sartak
http://sartak.org
git flux
/talks/dcbpw-2013/git-flux/
<p><img src="git-flux.png" /></p>
<p> There are a bazillion different ways to use git, which is of course a blessing and a curse. You can use it any way you like, but that puts the burden on you to figure out what's best.</p>
<p> Well we've already done the hard work of figuring out what's best. I'm here to explain how we use git at Infinity Interactive to ship great software to our clients.</p>2013-04-20Reinstating Class::MOP's Commit History in Moose
http://sartak.org/2013/03/reinstating-class-mops-commit-history-in-moose.html
<p>Ever use <code>git blame</code> or <code>git log</code> in the Class::MOP parts of the Moose repository? If so you've probably seen Dave Rolsky's mega commit <code>38bf2a25</code>.
</p>
<pre class="text code_snippet">$ git blame lib/Class/MOP/Package.pm
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 1)
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 2) package Class::MOP::Package;
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 3)
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 4) use strict;
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 5) use warnings;
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 6)
0db1c8dc (Jesse Luehrs 2011-04-17 19:11:28 -0500 7) use Scalar::Util 'blessed', 'reftype', 'weaken';
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 8) use Carp 'confess';
0db1c8dc (Jesse Luehrs 2011-04-17 19:11:28 -0500 9) use Devel::GlobalDestruction 'in_global_destruction';
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 10) use Package::Stash;
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 11)
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 12) use base 'Class::MOP::Object';
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 13)
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 14) # creation ...
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 15)
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 16) sub initialize {
38bf2a25 (Dave Rolsky 2010-12-27 08:48:08 -0600 17) my ( $class, @args ) = @_;</pre>
<p>This commit is a straight up copy of all the Class::MOP code into the Moose repository. Unfortunately it was committed as a single (gigantic) changeset.</p>
<pre class="text code_snippet">commit 38bf2a2585e26a47c919fd4c286b7716acb51c00
Author: Dave Rolsky <autarch@urth.org>
Date: Mon Dec 27 08:48:08 2010 -0600
Merged CMOP into Moose
...</pre>
<p>diffstat says this commit changed a solid fourteen thousand lines of code across 138 files. A most productive morning for Dave!</p>
<p>This is problematic because there is a rich and quite important four years' worth of history before December 27th 2010, which is when Class::MOP was merged into the Moose repository. All of that is effectively lost because this commit is a copy instead of a more delicate merge. I don't blame Dave one bit for this; I certainly did not object when this merger was going down, nor did I appreciate how valuable a clean commit history is.</p>
<p>Why does this copy merge <em>really</em> matter anyway? Well, for example, we needed to know exactly when Moose started forbidding bare references in attribute <code>default</code> (to prove someone wrong on the Internet, that noblest of goals). I had to resort to clumsily bisecting Class::MOP releases on metacpan to find that the restriction was added in version 0.33 (August 19th 2006). I was sad that I was not able to use my usual <code>git blame</code> or <code>git log</code> tools because all Class::MOP history leads to <code>38bf2a25</code>.
</p>
<p>In an ideal world, we would be able to tweak <code>38bf2a25</code>, two years later, to add a second parent commit (namely, the last commit in Class::MOP, <code>d004c8d5</code>). This would turn it into a merge of two commits, which is how we could inject the entire Class::MOP commit history into Moose's history. From then on git would inspect both histories to produce blame reports, commit logs, and so on. Just like any other merge commit.
</p>
<p>Alas, we cannot just go ahead and change the public Moose repository to include the Class::MOP history like that. Adding a second parent to <code>38bf2a25</code> would change that commit's SHA. That would cause a cascade of changes to every subsequent commit (and their SHAs) in the two years' worth of commits since. This would in turn break everyone's git clone of the Moose repository, as well as any external pointers to commits (such as RT tickets, email archives, etc), and the rest of the civilized world.
</p>
<p>However! you, my friend, can fix it for your own checkout without screwing anybody up. git has a tool for rewiring parent commits. Two, in fact. The original tool was called <code>git graft</code> but there is a slicker, more powerful replacement called, well, <code>git replace</code>. These tools allow you to tweak individual commits in your local repository without disrupting other commits and their SHAs. This means you can still freely and painlessly share your branches and commits with GitHub, other developers, and yo momma.
