Alex Miller Keynote "The Amazing Clojure"

Summary

このスクリプトでは、クロージャー言語のコミュニティにおける発表の概要が紹介されています。スピーカーはクロージャー言語のコミュニティの協力精神を高く評価し、クロージャー言語がどのようにして素晴らしい場所になるようになったかについて話します。また、クロージャー言語の様々な側面と、将来性についても触れています。さらに、クロージャー言語がどのようにしてデータサイエンスやインタラクティブな開発などの分野で優れた選択肢になる可能性についても議論されています。

Takeaways

• 🎉 大会の紹介と参加者の感謝の言葉から始まります。スピーカーは、コミュニティグループを集めるBig Tenアプローチを高く評価しています。
• 🏗️ スピーカーは、クロージャーを使用して迷路を生成し、それを解くというアイデアを持っていましたが、結果的にうまくいかなかったと反省しています。
• 📚 「The Mazes for Programmers」という本を参考に、迷路アルゴリズムについて話しました。特に、簡単なアルゴリズムと再帰的バックトレーカーアルゴリズムに焦点を当てています。
• 👀 スピーカーは、迷路の解くプロセスで使用されるセルの数が少なく、効率性が低いことに気づきました。
• 🧵 ユニCURSALな迷路について説明し、それらは再帰的バックトレーキングを使用して作成できますが、計算コストが高く、時間がかかることに触れています。
• 💡 ディクスターのアルゴリズムを使用して迷路を解決する方法を紹介しました。これは、迷路を塗りつぶし、最短経路を記録することで行われます。
• 🎂 Clojureの15周年を祝うセレモニーが開催され、コミュニティの貢献者に対して感謝の意が示されました。
• 🌟 Clojureの強みとして、問題解決へのアプローチ、言語の安定性、柔軟性、コミュニティの活発さ、そして高い給与レベルが挙げられました。
• 🛠️ Clojureのツールとエディタは豊富で、コミュニティによってサポートされており、新しい開発者にとってはスタートが容易になっています。
• 🚀 Clojureの未来の重点として、データサイエンスへの焦点、対話型開発の強化、JVMとの相互運用性の改善、そしてツールや開発者の経験の向上が挙げられました。
• 🌐 Clojureのコミュニティは活発であり、継続的なサポートとオープンソース開発者への支援が行われています。

Q & A

プレゼンターはなぜクロージャー言語に興味を持ったのですか？

プレゼンターはクロージャー言語がどのようにして複雑さを軽減し、システムを構築するプログラマーの問題に対処するのかに興味を持っており、それが彼らをクロージャー言語に惹きつけた要因です。

「The Amazing Closure」というプレゼンテーションの主なテーマは何ですか？

主なテーマは、クロージャー言語がどのようにして素晴らしいものであるか、そしてその言語を使って迷路（maze）を生成したり解決したりする実験についてです。

プレゼンターはなぜ迷路を生成するアイデアを諦めましたか？

プレゼンターは迷路を生成するアイデアを試みましたが、話の順序が分からないためにナラティブ構造を作り出すことが難しく、結果的に良い結果を出すことができなかったため諦めました。

迷路アルゴリズムの中で紹介された最も単純なアルゴリズムは何ですか？

最も単純な迷路アルゴリズムは、左上から始めて右へ移動し、行を作りながらランダムに右へ進むか下に分岐するかを決定するものです。

プレゼンターが好きな迷路アルゴリズムは何ですか？

プレゼンターが最も好きな迷路アルゴリズムは再帰的バックトレーカー（recursive backtracker）アルゴリズムで、これは迷路を渦巻き状に生成するもので美しく見えると感じています。

ユニカーサル迷路（unicursal maze）とは何ですか？

ユニカーサル迷路は、スペースを埋めるようなカーブで構成される迷路で、エントリーとエグジットが隣り合っており、一つのパスをたどるだけで迷路全体を通過できます。

プレゼンターはなぜクロージャー言語が好きなのですか？

プレゼンターはクロージャー言語がドメインの複雑性と人的誤解を軽減し、プログラミング言語の問題に対処するデザイン原則を持っているため好きです。

クロージャー言語の特徴として挙げられたものには何がありますか？

クロージャー言語の特徴には、不変データ構造、遅延評価、マクロ、ホモアイオスタシー、オープンマップ、オープン多相性などがあります。

プレゼンターはなぜリッチ・ひげ（Rich Hickey）に感謝していますか？

プレゼンターはリッチ・ひげがクロージャー言語を作成し、彼の問題解決へのアプローチが非常に重要だと考えているため、感謝しています。

クロージャー言語のコミュニティの特徴は何ですか？

クロージャー言語のコミュニティは活気に満ちており、多くの人々が協力して新しいアイデアを出し、助け合いながらプロジェクトを進めています。

クロージャー言語はどのようにして安定性を保っていますか？

クロージャー言語は新しい機能を追加する「accretion」、より寛容な「relaxation」、既存の機能を壊さずに進化させるための「新しい名前の使用」などの原則に従って安定性を保っています。

