Perl 6 - the future is here, just unevenly distributed

IRC log for #darcs, 2017-02-26

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

All times shown according to UTC.

Time Nick Message
02:16 drostie joined #darcs
03:57 mizu_no_oto joined #darcs
06:03 drostie joined #darcs
06:40 drostie joined #darcs
07:20 drostie joined #darcs
14:54 bfrk joined #darcs
15:05 leg joined #darcs
15:08 bfrk I think what darcshub needs most is a more inviting layout. (I am not trolling, if i had any expertise in web-design i would have contributed something long ago but i don't.) Take the front page: the long user list is useless. What is most interesting about a projects hosting site? You guess it: the projects. So most prominent should be a way to browse them, search them, etc.
15:18 sm bfrk: I don't know if that what it needs *most*, but it's needed for sure
15:19 sm have you used it much ?
15:20 bfrk No. I have done some work on darcs and that was via the mailing lists and the roundup tracker, mostly.
15:27 bfrk I think the value of things like github is mostly in the way issues and merge proposals work. It is important to have an easy to use UI for writing and viewing comments.
15:28 bfrk For the actual version control stuff darcs on the command line is the perfect tool (bugs aside)
15:32 bfrk Recently looked at vervis, prompted by a message on haskell-cafe. Despite being (I think) much less mature it made me feel much more welcome to log in and try things.
15:36 bfrk BTW, I am sure I have made an account on darcshub at some time but I can't remember my username. It would help if I could search users, it probably had "bf" in it
15:40 bfrk i found my user name. Let's see if I manage to recover the password.
15:40 bfrk Yes!
15:43 bfrk My browser tells me that the settings page of in my personal account are not secured. i think darcshub should secure all pages by default.
15:46 bfrk BTW, vervis is self-hosted at https://dev.seek-together.space/, but the author looks for somebody else to continue the project.
15:50 bfrk sm: I guess you are simon michael, the author it darcsden. Please understand that I *want* hub.darcs.net to succeed. I just can't offer much to make it better other than share my thoughts/ideas
16:03 notdan joined #darcs
16:20 sm hello bfrk, Yes I'm Simon. I'm not the original author, though I did some work on it
16:22 sm darcsden/hub and darcs unfortunately are always short of person power
16:22 bfrk sm: the usual problem ;-)
16:24 sm an energetic organiser can stir up more contributors, but I'm not sure it will ever be easy
16:25 sm we all love darcs, but it doesn't have an obvious growth path for the future
16:29 sm bfrk: but when I get patches, they get reviewed and deployed pretty quickly, and several of us here will mentor you, so consider fixing a few small things
16:30 bfrk I agree. That doesn't mean it cannot grow again. The most important hurdle IMO is to fix the patch theory. Something like what they did for pijul, so we need no longer fear conflicts.
16:31 bfrk sm: I will
17:12 drostie joined #darcs
17:16 Heffalump bfrk: or maybe pijul will displace darcs
17:18 Heffalump oh, bfrk = Ben Franksen?
17:27 Heffalump FWIW the darcs approach to patches still has significant advantages over pijul in some ways, so it's not clear
17:52 bfrk heffalump: yes it's me
17:53 bfrk I am intrigued. What are those advantages?
17:57 Heffalump to make a pijul patch type, you have to actually define pushouts, and you typically end up embedding the patches into a significantly more complex domain (e.g. for their equivalent of hunk patches, you end up with a graph structure that could itself have cycles)
17:58 Heffalump so in the darcs theory it's much easier to define new patch types (but handling conflicts will suck)
17:58 bfrk @sm The main problem for me wrt contributing is that I am not a web developer nor ever was. i do control systems...
17:58 * lambdabot smashes a lamp on The main problem for me wrt contributing is that I am not a web developer nor ever was. i do control systems...'s head
17:58 Heffalump lol at lambdabot
17:59 bfrk how did i prompt that?
18:00 Heffalump I think "@sm" was treated as short for "@smash" or something
18:00 Heffalump if you actually want to leave a message with the bot you'd use "@tell sm" or something like that - though I imagine he'll read scrollback anyway
18:01 bfrk ...and i just wanted to address sm
18:01 Heffalump :-)
18:01 Heffalump "sm: ..." is the normal shorthand for that, but it doesn't really matter, I just thought the result was funny
18:02 bfrk i do chat so seldom always feel like newby
18:02 Heffalump bfrk: btw we were hoping to organise another sprint soonish, any chance of you being able to make it?
18:03 bfrk i would love to but it's difficult; but tell me when you have dates etc
18:04 bfrk re pijul: i am aware of this difference and agree with your assessment