</p>
<p>To do this, start by cloning up a new copy of the Moose repository for playing around. Not strictly necessary, but caution is definitely warranted here. You want to make sure this procedure works before I, through you, potentially damage your working copy.</p>
<pre class="sh code_snippet">git clone git://git.moose.perl.org/Moose.git
<span class="synStatement">cd</span> Moose
</pre>
<p>Next, fetch Class::MOP's master so its commits also exist within the Moose repository.</p>
<pre class="sh code_snippet">git remote add cmop git://git.moose.perl.org/Class-MOP.git
git fetch cmop master</pre>
<p>Finally, fix the sledgehammer-merge commit <code>38bf2a25</code> to include both its original parent commit <strong>and</strong> the last Class::MOP commit. To do that we create an entirely new commit object that is exactly like <code>38bf2a25</code> except it has that second parent commit, <code>d004c8d5</code>. Then we use <code>git replace</code> to tell git to use the new SHA (hint: it's <code>f18fded8</code>) in place of <code>38bf2a25</code>.
</p>
<pre class="sh code_snippet"><span class="synIdentifier">NEW_MERGE</span>=<span class="synPreProc">$(</span>
<span class="synSpecial"> git cat-file commit 38bf2a25</span>
<span class="synSpecial"> </span><span class="synStatement">|</span><span class="synSpecial"> perl -ple </span><span class="synStatement">'</span><span class="synConstant">/^parent / && print "parent d004c8d565f9b314da7652e9368aeb4587ffaa3d"</span><span class="synStatement">'</span>
<span class="synSpecial"> </span><span class="synStatement">|</span><span class="synSpecial"> git </span><span class="synStatement">hash</span><span class="synSpecial">-object -t commit -w --stdin</span>
<span class="synPreProc">)</span>
git replace 38bf2a25 <span class="synPreProc">$NEW_MERGE</span>
</pre>
<p>All done. Enjoy your new (old) history!</p>
<pre class="text code_snippet">$ git log --grep associated_metaclass --format='format:%h %ad %an%n %s' lib/Class/MOP
cc03c2b Sun Feb 19 12:51:48 2012 -0600 Dave Rolsky
Weaken the associated_metaclass after cloning a method.
-- post-merge commit
aa5bb36 Mon Apr 25 10:38:05 2011 -0500 Jesse Luehrs
fix setting associated_metaclass and attribute on accessor objects
-- post-merge commit
09ea7f8 Sun Aug 10 18:24:02 2008 +0000 Yuval Kogman
package_name >= associated_metaclass->name
-- pre-merge commit!
5e60726 Sun Aug 10 17:42:29 2008 +0000 Yuval Kogman
add associated_metaclass to Method
-- pre-merge commit!</pre>
<pre class="text code_snippet">$ git blame lib/Class/MOP/Package.pm
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 1)
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 2) package Class::MOP::Package;
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 3)
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 4) use strict;
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 5) use warnings;
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 6)
0db1c8dc (Jesse Luehrs 2011-04-17 19:11:28 -0500 7) use Scalar::Util 'blessed', 'reftype', 'weaken';
6d5355c3 (Stevan Little 2006-06-29 23:28:32 +0000 8) use Carp 'confess';
0db1c8dc (Jesse Luehrs 2011-04-17 19:11:28 -0500 9) use Devel::GlobalDestruction 'in_global_destruction';
407a4276 (Jesse Luehrs 2010-05-10 23:20:29 -0500 10) use Package::Stash;
2243a22b (Stevan Little 2006-06-29 18:27:47 +0000 11)
f197afa6 (Jesse Luehrs 2010-05-10 21:13:19 -0500 12) use base 'Class::MOP::Object';
6e57504d (Stevan Little 2006-08-12 06:13:02 +0000 13)
6d5355c3 (Stevan Little 2006-06-29 23:28:32 +0000 14) # creation ...
