguish somehow between "track this existing repo" and "add a new repo"
20:13 notdan good point
20:14 sm a while back I was thinking darcs hub should consider repos related based simply on a naming convention
20:14 notdan maybe the 'track existing repo(s)' functionality can be implemented in the command-line tool?
20:14 sm just for simplicity
20:15 Heffalump sm: I bet said naming convention would conflict with mine :-)
20:15 Heffalump notdan: yeah, probably
20:16 Heffalump notdan: if we're going with the explicitly user-maintained list, I'd have it in a single file rather than a symlink farm - it's more economical
20:16 maerwald so, haskellers boast about controlling complexity... so what is this then: foo :: XMLGenT (HSPT XML IO) (XMLType (HSPT XML IO))
20:17 maerwald it's terrible monad burrito no one can follow
20:17 maerwald hsx is terrible
20:17 maerwald I can't figure out how to combine inline xml with xml from e.g. a file
20:17 sm maerwald: agreed! upgrading it was a nightmare. It works rather well though. Also thanks for the patches.
20:17 maerwald you can't just do "return", it will do random things
20:18 maerwald sm: and it breaks ghc_mod, error messages and whatnot
20:18 maerwald line numbers don't match anymore
20:18 maerwald etc
20:19 notdan it slows down my emacs for some reason
20:19 notdan it's a blast from the old times when we had PHP
20:20 sm I don't want to derail Heffalump/notdan's meeting, but interested in alternatives to HSP. The more I looked at it though, the more I felt it didn't make sense to switch
20:20 sm as long as you don't have to think about those types
20:20 notdan Heffalump: so, what is your final take on the indexing problem?
20:22 Heffalump I think if I was implementing it, I'd have an explicit text file either listing repositories or patterns that specific where to look for repositories. Add to it whenever darcsden discovers a repo that isn't already covered.
20:22 Heffalump s/patterns that specific/patterns that specify/
20:23 Heffalump probably just a list of specific repos is good enough for now
20:24 sm can that file be _darcs/prefs/repos (or sources, I'm not sure of the difference)
20:24 Heffalump it'd be a single global list, so somewhere in ~/.darcs I guess
20:24 Heffalump though if it's just for darcsden then ~/.darcsden might be better
20:24 Heffalump not sure
20:25 notdan OK, got it
20:25 sm is there also a smaller list of "repos related to this one" ?
20:26 notdan sm: hm, no. currently we just go over all of the repos and filter out those that are related
20:26 notdan it sounds slow, but actually unless we have thousands of repositories it shouldn't be too much of a problem
20:26 sm interesting, how do you figure out relatedness ?
20:27 notdan it's just the parent/child relation
20:27 Heffalump _darcs/prefs/defaultrepo defines the parent of a repo
20:28 sm I see, and maybe prefs/repos defines other parents/siblings/branches
20:28 * sm looks forward to trying it out
20:30 maerwald sm: the main point with hsp/hsx is... I cannot figure out how to make the front page configurable. You cannot just inject a string or an XML file into our DDXML type. You cannot do it with "return", there is no Read instance and no helper functions. It looks like you'd have to implement that totally from scratch which would involve understanding that terrible monad burrito
20:30 notdan wait, does darcsden currently look for siblings?
20:31 Heffalump I don't think so..
20:31 sm I think just parents and children
20:31 sm but a parent-child is really a sibling-sibling if you squint
20:32 notdan hah
20:32 sm just depends how you're using them
20:32 maerwald the funny part is... you can use string literals where DDXML is expected, but you can not use a variable of type String
20:33 sm we usually think of fork as creating a child, but it's really more general
20:33 notdan maerwald: you can do something like that http://hub.darcs.net/simon/darcsden/browse/src/DarcsDen/Pages/HubPages.hs#74
20:34 sm maerwald: yeah, what about that "cdata" helper
20:35 notdan Heffalump: ok, I have one more entry in my .org file that just says "daemonization", but I forgot what exactly I wanted to ask you
20:35 Heffalump authentication/privileges?
20:35 alexei___ joined #darcs
20:37 Heffalump to be honest I'd stear clear of that for now given the difficulties of making that work well. Have each user run their own copy however - I'd be inclined to say on-demand, i.e. when the cli tool is run, but still need to think about when it shoul die, if ever.
20:37 notdan i don't think that should be an issue, the darcsden-local process should always be run as current user, shouldn't it?
20:37 notdan ok, makes sense
20:38 notdan So the agenda for the nearest time is: 1. Rename `EInstance` to `DenConfig`
20:38 notdan 2. Write a blog post about using existential types for picking/packaging up backends
20:38 notdan 3. Merge with upstream once again; cherry-pick patches that can go into upstream right now.
20:38 Heffalump notdan: yes, if there's one per user, rather than one machine-wide
20:38 notdan 4. Make sure that the patches button doesn't cause the darcsden to list all the repositories/repository forks.
20:38 notdan 5. Redo the indexing of repositories.
20:39 notdan How does that sound? Is there anything else I am missing?
20:39 Heffalump sounds fine. How close are you to having the bulk of the work be mergeable? How many weeks are left of GSoC?
20:43 notdan 6 weeks left; I am not sure about the mergability
20:44 notdan it compiles with the released version of darcs, so it should be fully mergable i guess?
20:44 notdan btw, is it possible to have a mini-meeting during the week?
20:44 Heffalump I meant ready to merge in the sense of you think it's ready and it being acceptable to sm
20:45 Heffalump should be, only problem is I can't be 100% certain of being available on a specific day this particular week - most days I should be around at 8, but it's possible that I won't be on one or two.
