kmarekspartz

Expanding our capacity to respond to unforeseen changes

Adjusting my metrics to better reflect my goals.

A case of accidental prioritization.

In January, I started using Beeminder to track my blogging. For each post, I've been adding a data point with a value of one, with a goal pace of about four posts per week. However, because each post had the same value, I had inadvertently set my sights for quantity over quality. If it doesn't matter how well I write, quality suffers at the expense of quantity[^2].

[^2]: Once I realized I was doing this, both quantity and quality plummeted.

In order to get around this, I should change what is being measured. Naively, I could change my goal to be N words per week (where N is initially my historic rate), but this would result in a different accidental prioritization of quantity over quality: I would write longer posts, not necessarily better posts.

Instead, I'm developing a quality measurement using natural language processing, and incorporating it into mindfeed. I'm unsure how exactly to calculate quality, so instead of guessing, I'm developing a set of features each of which I'll mind[^1] with Beeminder separately. I'll then manually rate my previous posts and develop a model which can assign a rating to a post using the features. Eventually, I may ask some prototypical readers to rate my posts. I could then develop a model for each set of ratings.

[^1]: mind, verb: to track data with Beeminder.

One feature will be length, thought as mentioned above, it cannot be used on its own. I'll also look at vocabulary diversity using the count of types as well as an information-theoretic approach. I will also be tracking readability measures such as the Flesch–Kincaid index. Another aspect to consider is uniqueness relative to previous posts. This can be simple, such as how many new words were introduced in this blog post, or more complex, such as vector-space distance.

Measurements are important, as they do change behavior, but what we measure can sometimes have unintended consequences. Choose your metrics carefully.

Swift has improved the syntax for its Array type.

I keep forgetting to mention that Apple has fixed one of the issues I had with Swift. In Beta 3, the syntax for Array at the type level has changed from String[] to [String], which corresponds better to the value-level syntax for Arrays. Credit to ThriftySwift.

Edit: I should also link to Apple's blog post announcing the changes.

Danny Gratzer has been reviewing packages on Hackage.

Danny Gratzer, the organizer of HaskellMN, has been reviewing packages from Hackage to demonstrate his approach to understanding new code. So far he's reviewed logict, a logic programming monad, and extensible-effects, an alternative to the monad-transformer approach. They're good reads even if you don't know any Haskell, as he describes how he traverses files in a code base using find and grep.

Ron Garret discusses some additional issues with LinkedIn

LinkedIn was actually asking me, not for my LinkedIn password, but for my email password. At the risk of stating what should be obvious, you should NEVER EVER give your email password to ANYONE. Anyone who has your email password owns you.

– Ron Garret, Wow, LinkedIn [h]as gone seriously evil

Previously: Unbundling LinkedIn

Edit (June 2017): I caved and recently created a LinkedIn profile.

A friend of mine is working on an understandable regular expression implementation.

Andrew Clarkson is working on a pure Python implementation of regular expressions, with the intent to use it in a lexer for compiler implementation. One of his goals is to make it understandable and clear, one thing that doesn't often happen with regular expression implementations.

I'm looking for a good email client

I've been trying out email clients for a while now and I've not been too happy with any. I'm looking for something that gives me the following:

  • Multiple accounts with a unified inbox
  • Gmail-style threading (time-sorted and flattened)
  • Fast (both to receive messages and for user interactions)
  • Stable

What I've tried:

What I'm currently using:

  • offlineimap for archiving / backup
  • mu for command-line search
  • Mail Pilot for composition and organization
  • Gmail on iOS
  • Triage for checking non-Gmail accounts and sorting (archive/keep) large amounts of email.

Mail Pilot's scheduling features are nice, but it is very buggy and slow. Additionally, the clever foldering it uses makes it difficult to use with other email clients.

If I had an email client that did the above, but also seamlessly incorporated RSS/Atom feed reading, XMPP chat, and SMS I'd be elated. Maybe I should go back to emacs...

Tom Christie gave a guide on uploading to PyPI

Yesterday, my co-worker, Tom Christie, posted a guide on uploading Python modules to PyPI. I've done this a few times, but haven't had to deal with a package as complicated as the one he was uploading, since the packages I've uploaded are pure Python. His package compiles a C program which gets called by the Python library, and provides some pickled data to minimize recomputation.

One thing I noticed was the suggestion to use twine to minimize transmission of PyPI passwords in plaintext. I'm not sure why pip doesn't use HTTPS to interact with PyPI directly, but I suspect this to be fixed upstream before too long. For now, twine provides a workaround.

Opening a dialogue about depression within the development community

The culture around software development is not nice to developers. While society tends to compensate them well financially, their mental condition is often overlooked. Given how important mental condition is to successful development, one would think employers and educators would pay closer attention and provide resources for developers. This stems from lack of awareness, which in turn is due to stigma discouraging open discussion about these issues. Let's open up that discussion.

In August, Greg Baugues from Twilio will be giving a talk on developers and depression at the Minnesota PHP user group. Now, I'm not a fan of PHP, but this may be one of the most important events for developers in the area this summer. There are quite a few tickets remaining, so please register.

String { get { switch self { ... case let .JSONString(s): return “\“” + s “\“” ... } } } }


And that's it![^1]

This makes the operation they implemented as follows:

~~~ JavaScript
for item in json["blogs"]?["blog"] {
    let id = blog[“id”]
    let name = blog[“name”]
    let needspassword = blog[“needspassword”]
    let url = blog[“url”]

    println(“Blog ID: \(id)”)
    println(“Blog Name: \(name)”)
    println(“Blog Needs Password: \(needspassword)”)
    println(“Blog URL: \(url)”)
}

That's a lot easier, once a proper JSON library is implemented, anyway. Swiftz has the start of one, similar in style to what I'm proposing here.

[^1]: Or close to it anyway. I'm simplifying here. See Swiftz for a more complete JSON library.

Some people write fragile code and some people write very structurally sound code, and this is a condition of people. I think in almost any language you can write fragile code. My definition of fragile code is, suppose you want to add a feature–good code, there's one place where you add that feature and it fits; fragile code, you've got to touch ten places.

– Ken Thompson, Coders at Work, p. 467

Enter your email to subscribe to updates.