6d5355c3 (Stevan Little 2006-06-29 23:28:32 +0000 15)
6d5355c3 (Stevan Little 2006-06-29 23:28:32 +0000 16) sub initialize {
3be6bc1c (Yuval Kogman 2008-08-14 18:21:45 +0000 17) my ( $class, @args ) = @_;</pre>
<p>In closing, to answer my original question using my normal means...</p>
<pre class="text code_snippet">$ git log -p --reverse -S 'References are not allowed as default'
commit 148b469742669e1a506538200f624dcdaeeb510a
Author: Stevan Little <stevan.little@iinteractive.com>
Date: Wed Aug 16 21:32:05 2006 +0000
no ref in the defaults
...
+ (Class::MOP::Attribute::is_default_a_coderef(\%options))
+ || confess("References are not allowed as default values, you must ".
+ "wrap then in a CODE reference (ex: sub { [] } and not [])")
+ if exists $options{default} && ref $options{default};</pre>
<p>Sweet, sweet history.</p>2013-03-18Marketing 101
http://sartak.org/2013/03/marketing-101.html
<p>My girlfriend's friend just sent her a message:</p>
<blockquote>
<p>I'm reading an interesting article about the founders of GitHub. I was wondering, "Where have I heard about this?"</p>
<p>It's imprinted on my brain from all those hungover mornings drinking coffee out of that mug!</p>
</blockquote>
<a href="http://shop.github.com/products/octocat-mug">
<img width="512px" height="326px" src="http://cdn.shopify.com/s/files/1/0051/4802/products/MG_0914_1024x1024.jpg?63" />
</a>2013-03-16Give and Receive the Right Number of Gifts
http://www.perladvent.org/2012/2012-12-23.html
2012-12-23OX off the Web
http://ox.iinteractive.com/advent/2012-12-21.html
2012-12-21Configuration with OX
http://ox.iinteractive.com/advent/2012-12-16.html
2012-12-16Mooseを利用するに変えるハッカソン
http://sartak.org/2012/08/moving-to-moose-hackathon.html
<p>Oslo.pmが「<a href="http://act.yapc.eu/mtmh2012/">Moving to Moose Hackathon</a>(Mooseを利用するに変えるハッカソン)」という会議をノルウェーのプレーケストーレンにある山小屋で開催しました。私は参加したいでしたが、招待頂きました。Oslo.pm、どうもありがとうございます！
</p>
<p>会議のテーマは二つありました。1つは<a href="http://act.yapc.eu/mtmh2012/sponsors.html#perlrdf">RDF</a>に興味あるコミュニティが、Mooseを利用するにモジュールを変えるつもりでした。あと1つは、<a href="http://act.yapc.eu/mtmh2012/sponsors.html#p5mop">p5mop</a>を屈強化して、もう少しperl自体に追加できる状態に準備する予定たちました。
</p>
<h3>Day -3 (YAPC::EUの日)</h3>
<p>ドイツのフランクフルトで開催されたYAPC::EUに参加できました。初めてYAPC::EUだったし、イギリス以外ヨーロッパに行ったことがなかったから、とても嬉しいでした。YAPCで「<a href="http://sartak.org/talks/yapc-eu-2012/role-usage-patterns/">Mooseロール利用パターン</a>」というトークをやっていきました。今回のトークは、全ての前回よりよくできたな！と私は思います。きっと来年のYAPC::Asiaに行くことがてきるので、たぶん日本語版を話そうとします。今年来られないのはすみません！
</p>
<a href="http://www.flickr.com/photos/sartak/7838221830">
<img src="/img/blog/moving-to-moose-hackathon/yapc-audience.jpg" alt="" />
</a>
<h3>Day 0 (旅の日)</h3>
<p>フランクフルトからスタヴァンゲルに飛んで行きました。ホテルからフランクフルト空港に、そして飛行機に通った道が分かりにくいと思いました。Hauptbahnhof(中央というか駅)で乗り換えの標識がちょっと読めませんでした。ドイツ語は全然分からないのでした。しかも、空港のゲートを越えた後、バス(!)で飛行機に運ばされまして、道路から階段で直接に乗りました。ビックリした！</p>
<a href="http://www.flickr.com/photos/sartak/7880429950">
