February 2010
Feb. 22, 2010
node-v0.1.30 (via) A very significant new release of Node.js: the Twisted/Dojo-style Promise abstraction has been removed entirely, causing backwards incompatible changes to a bunch of core APIs. This means the pseudo-blocking Promise.wait() method is gone too, making it even harder to accidentally block your event loop. Instead, user-level libraries are encouraged to add Promise-style abstractions. I’m pleased to see Node sticking to the low-level stuff.
“Do” it fast! Tim Caswell’s Do library has been upgraded for compatibility with Node v0.1.30, and now has a clever Do.convert() method which wraps Node’s low-level APIs with the Do libraries “continuable” abstraction.
Feb. 25, 2010
Hg Init. Joel Spolsky’s guide to Mercurial. If you’re still using a non-distributed version control system like Subversion, this is a great introduction to the new world order.
Using Bash’s History Effectively. The HISTIGNORE environment variable is particularly useful, allowing you to suppress certain commands by specifying a pattern. This article has a tip for causing a command to be omitted from the history if you prefix it with a space.
Feb. 26, 2010
Internet Explorer Cookie Internals (FAQ). Grr... IE 6, 7 and 8 don’t support the max-age cookie argument, forcing you to use an explicit expiry date instead. This appears to affect the cache busting cookie pattern, where you set a cookie to expire in 30 seconds for any user who posts content and use the presence of that cookie to skip caches and/or send their queries to a master instead of slave database. If you have to use expires, users with incorrect system clocks may get inconsistent results. Anyone know of a workaround?
PiCloud. An interesting twist on cloud computing for Python. “import cloud; cloud.call(my_function, arguments)” serialises my_function and its arguments, pushes it up to one of their EC2 servers and hands you back a job ID which you can poll (or block on) for a response. They suggest using it for long running tasks such as web crawling or image processing.
Django Advent: Scaling Django. Mike Malone’s advice on scaling Django applications, including taking advantage of new features in 1.2.
Feb. 28, 2010
kriszyp’s node-promise. Another elegant approach to managing asynchronous flows in Node, including running things both in parallel and serial.
ClearMaps: A Mapping Framework for Data Visualization. An open source library for map visualisations using ActionScript, with an Adobe AIR based encoding tool for translating data from shapefiles in to vector data suitable for use with the library.
Unit Testing Achievements. A plugin for Python’s nose test runner that adds achievements—“Night Shift: Make a failing suite pass between 12am and 5am.”
Johnny Cache. Clever twist on ORM-level caching for Django. Johnny Cache (great name) monkey-patches Django’s QuerySet classes and caches the result of every single SELECT query in memcached with an infinite expiry time. The cache key includes a “generation” ID for each dependent database table, and the generation is changed every single time a table is updated. For apps with infrequent writes, this strategy should work really well—but if a popular table is being updated constantly the cache will be all but useless. Impressively, the system is transaction-aware—cache entries created during a transaction are held in local memory and only pushed to memcached should the transaction complete successfully.
Node.js, redis, and resque (via) Paul Gross has been experimenting with Node.js proxies for allowing web applications to be upgraded without missing any requests. Here he places all incoming HTTP requests in a redis queue, then has his backend Rails servers consume requests from the queue and push the responses back on to a queue for Node to deliver. When the backend application is upgraded, requests remain in the queue and users see a few seconds of delay before their request is handled. It’s not production ready yet (POST requests aren’t handled, for example) but it’s a very interesting approach.
Notes from a production MongoDB deployment. Notes from running MongoDB for 8 months in production, with 664 million documents spread across 72 GB master and slave servers located in two different data centers.