20:46 Heffalump can you just try to grab me when you're ready?
20:46 notdan Heffalump: certainly!
20:46 notdan as for the mergability, I would say a week or two
20:46 notdan What I really want to do is to have more or less working thing by the end of the next week.
20:47 Heffalump yes, that'd be great
20:47 notdan Most likely without some useful features I'd love it to have, but working
20:47 notdan working in the sense that I can do `cabal install` and use it
20:47 Heffalump did you merge the cli tool into the main repo btw?
20:48 notdan that's why I wanted to figure out critical parts and implement them in the simplest way possible, and that's why I wanted to have an additional meeting
20:48 notdan Heffalump: not yet, but that shouldn't be a problem
20:49 Heffalump notdan: makes sense re meeting
20:50 notdan yeah we don't have to set it up - this week we had some good chat about runtime flags for example
20:51 sm notdan: don't forget review/merge time for you and me - though everything you showed me so far merged easily
20:53 notdan sm: yeah most definitely!
20:53 notdan i don't know exactly when i will be ready tho :(
20:53 Heffalump just focus on the things you need to be ready, and don't worry about precise timescales
20:54 Heffalump they'll be wrong anyway
20:54 notdan hah true :)
20:55 sm :)
20:55 mizu_no_oto joined #darcs
20:56 sm I will try to keep merging eagerly from your patch queue - push there whatever is ready
20:56 sm last minute merging usually derails things
20:58 sm I think I asked if I could merge what's there now, and I might have forgotten your answer - is that stuff good to go ?
21:00 sprang joined #darcs
21:03 alexei___ joined #darcs
21:06 notdan yeah, i think they are good to go
21:06 sm ok thanks
21:06 notdan the instance based on  http://hub.darcs.net/co-dan/darcsden is running at http://den.updog.xyz/den/
21:07 sm notdan: and these are the darcs 2.10.1 compatible ones right
21:08 notdan yes
21:08 sm ok merged, thank you
21:11 sm maerwald: merged yours too, thanks, and especially for the doc fix
21:11 maerwald sm: I think it works with cdata, hacking something up right now
21:11 maerwald sm: my idea is this: if there is a file "frontPage.xml" in the public/ dir, then it will use that instead of the inline xml
21:12 maerwald configuring that in the .conf file doesn't seem like a good idea
21:13 notdan ok, I wrote up a summary of the meeting on the wiki http://darcs.net/GSoC/2015-Darcsden
21:14 notdan maerwald: do you want to read frontPageContent from the file or the whole front page?
21:21 maerwald notdan: e.g.: frontPageContent = <p class="blurb"><% cdata $ pack frontPageXML %></p>?
21:22 notdan sorry, I meant to ask whether you wanted to customzie just the 'frontPageContent' part or the whole 'frontpageP'?
21:22 maerwald I'll do it for siteLink, siteFooter, frontPageContent all in separate files
21:23 notdan oh, okay
21:23 maerwald because those are the only things that are currently part of the Settings
21:23 sm great. I love runtime templates. It will just touch disk more so serve requests a bit slower, right
21:23 notdan I was just worried you wanted to modify frontpageP, because I did that in my branch and it would be painful to merge
21:23 sm first merged to trunk wins! :)
21:24 notdan would be nice to have pre-compiled templates like in snap
21:24 notdan would actually make a lot of sense since we are already using snap
21:24 sm isn't that what we have ?
21:24 maerwald wait
21:24 maerwald what
21:25 notdan sm: I meant Heist templates, like in snap-framework
21:26 sm heist templates are dynamic, right ? I wanted to try them but couldn't figure them out
21:27 notdan You can compile them at "load time" IIRC
21:27 notdan http://snapframework.com/docs/tutorials/compiled-splices
21:27 sm they have a several variants, I know
21:27 sm right: The main idea is to do most of your splice processing up front at load time. There is still a mechanism for rendering dynamic information at runtime, but it is faster than the fully interpreted approach
21:29 sm it's dynamic templates, optimised so only the dynamic bits are regenerated each time
21:33 notdan alright, the new meds are making me wonky so i'll go hit the sack
21:33 notdan see you tomorrow, and good chat today :)
21:33 sm night notdan !
21:33 sm good work
21:39 maerwald notdan: why does "frontPageContent = <% whenXML frontPageXMLExists $ cdata $ pack frontPageXML %>" not work? It works if there is <p class="blurb"></p> around the <% ... %> thing
21:41 maerwald just use the div tag?
21:42 sm yeah.. <% %> doesn't create an element, it's just for interpolation
21:43 sm there might be a workaround so you don't have to add extra tags, but I forget it
21:44 byorgey joined #darcs
22:05 byorgey joined #darcs
22:30 mizu_no_oto joined #darcs
22:45 byorgey joined #darcs
22:54 maerwald sm: I think I've managed it
22:55 maerwald it's not particularly beautiful, but inline xml is never
22:56 byorgey joined #darcs
22:57 maerwald http://hub.darcs.net/maerwald/darcsden/patch/20150712225558-0bd70
23:00 maerwald the type signature of maybeXML is horrible, but...
23:01 maerwald oh, I think I can even shorten it
23:04 maerwald http://hub.darcs.net/maerwald/darcsden/patch/20150712230230-0bd70
23:07 maerwald I hope the divs don't break anything, like style sheet etc
23:07 maerwald someone should double check
23:33 dixie joined #darcs
23:36 Igloo joined #darcs

| Channels | #darcs index | Today | | Search | Google Search | Plain-Text | summary