<img src="/img/blog/moving-to-moose-hackathon/flight-in.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880429316">
<img src="/img/blog/moving-to-moose-hackathon/oslo-bus.jpg" alt="" />
</a>
<p>結局、プレーケストーレンの山小屋に無事に到着しました。</p>
<a href="http://www.flickr.com/photos/sartak/7880431538">
<img src="/img/blog/moving-to-moose-hackathon/lodge-lake.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880431082">
<img src="/img/blog/moving-to-moose-hackathon/lodge-interior.jpg" alt="" />
</a>
<p>参加者の皆さんが、自己紹介したり、喋ったり、一緒に夕食を食べました。p5mopに興味ある参加者が「基本的に、どうやってroleという概念はp5mopの仕組みに合う？」というを夜中すぎまで論じました。</p>
<p><a href="http://twitter.com/robinsmidsrod">@robinsmidsrod</a>さんが「日本語で入力する方法を説明しませんか？」と聞いて、私はもちろん嬉しくなりました ;) その後、XPathとか<a href="http://p3rl.org/XML::Rabbit">XML::Rabbit</a>とか<a href="http://ja.wikipedia.org/wiki/Common_Locale_Data_Repository">Unicode CLDR</a>とかとかについて喋りました。
</p>
<h3>Day 1 (登山の日)</h3>
<p>早起きして、プレーケストーレンという有名な崖にハイキングしました。眺めはとっっってもきれいでした！！！</p>
<p>旅立ちは10時から16時までかかりました。たいへん疲れました。（四日後でもまだ足が痛い！）</p>
<a href="http://www.flickr.com/photos/sartak/7880432218">
<img src="/img/blog/moving-to-moose-hackathon/picnic.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880432884">
<img src="/img/blog/moving-to-moose-hackathon/rocky-forest.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880376022">
<img src="/img/blog/moving-to-moose-hackathon/wave-of-trees.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880377090">
<img src="/img/blog/moving-to-moose-hackathon/focused-cliff.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880435618">
<img src="/img/blog/moving-to-moose-hackathon/pulpit-rock.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880377758">
<img src="/img/blog/moving-to-moose-hackathon/topdown-cliff.jpg" alt="" />
</a>
<p>その後で、<a href="http://ja.wikipedia.org/wiki/フラックス_(ゲーム)">Fluxx</a>(zombie版とcthulhu版組み合わせられた)というカードゲームをやりました。計画者はプログラマーですから、法則が面白い。しかもルールが進化することまであります。
</p>
<a href="http://www.flickr.com/photos/sartak/7880437426">
<img src="/img/blog/moving-to-moose-hackathon/stevan-peek.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880438210">
<img src="/img/blog/moving-to-moose-hackathon/fluxx-rules.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7880438674">
<img src="/img/blog/moving-to-moose-hackathon/pro-perigrin.jpg" alt="" />
</a>
<h3>Day 2 (ハッカソンの一日目)</h3>
<p>朝食から昼食まで私が「Mooseロール利用パターン」というトーク拡張版を話してあげました。昼食の後で、<a href="http://twitter.com/pmichaud">@pmichaud</a>さんが<a href="http://pmichaud.com/2012/pres/mtmh2012-nqp/slides/start.html">NQP言語</a>について解説しました。プログラミング言語が作りたい方は、NQPを使うことで、きっととても早く計画できると思います。
</p>
<p><a href="http://twitter.com/perlyarg">@perlyarg</a>さんとMooseのロールテストをp5mopに変更して始まりました。p5mopではロール合成は不完全ですから、そのテストは便利だと思います。
</p>
<p>最初のp5mopの拡張として<a href="https://github.com/stevan/p5-mop/commit/04997c0c93c7">mopx::instance::tracking</a>は、私が書いたことができました。Moose版よりも簡単でした。p5mopのClassというメタクラスをサブクラスして、create_instanceというメソッドにオブジェクトを見逃さない挙動を追加されます。
</p>
<p>夕飯の後で、昨日の旅だし、朝のトークをしたから、たいへん疲れました。で、私がやったことは、Fluxxだけでした。</p>
