Log file Viewer

[03:23:28] <Ivo> !logs

[03:23:28] <pmxbot> http://chat-logs.dcpython.org/channel/pypa-dev

[04:23:20] <pypabot> Deployed 07aa364 to Warehouse (Production)

[15:10:47] <pypabot> Deployed 0d6aa59 to Warehouse (Production)

[16:50:14] <nlh> ping dstufft

[16:51:20] <dstufft> nlh: pong

[16:52:08] <nlh> hello... actually, I think I might have solved my own problem....

[16:52:29] <dstufft> :D

[16:59:27] <nlh> dstufft: so... i have the live reloading of assets working... which is sweet. Thankyou

[17:00:00] <nlh> the only issue is that i keep getting a builtins.ValueError

[17:00:03] <nlh> ValueError: components/font-awesome/css/font-awesome.css is not able to be cache busted.

[17:00:13] <nlh> via this workflow;

[17:00:18] <nlh> change file... save file

[17:00:21] <nlh> refresh page

[17:00:23] <nlh> value error

[17:00:26] <nlh> refresh page again

[17:00:29] <nlh> see changes

[17:01:23] <dstufft> nlh: I think the problem is that compilation takes a minute or so and in the meantime you're in an inconsistent state

[17:01:29] <nlh> ah right, ok

[17:03:12] <dstufft> maybe the error should be relaxed so that it doesn't happen in development, or development should get an incremental css compilation or something

[17:04:00] <nlh> if it's not too much work....

[17:04:08] <nlh> it's annoying, but I can put up with it :P

[17:05:05] <nlh> playing around with it - it seems like it takes around 4-5 seconds on my machine

[17:05:22] <dstufft> relaxing the error isn't hard but it might give weird results, the error basically means "the manifest file that we tried to read didn't have this URL in it"

[17:05:34] <dstufft> so you're going to get an un-cachebusted url

[17:05:41] <dstufft> which may or may not be cached

[17:06:32] <dstufft> I'll poke at it and see if I Can make it so it doesn't do caching in dev and it'll just fall back

[17:06:40] <nlh> ok, thanks

[17:39:10] <dstufft> nlh: https://github.com/pypa/warehouse/pull/879 that should remove the ValueError

[17:39:26] <dstufft> and it disables the static file cache in development

[17:39:46] <dstufft> if you're fast enough, you might still have to double refresh to actually get the updated content though

[17:40:29] <dstufft> (it basically turns the ValueError into a sometimes it does the right thing, and sometimes you get a 404 of the static asset and sometimes you get the old version of the static asset)

[17:42:25] <nlh> ok, that's great. Thanks

[17:46:29] <dstufft> nlh: updated https://github.com/pypa/warehouse/pull/879 with a full description of the caveats, can you +1 (or merge it) or -1 (or close it)

[17:47:39] <nlh> thanks dstufft - i am happy with those trade offs

[17:47:48] <dstufft> awesome

[17:48:41] <dstufft> I gotta go and work on a kitchen table I'm building my wife. I'll deploy whatever is on warehouse-staging to prod before I go though.

[17:48:57] <nlh> ok, thanks dstufft ... happy table building :D

[17:49:07] <nlh> also - if you have a minute - can you take a read of the article I sent you?

[17:49:11] <nlh> No rush :P

[17:49:16] <dstufft> whicih article is that?

[17:49:20] <dstufft> don't think I got it

[17:50:20] <nlh> ok, I'll send it to you

[17:50:37] <dstufft> oh

[17:50:41] <pypabot> Deployed 6fded23 to Warehouse (Production)

[17:50:42] <dstufft> wait I remember now

[17:56:01] <dstufft> nlh: I read it before (and just re-read it now), it looks good to me! Though one possible niggle we'll need to work out (not specific to the article I don't think... though it pointed out to me a possible problem). I don't think l20n.js supports all the way back to IE8. I think they currently only support the "evergreen" browsers (e.g. the ones that auto update themselves). So that might mean either people using an older browser can't get

[17:56:02] <dstufft> translations (though that's kind of sad :( ) or that I'll need to see about either adding support for that to l20n.js OR porting l20n.js to Python to use it serverside instead of client side

[17:56:41] <nlh> hmmm... that is a real problem

[17:57:03] <nlh> especially given that many of our users on older browsers are not in english speaking countries

[17:57:06] <nlh> i.e china

[17:57:07] <dstufft> yea

[17:57:42] <dstufft> porting it to Python might also solve https://github.com/pypa/warehouse/issues/833

[17:58:44] <dstufft> I'll have to poke at that later

[17:58:56] <nlh> it sounds like the python port would be a lot of work, no?

[17:59:11] <nlh> are there any 'off the shelf

[17:59:17] <nlh> ' alternatives we can look into?

[17:59:37] <nlh> should I log an issue for this?

[17:59:47] <nlh> so there is an open discussion?

[17:59:56] <dstufft> Hmm, a Python port might not work out entirely either, because Python won't be able to do the whole responsive translation thing

[18:00:13] <dstufft> since Python won't know for example, what the view port size is

[18:00:19] <dstufft> yea, an issue is probably relevant

[18:01:16] <nlh> well, the responsive translation is *really* nice, but if given the choice I'd rather support older browsers and give this up, than the other way around

[18:01:49] <dstufft> I don't know JS very well to know how hard it would be to add support for older browsers to l02n.js

[18:01:59] <nlh> In our layout there are not that many points where I am expecting a fixed length string

[18:02:07] <nlh> probably just the top menu (off the top of my head)

[18:02:11] <dstufft> Okay

[18:02:15] <nlh> everything else is pretty flexible

[18:02:22] <dstufft> there's also the whole, going back to gettext thing we could do

[18:02:24] <nlh> maybe the sidebar menu on the packages page

[18:02:37] <nlh> *package page

[18:03:12] <nlh> but even then, the text just wraps, which is okay

[18:03:29] <dstufft> L20n vs gettext is like -> <div data-l20n-id="home">Home</div> vs <div>{{ _("Home") }}</div>

[18:04:45] <dstufft> though L20n has some other benefits over gettext in that it sort of natively understands HTML

[18:04:56] <dstufft> but gettext has the benefit that there are a _ton_ of tools for it

[18:07:00] <dstufft> the natively understanding HTML is nice for things like links, in gettext if you want an embedded link you have to pass it in as a variable and then the translator needs to do: Some translated <a href="%(url)s">link</a>, in L20n the translator only needs to do: Some translated <a>link</a> (on the flipside, currently l20n doesn't allow links to be re-ordered, but the gettext solution does

[18:07:35] <dstufft> I feel very unqualified to answer how important particular features are of l20n vs something else :/

[18:08:06] <nlh> I don't have a ton of experience in this field either. Is there anyone in particular we should talk to to help us weigh up the pros and cons?

[18:08:42] <nlh> re: gettext... I imagine that people who do translations are pretty used to '<a href="%(url)s">link</a>'

[18:08:58] <nlh> i.e. they would probably have dealt with it before

[18:08:59] <dstufft> Generally I ask jezdez :D He suggested l20n.js to me, but idk if he has time to do a big discussion about it or not

[18:09:01] <dstufft> yea

[18:09:21] <nlh> ok, we'll I'll open the issue and we can go from there

[18:09:26] <dstufft> Okay

[18:09:50] <dstufft> Going to go ahead and go work on this table now :D thanks again!

[18:10:16] <nlh> have fun :D