Repository navigation
Recent Twitter char counting changes 😭 #8
Description
evidently, along with the recent switch to 280 chars, twitter also changed the way they count chars. (IRC discussion.) it's no longer simply unicode code points, it's some weighted thing i don't understand yet. one big effect is that emoji seem to now count for two chars, not one.
this made e.g. @tantek's recent bridgy publish attempt fail. we tried to publish this content:
This is meditation. #FBF to last Saturday, in the zone, racing the Mt. Tam Trail Run half marathon with singular focus. 📷 chasquirunner.com
Awake the night before @TheNorthFace #ECSCA, thinking back to last Saturday’s challenges and triumphs, this…
http://tantek.com/2017/321/t2/last-saturday-racing-tam-half-marathon
which twitter 403ed it with "Tweet needs to be a bit shorter."
if you count chars normally, it's 319 total, - 17 + 23 for chasquirunner.com, - 68 + 23 for the permalink, == 280.
however, if you paste it into the twitter UI, that says it's 3 chars over. deleting the camera and ellipsis emoji each drop it by 2. i haven't found the last extra char yet.
goddammit.
Activity
snarfed commentedon Nov 18, 2017
ok, after reading the docs, i understand the new way. some chars (code points) now count for two instead of one, and they have a data driven config that determines which are which. it'll take a little work to implement, but not a ton. seems doable.
https://developer.twitter.com/en/docs/developer-utilities/twitter-text
update for twitter's new weighted char counting and 280 char limit
update for twitter's new weighted char counting and 280 char limit
update for twitter's new weighted char counting and 280 char limit
update for twitter's new weighted char counting and 280 char limit
switch brevity to github master for new twitter weighted char counting
new test case for weighted char counting, all chinese chars
snarfed commentedon Nov 24, 2017
reopening, this is still buggy. :/
added a failing test case in karadaisy/brevity-testcases#3. the problem is that the new
str_length()returns weighted count, but we also use it to index into the python string, which we shouldn't. those are separate measurements that can't be mixed.