18:05 bfrk But: i see a much bigger problem with pijul patches
18:06 bfrk And that is how to design a UI for resolving de-normalized (conflicting) graphs
18:07 bfrk gosh, i cant even spell newbie right
18:07 Heffalump true
18:08 Heffalump (sprint: current thoughts are Paris, but I'd also be quite happy to try to organise one in Berlin sometime..)
18:08 bfrk that would be extremely cool!
18:13 bfrk wrt chatting: i use pidgin now since xchat is no longer in my distro and xchat-gnome is broken; does anyone know how to explain pidgin that I am chatting in english and not in german?
18:16 bfrk wrt pijul: a while ago i have been working on a prototype for something inspired by the pushout paper (pijul is/was inscrutable to me) as well as the focal system (no longer online); how to linearize an arbitrary graph of changes was the main stumbling block
18:17 Heffalump I think if there's no unique linearization, that's a conflict (in pijul)
18:18 bfrk Of course. But you want to arrive at some linear thing eventually, right?
18:18 amgarchIn9 joined #darcs
18:18 bfrk i.e.  resolve the conflict
18:21 bfrk BTW, in my prototype there were no conflicts, not at the patch level. Just de-normalized (non-linear) graphs; that what i took from focal
18:23 bfrk the traditional systems (git etc) force you (more or less) to resolve conflicts as soon as they are encountered, that makes the task easier
18:25 bfrk have to leave for a few minutes
18:29 bfrk (back now) what i wanted was a simpler theory; i ended up with patches as a commutative group. IOW you can do anything you like, forward and backward
18:32 bfrk (i was severely burned by trying to prove things about darcs patch theory: inverse semigroups are just too difficult for my taste)
19:10 amgarchIn9 joined #darcs
19:15 Heffalump bfrk: right, I think pijul has patches to resolve a conflict by selecting an ordering
19:18 jeltsch joined #darcs
19:19 bfrk (With a group of patches, everything can occur not only zero or more times, it can also occur with a negative "count". You can delete a line (or chunk, or file) before it is created. Perhaps that makes the question of how to linearize the graph even more difficult than in pijul.)
19:22 bfrk Some problems with selecting an ordering: where do you start? what to do about disconnected subgraphs? how to handle circles?
19:24 Heffalump I think pijul has most of those problems. Circles are the biggest one.
19:24 Heffalump I think disconnected subgraphs are just gone from the live copy
19:27 pointfree sm: Sorry for the delayed response to your ping and code review, I was swamped with tasks at the time but that's cleared up now.
19:27 pointfree I agree with all of your code review points.
19:27 pointfree Hm should I rebase it or unrecord it?
19:29 Heffalump rebase/amend-record/unrecord+record amount to the same thing, replacing it with a new version
19:29 Heffalump rebase is just there for when you have a stack of patches you want to work with
19:29 bfrk Heffalump: Hmm. Perhaps pijul can get away with discarding disconnected subgraphs, because it limits the operations a user can perform on the graph.
19:30 pointfree Heffalump: thanks
19:36 bfrk Heffalump: The question is: (dis-)connected with what? Is there a special anchor for the first line/chunk? Or the last? Ah, i now remember having read on the pijul website that conflict resolutions are not invertible; perhaps this assures that you never loose the anchoring.
19:37 Heffalump in general pijul patches aren't invertible, that's right
19:40 bfrk Makes sense. Anyway disconnected subgraphs are that problematic anyway, even if you keep them. Just offer them for ordering (in any order), then the user can discard them if so desired.
19:41 bfrk The real problem remains how to represent non-linear subgraphs to the user in such a way that she can make a meaningful, informed decision about the desired order.
19:42 Heffalump agreed, though in the common case it's just like conflict resolution
19:42 Heffalump (normal two-way conflict resolution)
19:43 Heffalump I think pijul views one of the algorithmically hard problems as identifying the semantic intent behind a new linear file that's intended as a conflict resolution
19:43 Heffalump (I may be misrepresenting pijul here, it's how I understood something)
19:45 bfrk Imagine a pull with many deep conflicts, perhaps between different conflicting conflict resolutions. The things that make darcs runs amok and consume 100% cpu for years. A user certainly can do better but only if he can make sense of it all.
19:45 Heffalump I think the important thing is that conflict resolutions themselves merge icely
19:45 Heffalump *nicely
19:46 Heffalump so you go back to identify smaller sets of conflicts, resolve those, and work forward again
19:47 bfrk Yes, I guess that's how to play it: resolve bottom up. But sometimes you need the big picture to know what to do in the small.
