#microformats 2019-08-25

2019-08-25 UTC
#
@tw2113 New post set sail: Adding my own pronouns as part of potential microformats2 spec https://trxtp.com/62 (twitter.com/_/status/1165417479738249216)
KartikPrabhu, [tantek], mauz555, [KevinMarks], [Rose], [Michael_Beckwit, [bdesham], [asuh] and [snarfed] joined the channel
#
[tantek] [snarfed] just checked that jasonbriggs post and it has a u-url now to its permalink (as blog posts should). so the fix is to fix the h-entry in general. I think part of the answer here is to check any h-entry with the indiewebify h-entry validator and if it has errors, then blame/fix that first, before blaming anything else. Lots of things can fail when markup is bad
mauz555 joined the channel
#
[snarfed] sure. all fine debugging advice. the real interesting part here is the original (failed) authoring experience, not the debugging
#
[tantek] it's not a normal "authoring experience". it's a microblog post which should have been just fine, so somehow it was broken by alterations, that's what's so odd about it
#
[tantek] point being if someone is a developer skilled enough to mess with something that's already working, the right answer is to set them up with regression testing so they don't introduce a regression. not an "authoring experience"
#
[snarfed] wait, microblog meaning micro.blog? are you sure? if so than ok yes, i have no idea what happened there
#
[tantek] yes, check out the styling. it's a self-hosted microblog
#
[tantek] yes with the . 🙂
#
[snarfed] ok. random. backing away slowly
#
aaronpk 🤔 micro.blog can't be self-hosted
#
[tantek] well, curious. how did a microblog post get "broken"?
#
[tantek] I mean, it can be domain mapped
#
[tantek] or maybe in a directory
#
aaronpk it can be domain mapped, i've never seen it done in a subfolder, but in either case, the owner does not have control over the HTML still afaik
#
aaronpk ok that site is not hosted by micro.blog
#
[snarfed] just to double check, you're assuming it's altered m.b based on the url path and some similar styles?
#
aaronpk checked the IP address
#
aaronpk I think it's just styled to look like micro.blog. it's not even a 100% accurate reproduction
#
[snarfed] could be an html export. but yeah that seems more likely
#
KartikPrabhu its the second post here https://micro.blog/jasonrbriggs
#
KartikPrabhu not sure if that is a POSSE
#
aaronpk ok yeah the fact that on micro.blog they all have text like "Microblog posted at 12 Aug 2019 19:57:38" leads me to believe it's a homegrown site that micro.blog is importing from
#
[snarfed] m.b easily syndicates feeds. mine does the same: https://micro.blog/snarfed
#
[snarfed] yeah. so it may well have been an original failed mf2 authoring experience after all
#
[tantek] not sure about "easily syndicates feeds" — we never did figure out the photo syndicating problem(s)
#
[snarfed] it definitely easily syndicates feeds. that doesn't preclude occasional bugs 😎 happens to everything now and then
#
[snarfed] witness bridgy Twitter replies regressing for the last week and Instagram broken entirely, sigh
[Michael_Beckwit, [jeremycherfas], IWSlackGateway1, gRegorLove, [dshanske], [Rose], [jgmac1106], GWG, [tantek], [bdesham], KartikPrabhu, [grantcodes] and [vendan] joined the channel; SebDiscord[m] left the channel
#
[vendan] spruced up my microformats parser comparison tool some: https://mf2.andyleap.dev
#
[vendan] tell it to parse a chunk of html, and it'll render them in pretty printed json in an accordian
#
[vendan] php and python parsers for now, but should be relatively simple to add more 🙂
#
KartikPrabhu [vendan] if the goal is to compare the parsers then the accordian is a bit annoying since you can't have multiple ones open to compare
#
KartikPrabhu but otherwise looks neat
#
vendan yeah, still working on that some, but part of the goal is to make is so you don't need to eyeball both to compare them
#
KartikPrabhu got it
#
KartikPrabhu so something like you pick a reference parser output and the others show the difference?
#
vendan yeah, might do something like "calculate minimum set all the parsers agree on" and then diff against that
#
KartikPrabhu sounds good
#
vendan I'm also going to be pulling in the test suite, and having it's "expected output" be the comparison base for those
#
vendan initial goal for this stuff is to serve as a tool for evaluating/refining https://github.com/microformats/tests
#
KartikPrabhu :thumbsup:
#
[tantek] vendan++
#
Loqi vendan has 1 karma over the last year
#
[vendan] got a select list filled with all the tests now, and if you click a test, it'll parse that test. Still need to get it to render the test html to the textarea
ingoogni, gRegorLove, [KevinMarks] and [vendan] joined the channel
#
[vendan] and now it renders the test html to the text area, diffs the parser outputs against the test's expected output, and turns the parser's header red if there is a difference
#
[vendan] sadly, it's triggering a lot on the php parser cause I think PHP's json output is a little... wonky
#
[vendan] php serializes empty array as `[]`, even though it'd actually be used as an associative array, and thus `{....}` if there was anything in it
#
aaronpk where are you seeing that? I thouoght the php parser fixed that ages ago
#
[vendan] ehhh, might just be cause of something I'm doing?
#
[vendan] ```
#
[vendan] <?php
#
[vendan] use Mf2;
#
[vendan] namespace MF2TestBench;
#
[vendan] require 'vendor/autoload.php';
#
[vendan] $html = file_get_contents('php://input');
#
[vendan] echo json_encode($output);
#
[vendan] $output = Mf2\parse($html, 'http://example.com/');
#
[vendan] ugh, really? can't edit messages...