<a href="http://www.flickr.com/photos/sartak/7880533504">
<img src="/img/blog/moving-to-moose-hackathon/artsy-stevan.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7879978022">
<img src="/img/blog/moving-to-moose-hackathon/perigrin-workstation.jpg" alt="" />
</a>
<h3>Day 3 (ハッカソンの二日目)</h3>
<p>先日書いたDTraceプローブをPerlに追加する<a href="https://rt.perl.org/rt3/Ticket/Display.html?id=114638">パッチ</a>をやっと送りました。p5pが受け入れたので、5.18からop-entry、loading-file、loaded-fileというプローブを使うようになります。
</p>
<p><a href="http://twitter.com/RubenVerborgh">@RubenVerborgh</a>さんが「正規表現を使いこなす教師いますかー！」と聞いて、「私が手伝ってみてあげましょう」と私は答えました。問題は、テストが9秒以上で実行するので、ちょっと遅すぎました。役800回呼び出したregcomp(正規表現をコンパイルする関数)というopcodeに2秒以上部分があることを<a href="http://p3rl.org/Devel::NYTProf">Devel::NYTProf</a>で見つけました。でも、ちゃんと<code>qr//</code>を使ったので、regcompがほとんど呼ばないはずだと思いました。残念ですがDevel::NYTProfがopcode内に見ることができませんが、DTraceは各C関数でも見られます。そして私がperl providerのsub-entryとsub-returnというプローブをトレースして、pid providerのC関数のentryとreturnで、遅いマッチだけ内regcompをトレースしました。発見として、700部分のregcompは速いですけど、あと100部分のregcompはその速いregcompより100x以上遅い。なに。。。？後程、遅すぎた正規表現は記録括弧がありまして、その括弧を<code>(?:...)</code>に変更することで、とてみ速くなったことができました！実行は9秒から2秒まで減りました。
</p>
<p>もう一つ問題に手伝ってあげました。AUTOLOADを使っているモジュールには、「URI::Namespace=HASH(0x7ff5fa032360) contains invalid characters for a type name. Names can contain alphanumeric character, ":", and "."」というエラーが発生していました。結局、原因は<code>$obj->type</code>呼び出して、AUTOLOADを実行するはずがあるけど、実は名前として<code>$obj</code>を渡して、タイプを定義する<code>Moose::Util::TypeConstraint::type($obj)</code>を呼び出されました。<code>no Moose::Util::TypeConstraint</code>とか<code>use namespace::autoclean</code>を追加すると、typeというメソッドはAUTOLOADを実行するので、よく解決しました。
</p>
<a href="http://www.flickr.com/photos/sartak/7880605988">
<img src="/img/blog/moving-to-moose-hackathon/doy-nothingmuch.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7886214546">
<img src="/img/blog/moving-to-moose-hackathon/reminiscing.jpg" alt="" />
</a>
<h3>Day 4 (ハッカソンの三日目)</h3>
<p>p5mopは<code>has $foo</code>宣言でアトリビュート定義します。メソッドを実行する時に、PadWalkerでオブジェクトの値をその変数に代入されます。メソッド内には、正しく変数を使うことができます。でも、<a href="http://twitter.com/doyster">@doyster</a>さんたちがbootstrapを書きなおした後で、$self(PadWalkerで代入された変数)は時々間違うオブジェクトにあるが$::SELF(localで代入された変数)は正しいオブジェクトにあることになってしまいました。@doysterと<a href="http://twitter.com/stevanlittle">@stevanlittle</a>と<a href="http://twitter.com/nothingmuch">@nothingmuch</a>と私とデバッグしようとしましたが、結局原因を見つかれませんでした。
</p>
<p>しかし、<a href="http://twitter.com/perldition">@perldition</a>さんが書いている新しいモジュールにすることで、解決できるはずです。そのモジュールはlexical padで手動的に管理できます。具体的に、メソッド実行時に、カスタムpadをスタックにプッシュされ、そのブロックがオブジェクトの値を使えるになります。メソッドが返す時に、カスタムpadをポップされます。このモジュールでは、変数汚染が抜けられるはずです。
</p>
<a href="http://www.flickr.com/photos/sartak/7886214156">
<img src="/img/blog/moving-to-moose-hackathon/roof-lawn.jpg" alt="" />
</a>
<h3>Day 5 (帰宅の日)</h3>
<p>空港まで無事に着きました。でも、チェックインしようとしたら、「目的地：London/Gatwick」と表示されました。クソォォ！！と思って、不安感じきました。London/<em>Heathrow</em>からボストンまで飛ぶ飛行に乗り換える予定たったのです。Heathrowのターミナル5に乗り換えるのは、二時間でもちょっと足りないのですね。GatwickからHeathrowまで乗り換えるのは絶対無理です！原因は、予約には「London」だけ示して、ロンドンは航空5つまであると、全然考えなかったのです。