19:48 pointfree pem__: ping
19:49 bfrk Perhaps this is all solvable. I hope so. One of the reasons i don't like the approach wrt development in pijul is that they optimize before we have a chance to get any experience with this sort of conflict resolution. We need to play with it.
19:55 bfrk One of the things I have spent some time with is finding better diff algorithms. Thta could perhaps help with the problem of finding the intent behind a resolution presented as a new version of a file.
19:56 bfrk You certainly cannot do this in O(n*log n)
19:57 drostie joined #darcs
20:05 bfrk In fact, I have always assumed that this is practically impossible. This is why I think an interactive conflict resolver is important. My hope is that this allows to reduce the problem to one of alternately choosing between chunks and editing them, making clear in process what the intention is.
20:12 bfrk In short: with darcs or pijul you must somehow deal with (perhaps deeply) *nested* conflicts, simply because the systems are designed to allow such a situation. Pijul solves the algorithmic problem of efficiently representing and accumulating such conflicts. How to properly present them to the user for resolution remains an open problem (IIUC).
20:49 bfrk joined #darcs
20:53 Heffalump bfrk: btw there's a #pijul you might want to join
20:56 jeltsch sm, pointfree: Hi! How did things go regarding the changes to Markdown on darcs hub?
20:57 bfrk Good idea. (BTW, I am currently trying to write a quickcheck test for Darcs.UI.Email.qpencode to find why I cannot reproduce Timo's problem)
21:01 pointfree jeltsch: I fixed it here: http://hub.darcs.net/simon/darcsden/compare/pointfree/darcsden-markdown-ext but I need to do some things sm pointed out in his code review, and then sm will need to deploy it.
21:20 jeltsch pointfree: Great to see that there is progress at this front. :-)
21:20 jeltsch By the way, I noted that there are other Markdown issues.
21:21 jeltsch 1. Block quotation are not rendered properly. Instead of generating a blockquote element, the “bird tracks” (>-signs at the beginning) are copied over to the HTML text.
21:22 jeltsch 2. Definition lists are not handled. They are not standard Markdown, but Pandoc supports them, and it would be good to have support for them in darcs hub.
21:22 jeltsch What do you think?
21:45 pointfree sm: You mentioned basing darcsdenMarkdownExtensions off of a set of extensions from a set of extensions in upstream pandoc. https://irclog.perlgeek.de/darcs/2017-02-24#i_14156971
21:46 pointfree sm, jeltsch: how do you feel about using multimarkdownExtensions https://github.com/jgm/pandoc/blob/ae8ac926a43ed48316081b7272701fba3884dbf5/src/Text/Pandoc/Extensions.hs#L234 as the base and then we can use Set union and Set difference to add and remove extensions as desired?
21:48 jeltsch I always use the Pandoc defaults, which apparently cover definition lists. Is Pandoc using MultiMarkdown by default?
21:51 pointfree jeltsch: pandoc uses this as its default set of extensions: https://github.com/jgm/pandoc/blob/ae8ac926a43ed48316081b7272701fba3884dbf5/src/Text/Pandoc/Extensions.hs#L137 and that includes Ext_smart (smart quotes)
21:51 jeltsch What are smart quotes?
21:55 pointfree jeltsch: https://irclog.perlgeek.de/darcs/2017-02-23#i_14151386   https://newrepublic.com/article/113101/smart-quotes-are-killing-apostrophe
22:01 dolio joined #darcs
22:30 bfrk left #darcs
23:11 sm pointfree: no problem. Re multimarkdown, let's not jump to another markup language; markdown is standard and what we had before, let's get it working again
23:12 sm I think all we need is pandoc's defaults plus Ext_autolink_bare_uris, no ?
23:13 sm minus Ext_smart
23:13 jeltsch sm, pointfree: Aren’t blockquotes via bird tracks standard Markdown? If yes, why are they currently ignored?
23:14 sm we are currently using github markdown, which I think wasn't really intended
23:18 sm until end of 2015 we used: readMarkdown def
23:19 sm and I think now we want: readMarkdown def { readerExtensions = insert Ext_autolink_bare_uris pandocExtensions }
23:25 sm night all
23:36 leg joined #darcs
23:37 pointfree sm: jeltsch's README.markdown is also having problems with angle brackets after bare URI's. I think the source of a lot of these problems is different markdown styles all being interpreted with the same set of extensions. Pandoc is written to interpret different markdown styles based on file extension...
23:37 pointfree Maybe that's what we should do?
23:48 pointfree That said, I don't think it would be a good idea to always go from: any --> html
23:49 pointfree e.g: PDFs' should not be converted to HTML
23:50 pointfree I think the best place to choose how something should be rendered is through the MIME map settings, somehow. http://hub.darcs.net/simon/darcsden/browse/src/DarcsDen/MimeMapSettings.hs

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