#
[vendan] ```
#
[vendan] namespace MF2TestBench;
#
[vendan] <?php
#
[vendan] require 'vendor/autoload.php';
#
[vendan] use Mf2;
#
[vendan] $html = file_get_contents('php://input');
#
[vendan] $output = Mf2\parse($html, 'http://example.com/');
#
[vendan] ```
#
[vendan] echo json_encode($output);
#
[vendan] using latest composer installed version as of 2 days ago or so
#
aaronpk can't edit messages because edit doesn't mean anything in irc ;-)
#
[vendan] true
#
aaronpk ah there's a flag to switch that https://github.com/microformats/php-mf2#generating-output-for-json-serialization-with-json-mode
#
Loqi [microformats] php-mf2: php-mf2 is a pure, generic microformats-2 parser for PHP. It makes HTML as easy to consume as JSON.
#
[vendan] looks like I need to turn jsonMode on
#
[vendan] yeah, that
#
[vendan] yeah, that cleans up the php parser for most of the tests
#
[vendan] so far looks like there's some issues with whitespace stuff
#
[vendan] based off of the parsing spec, the tests seem to be correct, and the parsers are being too generous about sanitizing whitespace
#
KartikPrabhu welcome to the deep end of whitespace parsing :P
#
KartikPrabhu here's your life-guard Zegnat
#
[vendan] ehh, fortunately, the specs are pretty clear that it shouldn't be collapsed
#
KartikPrabhu yes. and mostly it was found to not be very useful
#
KartikPrabhu long story
#
KartikPrabhu not sure where it is documented
#
[vendan] http://microformats.org/wiki/microformats2-parsing-issues#When_to_collapse_whitespace_in_properties
#
KartikPrabhu https://github.com/microformats/microformats2-parsing/issues/15
#
[vendan] (which also means it's in a bunch of the parsing rules, as `textContent`
#
KartikPrabhu yes. which is also ambiguous enough
#
[vendan] textContent isn't really that ambiguous, it's pretty strongly defined, it's just going to have a bunch of newlines and such in there
#
[vendan] and also be missing a bunch of newlines
#
[vendan] (if you are thinking of `<br>` and such)
#
KartikPrabhu well, I don't understand the definition so...
#
KartikPrabhu here is another bunch of tests for whitespace while we were deciding the "correct" behaviour https://pin13.net/mf2/whitespace.html
#
[vendan] concat all text node descendents, in tree order
#
KartikPrabhu yeah that turned out to be not useful/ideal in practice
#
[vendan] not saying otherwise, just saying that `textContent` is non-ambiguous. It's just not what most people would expect from looking at the displayed value in a browser
#
[vendan] there's a bunch of other things, like you can hide a node and it won't be visible in a browser, but textContent would have it
#
[vendan] I don't personally have a preference either way, I just think it needs to be encoded into the parsing spec and tests set up for it
#
KartikPrabhu [vendan]: yes. http://microformats.org/wiki/textcontent-parsing is a draft spec for it. not sure which parser is implementing it fully. Possibly phpmf2
#
KartikPrabhu mf2py might also be doing some parts of it. I have forgotten how much
#
KartikPrabhu the tests might not have any of this
[Michael_Beckwit joined the channel
#
[vendan] yeah, doesn't match up with the tests
#
KartikPrabhu [vendan]: maybe as a first pass ignore the whitespace diffs?
#
KartikPrabhu since the rest should be fairly speced out
#
[vendan] mostly everything else is matching
#
[vendan] like, so far I'm seeing whitespace issues and the php parser doesn't understand MFv1 geo
#
KartikPrabhu aah so the tests probably do not test for some new stuff. like including the img alt and all that
#
[vendan] `Latest commit 92b5893 on Sep 12, 2018`
#
[vendan] so yeah, probably a little out of date
#
KartikPrabhu oh hmm the mf2py you are using also does not do alt-text parsing, weird
#
KartikPrabhu oh that is still behind a flag in mf2py
#
[vendan] also looks like there might be an issue with mf2py not doing relative path resolving for photos
#
KartikPrabhu [vendan]: why is php showing fail on the microformats-v1/geo/abbrpattern test?
#
KartikPrabhu it "looks" the same except for ordering
#
[vendan] ehhh, all the stuff inside the items array should be red? which means it's "deleted" by the diff of expected -> php output
#
[vendan] php parser doesn't seem to handle the v1 geo stuff at all
#
KartikPrabhu oh!! I totally mis-read the red part as being the "wrong part"
#
KartikPrabhu but "red = missing"
#
[vendan] yeah, it's like a red/green diff of a commit on github or something
#
[vendan] wide open to suggestions for rendering that kind of thing better
#
KartikPrabhu got it
#
[vendan] also, still building out all the diffing stuff some
#
KartikPrabhu also mf2py is not resolving the relative path for photos because those test don't have a base url
#
[vendan] I specify a base url for parsing of `http://example.com` for everything
#
[vendan] and it works for `<img class="h-card" alt="Jane Doe" src="jane.jpeg"/>`
#
[vendan] just not for the object `data="jane.jpeg"` ones
#
KartikPrabhu aah so mf2py might not be doing it for those hmmm
#
KartikPrabhu that is a bug!
#
[vendan] woot! win #1 for better test suite stuff!
#
Loqi 😊
#
[vendan] next steps: clean up diffs, implement "base output calculation" for diffs of non-test suite parses, add more parsers, and start cleaning up and submitting changes for the test suite
#
KartikPrabhu yup confirmed. you should file that bug at https://github.com/microformats/mf2py with the example
#
KartikPrabhu vendan++
#
Loqi vendan has 2 karma over the last year
#
[vendan] bug filed: https://github.com/microformats/mf2py/issues/153
#
KartikPrabhu there also seem to be conflicting datetime formats