</p>
<p>とにかく、カスタマーサービスに「飛行機の予約を変更してくれないか？」と聞いて、「新しい予約を取るしかできません」と答えました。で、新しくて高い予約を取りました。フランクフルトの代わりに、オスロで乗り換えまして、目的地はHeathrowにしました。</p>
<p>怒りを減るために、Heathrowでビールと二番好きなチキンカツカレーを食べました。結局、定刻通りに家に帰ったことができました。</p>
<a href="http://www.flickr.com/photos/sartak/7902234762/">
<img src="/img/blog/moving-to-moose-hackathon/tau-rainbow.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7902236444/">
<img src="/img/blog/moving-to-moose-hackathon/outbound-ferry.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7902233400/">
<img src="/img/blog/moving-to-moose-hackathon/comfort-food.jpg" alt="" />
</a>
<a href="http://www.flickr.com/photos/sartak/7902238394/">
<img src="/img/blog/moving-to-moose-hackathon/high-life.jpg" alt="" />
</a>2012-08-31Moose Role Usage Patterns
/talks/yapc-eu-2012/role-usage-patterns/
<p>Roles are one of the most exciting and powerful features provided by Moose, but also one of the most misunderstood. This talk will explore, in depth, some common usage patterns (and antipatterns) for roles and how best to use them in the design of your classes. I will also talk about the philosophy of roles and how they fit in with the larger OO toolset.</p>2012-08-22Why Dragon Quest Never Took Off in America
http://sartak.org/2012/08/why-dragon-quest-never-took-off-in-america.html
<p>This is the <em>first</em> dungeon.
</p>
<a href="http://www.flickr.com/photos/sartak/7713651052/">
<img src="/img/blog/why-dragon-quest-never-took-off-in-america/maze-1.jpg" alt="" />
</a>
<p>Most other games' dungeons are a joke. But Dragon Quest I made me bust out the graph paper immediately.</p>
<p>I think I'm going to love this game.</p>2012-08-04My Study Calendar
http://sartak.org/2012/07/my-study-calendar.html
<p>I've been studying Japanese for three years now. There are four tools that I have found to be invaluable in making me productive with the language. (I'm still trying to figure out how to be modest about it.)</p>
<p>The first is <a href="http://ankisrs.net"><strong>Anki</strong></a>, which I've discussed at length in some <a href="/2010/01/on-learning.html">old</a> <a href="/2010/04/learning-japanese-with-sentences.html">blog</a> <a href="/2012/02/how-many-sentences-teaches-a-word.html">posts</a>. Anki is the lynchpin of my study. In the right hands, it offers more than its primary focus of maintaining knowledge efficiently. It <a href="/2012/03/near-miss-words.html">secretly</a> <a href="/2012/02/how-many-sentences-teaches-a-word.html">doubles</a> as an external, <em>queryable</em> database that represents exactly the Japanese that I understand right now (with enough historical information that I can fudge in order to make progress charts). Anki is the <i>i</i> in the <a href="http://en.wikipedia.org/wiki/Input_Hypothesis#Input_hypothesis"><i>i+1</i> input hypothesis</a>.
