The big thing here was to start up some JS to use the Bookie Parser api in order to load the readable content of a website as you’re bookmarking it from the edit page. In this way, users of the bookmarklet will have a better experience as they can now see their article, but it’s shown in cleaned up readable form. I need to clean it up and catch some edge/error cases, but it’s a start. Once it’s solid we can then use that content to store the page content and have immediate readable results instead of waiting for the next cron job to run in the background.
I spent some time cleaning up the CSS. I did some research on the most readable fonts for screens and surprisingly, it seems that sans serif wins on digital displays. So I updated the CSS and combined with some work on the Bookie main CSS files to make the readable pages a bit nicer. I’ve still got some more cleanup to do, but it reads a bit nicer now.
I also fixed the html generated to not have the empty body tag. It was due to the way the readable parsing library was giving me a full html document of content. See the updates over there for some bigger updates.
Finally, I added a form on the main page so you can try it out on a url just by entering it. So if you’re just curious what it does, go try it out!
Just added a ping command. It should help make sure that the configuration is correct for new users. It’s also a nice start to a non-admin specific api command. A little bit of cleanup aside from that, but nothing major.
Currently, Bookie uses a library called decruft for parsing html pages for the actual important article content. The bookie_parser project is using a different fork of that called readability_lxml. The author is a bit open to merging changes in and actually says she’s in ‘maintenance mode’. Since I kind of want a really decent library for this, it’s an important feature, I started hacking on it. In the process, this is where my week of hacking went.
First I updated it to allow me to get back only a partial html document vs an entire <html> doc. I then fixed some bugs, started cleaning up the code (adding tests, making the command line client all nice and argepare’y) etc. In the process I noticed that there’s a big branch in Github that adds a ton of things like multiple page document support and such. I’ve started to try to pull his branch into my work and the origin author’s code. It’s a LOT of git cherry-pick and really a pain since I want to clean up the code as I go. Unfortunately, this just means that Git gets confused on future merges since the code’s changed between commits. Ugh!
I’m about half way done though and I hope this will leave us with one solid library to do this parsing. I’m hoping to kind of take over stewardship of the library as I complete this work. It should hopefully make Bookie and bookie_parser all the more awesome.
The coming week
hacking time will be a bit less since I’ve got a presentation to prepare for. Next week’s status report might be a bit light and boring, but hey, maybe I’ll scrounge up some more beta users of Bookie while at the conference.
Ok, I’m overdue for a ‘weekly’ status report. I’m going to try to kick this back into gear as it helps you out there track things and me feel like I’m moving forward by writing down all the little things I’ve done over the last bit.
This is meant to start taking over the work of reading the page content and readable parsing the important content out. It was a chance to play with Tornado and Heroku. This also means that in the future I’ll be able to scale out the readable processing serperatly from the main Bookie website and host. It’s pretty bare bones right now and doesn’t directly talk to Bookie, but I’ll look at adding that integration soon as the API stabilizes and I get more tests going in it.
So far the Heroku bit has been pretty awesome. I have to deal with the fact that the app gets shut down and has to restart on first request, but hopefully that gets better as traffic and use picks up. You can tinker with it at http://readable.bmark.us
I’ve been wanting to start up a command line client for some of the Bookie work. The big thing is that I need tools to help manage invites and such. So it’s currently very admin centric, but eventually I’d like to get this into a ncurses cool command line interface to pull up recent bookmarks and even do some quick searches via the API. Aren’t API’s cool. This will also contain the reference Python API implementations so we’ll have two implementations soon. One in JS and one in Python.
I’ve got a beta version (which is really an alpha) up on PyPi so you can
$ pip install bookie_api
$ bookie ping
Build baby build
I spent some quality time with http://build.bmark.us to get the JS tests running via grover and phantomjs and that’s awesome. I also added the new projects into the builder as well. So, while I don’t have all the tests I need, at least now the ones I do have run consistantly.
Other little tweaks
Prettied up the new user invite email and landing page
Fixed a bug with dupe tags in the tagcontroller
Added more icons from the fontawesome set to pretty up the ui, especially the account page.
Lots of changes to the make/build steps for JS and CSS including actually doing the pyscss transition.
Everything is now on the final stable release of YUI 3.5. It’s been a good ride through the development releases.
I’ll be giving a talk at Penguicon on using YUI for JS app development. If you’re in the area stop by. This is Friday April 27th, at 6pm. Then on Saturday I’ve got a Bookie mini-sprint going on. I’ll probably be hacking most of the weekend. Feel free to stop by and check things out.