sjn tonyo: can the resolution speed somehow be made "ok" with a index format that is in a text (human-editable) format? Having an easy-to understand format can make a _huge_ difference when it comes to making it easy for new people to contribute and/or troubleshoot... 00:01
(this is especially true if there's an ambition to make the raku.land locally deployable (e.g. for managing business-internal packages)) 00:03
00:46 Nemokosch joined
Nemokosch Hi, something else: I think the version selection on the page of a dist is broken 00:47
raku.land/zef:lizmat/DirHandle take this dist for example 00:48
0.0.8 is preselected. Open it, select 0.0.7. It will appear as selected but 0.0.8 will still be disabled from selection and it will still have the "selected" attribute in the HTML...
and you will never be able to make it show
in fact, you don't even have to select the previous version, it's enough to open the drop-down and hover the previous one for a brief moment 00:50
00:53 Nemokosch left 00:54 Nemokosch joined
Nemokosch uh, wait... is this the actual UX? :D 00:54
if one clicks "Go", the selection changes... can't be coincidental, right? 😅 00:55
Then what can I say, I find it unintuitive. Why would you require an extra button only the update the selection of a dropdown. 00:56
I thought it would trigger the download. But there is a separate link for the download, and that link doesn't even change for some modules, regardless the version (maybe CPAN being the common theme) 00:58
00:58 Nemokosch left
jjatria I seem to remember there was a reason we had to go with the compromise of requiring an additional "Go" click, but we'd welcome anyone coming up with a better solution. I cannot find that reason if it ever existed, though... 11:41
What modules are you thinking of? I can see the download link update appropriately in the CPAN version of HTTP::Tiny, for example: raku.land/cpan:JJATRIA/HTTP::Tiny 11:42
JRaspass The alternative would be to require javascript, atm it's just a simple form, nothing complicated 13:58
tonyo sjn: it can be converted to human readable format using the module 17:19
zef eco is also capable of being a private repository for organizations 17:20
and then be indexed by private instances of RL without any modification or special handling in RL aside from an API key being provided to the endpoint
if you know of an org that would like a private auth please let me know and i will work with them on configuring it/setting it up 17:22
sjn is a little frustrated about the increased use of opaque file formats for keeping track of information that doesn't _need_ to be it... :-| 17:39
lizmat sjn zef performance is to a great extent affected by the JSON file format used currently, so there is a need 17:46
sjn what bottlenecks has profiling revealed? 17:54
tonyo the time to search and resolve is directly tied to the size of the JSON blob zef receives 17:55
sjn sure, as expected. though that's not what I asked about :)
tonyo so it's slow right now, 3500 modules, every time that number doubles so does the parsing. the bottleneck is json parsing 17:56
lizmat tonyo: you're aware of raku.land/zef:lizmat/JSON::Fast::Hyper ?
tonyo i am but it effectively needs to use whatever is in core to avoid problems with what zef thinks vs what rakudo internals 17:57
sjn to me "it's slow" is still a good enough reason to take another hard look at what's taking time in the language to see if it can be sped up somehow :) 17:58
tonyo so if we can put that in core then it may bide us some time until the format is needed
sjn: there's been a lot of eyes on the json parser, it's currently written in nqp
sjn: converting the index to a human readable format is trivial in this case, i'm not sure what the resistance is about but am open to more discussion of it 18:25
lizmat tonyo: fwiw, there's no resistance from me as long as there's an API for it 18:29
tonyo thank you lizmat 18:30
JRaspass likewise 21:44
21:47 Nemokosch joined
Nemokosch I also think the demand to move away from the "current API" (ie. a huge monolithic JSON file exposed) is very much justified 21:48
I suppose it's a main reason why a `zef search` or any sort of module resolution is so slow 21:50
and whenever I want to quickly check if a dist got uploaded indeed, it's not the most convenient API to work with anyway 21:51
for the wrong archives - honestly, when I saw it, I didn't assume it was a very specific thing. Now, I don't see anything similar. I guess it cannot be such a big thing, then. 21:56
(It was already after the realization that one has to press the "Go" button, as much as I can recall) 21:57
21:57 Nemokosch left

Powered by Cro and the Raku® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!