</p>
<p>The second tool is the <em><a href="http://www.amazon.com/gp/product/0824835921/ref=as_li_ss_tl?ie=UTF8&camp=1789&creative=390957&creativeASIN=0824835921&linkCode=as2&tag=sartak-20"><strong>Remembering the Kanji</strong></a></em> system for learning kanji and its community-created companion website <a href="http://kanji.koohii.com">Reviewing the Kanji</a>. Last weekend I finished the <a href="http://www.amazon.com/gp/product/0824831675/ref=as_li_ss_tl?ie=UTF8&camp=1789&creative=390957&creativeASIN=0824831675&linkCode=as2&tag=sartak-20">third book</a> in the series which covers characters 2049(此) through 3030(鬱). This is essentially college-level proficiency. So by completing RTK3 I've finished systematically studying and learning the kanji themselves. What's left is acquiring vocabulary, and lots of it. Happily, the fruits of all the Heisig study is that kanji helps <em>massively</em> rather than hinders.
</p>
<p>The <em>fourth</em> is my own <strong>sentence corpus</strong> system, which I still have not devoted the appropriate amount of time to explain. Some day, but not today!
</p>
<h2>A Calendar</h2>
<p>The third tool is the focus of this article. It's important to expose oneself to a variety of media. There's no substitute for immersion in acquiring a second language. I need to make sure that I'm regularly reading articles and novels, practicing writing, watching movies, learning song lyrics (for <a href="http://www.flickr.com/photos/sartak/4628540837/">karaoke</a> purposes), playing video games, and everything else I enjoy doing. Just... in Japanese.
</p>
<p>To track my progress, to keep me honest about my immersion, and to decide what kinds of study I should do today, I maintain a <strong>study calendar</strong>. It used to take the form of a cheap whiteboard, inspired directly by <a href="http://gilesbowkett.blogspot.com/2011/01/my-habit-calendar-system-levelling-up.html">Giles</a> <a href="http://gilesbowkett.blogspot.com/2010/02/calendar-win-rapid-course-correction.html">Bowkett</a>.
</p>
<a href="http://www.flickr.com/photos/sartak/5998087904/in/photostream">
<img src="/img/blog/my-study-calendar/whiteboard.jpg" alt="" />
</a>
<p>The whiteboard calendar served me well for over a year. But it had many glaring problems:</p>
<ul>
<li>As you can see, it was messy. It would only support about a dozen goals before it became a wreck. Actually, more like five goals.</li>
<li>I could look at it and update it only when I was home. This meant that when I traveled my study went to shit.</li>
<li>It didn't support metadata very well, though I certainly tried! Those numbers on each day represent the number of sentence cards I added to Anki (red), the number of words I learned (blue), and the number of kanji I learned (green).</li>
<li>It supported daily (horizontal) and weekly (vertical) goals well, but any other schedule was a pain to manage.</li>
<li>It wasn't computerized so I couldn't extract any statistics. Not even streaks. Of course I also threw away all my data every month, which led to weird mental side effects depending on how far into the month it was.</li>
<li>It was... accident-prone.</li>
</ul>
<p>I'd been trying to get away from the physical whiteboard calendar because I'm whatever the opposite of a Luddite is. I've collected quite a few crummy habit-building iPhone calendar apps.</p>
<h2><a href="http://tdp.me">The Daily Practice</a>
</h2>
<p>Thankfully (and finally!), a worthy replacement has come out, and I've been on it exclusively for six months now. I even threw away the whiteboard in a symbolic gesture of victory. The new system is called <a href="http://tdp.me">The Daily Practice</a> and it looks like this (plus purple translations I added in post):
</p>
<img src="/img/blog/my-study-calendar/full.png" alt="" />
<p>As you can see it solves the issues I had with my whiteboard. The quotation icons indicate notes, so I can track what I'm actually reading, watching, writing, etc. I can consult this calendar and update it from anywhere I have an internet connection, including from my phone. The lightly-shaded blocks are "free" days based on the independent schedule each goal has.</p>
<p>This is the current set of commitments I demand of myself. I wrote them in Japanese because that is most definitely how I roll.</p>
<dl>
<dt>2ch</dt>
<dd>Read <a href="http://en.wikipedia.org/wiki/2channel">2chan</a> every day. "<em>It has gained significant influence in Japanese society, comparable to that of traditional mass media such as television, radio, and magazines</em>". I'd be robbing myself of important arguing and trolling tactics if I ignored this!
</dd>
<dt>記事 (article)</dt>
<dd>Read an article-length document every day. This usually comes in the form of a <a href="http://ja.wikipedia.org/wiki/メインページ">Japanese Wikipedia</a> article or a blog post.
</dd>
<dt>書く (write)</dt>
<dd>Write every day. Sorry about <a href="http://twitter.com/search/from%3Asartak%20lang%3Aja">all the Japanese tweets</a>. :) (not really sorry)
</dd>
<dt>アンキの勉強 (anki)</dt>
<dd>Do my Anki reviews every day. I have an unbroken streak of 164 days, which is how long I've been using The Daily Practice. It's also the <a href="http://twitter.com/jshirley/status/228887804485197824">longest streak on the site</a>!
</dd>
<dt>漢字 (kanji)</dt>
<dd>Learn at least one kanji every day. This uses the Heisig system I described above as tool number two. It has a recycling widget because I'm done! Once my last checkmark falls off the weekly calendar, so too will the goal itself. But the data is still there. That's a great bit of design.</dd>
<dt>カードを加える (cards)</dt>
<dd>Add sentence cards to Anki every day.</dd>
<dt>単語 (vocabulary)</dt>
<dd>Learn at least one word every day. I keep track of my active vocabulary <a href="https://github.com/sartak/vocabulary/commits">on github</a>.
</dd>
<dt>ゲーム (game)</dt>
<dd>Play a video game every <em>other</em> day. Currently playing Final Fantasy VII on my newly-imported PlayStation 2.
</dd>
<dt>読む (read book)</dt>
<dd>Read part of a book or manga every <em>other</em> day. <a href="http://www.goodreads.com/user/show/5099248-shawn-moore">Currently reading</a> Dune and Harry Potter and the Goblet of Fire.
</dd>
<dt>映画を観る (movie)</dt>
<dd>Watch a movie or television every <em>fourth</em> day. Mostly I watch American movies that I already like dubbed into Japanese. But I've noticed I can follow along with new content too, watching Miyazaki and Kurosawa.
</dd>
<dt>歌詞 (lyrics)</dt>
<dd>Learn part of a song every <em>week</em>.
</dd>
<dt>漢検 (kanken test)</dt>
<dd>Take the <a href="http://en.wikipedia.org/wiki/Kanji_kentei">kanji aptitude test</a> every <em>week</em>. I have a <a href="http://traveljapanblog.com/wordpress/2009/04/kanken-ds-3-deluxe/">kanken DS game</a> and I'm currently struggling with <a href="http://en.wikipedia.org/wiki/Kanji_kentei#Level_7">level 7</a>.
</dd>
<dt>資料を集める (content)</dt>
<dd>Acquire new Japanese content every <em>week</em>. This includes books, movies, music, games. I've got to keep topping up my Japanese gas tank to keep studying.
</dd>
</dl>
<p>The numbers on the far right are streak counts. You'll notice that most of the daily goals have a streak of around 30, the every-other-day goals around 20, and the less frequent goals around 10. This is because about a month ago I noticed my study calendar looked like this:</p>
<img src="/img/blog/my-study-calendar/barren.png" alt="" />
<p>As you can see, this time period is barren. There are multiple days where I did only the bare essentials: my Anki reviews and learning a couple kanji. It was depressing to look at.</p>
<p>So <a href="http://twitter.com/sartak/status/220287719706853376">I fixed it</a>! And now a month later I still haven't broken the chains in my study calendar. My appetite for study waxes and wanes, but externalizing what I have <em>actually</em> done encourages honesty and leads me to course-correct in this way. The Daily Practice answers the question "What should I do today?" for me. Without it, the answer is usually "goof around on the internet".
</p>2012-07-28