クロージャー言語の柔軟性はどのように表現されますか？

クロージャー言語はJVM、JavaScript、CLRなどへのデプロイが可能で、様々な環境での使用に柔軟性があります。

クロージャー言語のコミュニティはどのようにして助成されていますか？

クロージャー言語のコミュニティは多くの企業と個人開発者の支援によって助成されており、これは他の言語コミュニティと比較してもなかなかの数の支援者を持っています。

クロージャー言語の将来についてプレゼンターはどのような考えを持っていますか？

プレゼンターはクロージャー言語がデータサイエンス分野でもっとも使われるべき言語であり、コミュニティがこの分野での地位を強化するべきだと考えています。

クロージャー言語の教育と普及においてどのような課題がありますか？

教育と普及においては、言語の選択が既に根強く、学校やブートキャンプでの教育が既に言語の選択に影響を与え、新しい言語を取り入れるサイクルを打破するのが難しいという課題があります。

クロージャー言語の学習においてはどのような障壁があるとプレゼンターは言及していますか？

プレゼンターはクロージャー言語の学習においては、プレフィックススタイルの記法、カッコの使い方、関数型プログラミング、不変データ、再帰的な思考などがあると言及しています。

クロージャー言語が持つとされる特徴とそれに反する誤解はどのようなものですか？

クロージャー言語は学習リソースやツール、エディターが豊富であるとされる一方で、適切なツールを見つけることが難しいという誤解もあります。また、学習曲線が緩やかであると誤解されることがありますが、実際は関数型プログラミングや不変データを学ぶ必要があるため学習曲線はあります。

クロージャー言語が今後重点を置くべき分野としてプレゼンターはどのような分野を挙げていますか？

プレゼンターは今後クロージャー言語が重点を置くべき分野として、データサイエンス、インタラクティブな開発、JVMとの相互運用性、ツールの向上、新しい開発者体験の向上などを挙げています。

クロージャー言語のコミュニティが今後どのようなイベントに参加する予定ですか？

クロージャー言語のコミュニティは、将来のイベントとして2024年4月27日と28日にクロージャーコンファレンス（Clojure Conj）に参加する予定があり、また2024年9月21日と22日にStrange Loopカンファレンスに参加する予定です。

Outlines

00:00

🎉 クロージャーコミュニティへの感謝と講演の紹介

講演者はクロージャーコミュニティへの参加を楽しんでおり、特にビッグテンアプローチと呼ばれるコミュニティグループの取り組みを高く評価しています。講演のテーマは「驚くべきクロージャー」で、迷路を生成し、解決するアイデアを元に話す予定でしたが、実行が難しくなり、代わりに迷路について話すことにしました。

05:01

🧩 迷路のアルゴリズムとコードデモ

第1部では迷路に関する話が行われ、1913年のウェブスター辞典による迷路の定義が紹介されています。次に迷路生成アルゴリズムについて説明され、特に単純な右向きのアルゴリズムと再帰的バックトレーカーアルゴリズムが紹介されています。コードデモではRebelを使用して迷路を可視化し、迷路の生成過程をステップバイステップで説明しています。

10:04

🔄 ユニカーサル迷路と迷路の解決

ユニカーサル迷路について話し、これは空間を埋める曲線であり、再帰的バックトレーキングを使用して生成されると説明します。しかしながら、デモンストレーション中に環境が壊れてしまい、デモが中断されることになります。

15:06

🤔 迷路の解決方法とクロージャーの紹介

迷路の解決方法としてDijkstraのアルゴリズムと再帰的なバックトレーキングについて説明し、その後にクロージャー言語の紹介とその15周年を祝うセレブレーションが行われています。また、クロージャー言語の創設者であるRich Hickeyに感謝の言葉を述べています。

20:07

🌟 クロージャー言語の強みとコミュニティ

クロージャー言語の強みとして、ドメインの複雑性と誤解を扱う方法、変数の中心によるプログラミングの問題点、不変データ構造のサポート、情報の多様性、名前の弱いサポートなどが挙げられます。クロージャー言語はこれらの問題に焦点を当てており、そのために開発された言語です。

25:07

📈 クロージャー言語の成長と評価

クロージャー言語が仕事で使用される割合が増加し、大企業での利用も増えていると報告されています。クロージャー言語のダウンロード数は月間200万回を超え、ユニークIPからは25万件ものダウンロードがされています。また、クロージャー言語を採用している企業が挙げられ、その中にはニューバンクが1200人以上のクロージャー開発者を抱えていることが紹介されています。

30:09

🛠️ クロージャー言語の柔軟性とツール

クロージャー言語はJVM、JavaScript、CLRなどへの展開が可能で、その柔軟性と多様なツールが紹介されています。また、エディタとツールが豊富な環境を提供していると同時に、コミュニティが活発であることも強調されています。

35:10

💰 クロージャー言語の経済的価値とコミュニティの支援

クロージャー言語は経済的価値が高いとされ、その安定性と柔軟性、そしてコミュニティの活発さはその理由の一つとされています。また、クロージャー言語に対する様々な形の支援が行われていると報告されており、その中にはニューバンクが月間で約40人の開発者を支援していることも紹介されています。

40:10

🚀 クロージャー言語の将来性と課題

クロージャー言語が抱える課題として、既存の言語への固定観念、教育機関への普及の難しさ、親和性の欠如などが挙げられます。一方で、クロージャー言語はデータサイエンス分野での採用を目指すべきであり、その柔軟性とコミュニティのサポートがその成功につながると述べています。

45:12

🛑 クロージャー言語の強みと発展の鍵

クロージャー言語は実際の問題解決にフォーカスしており、開発者の迅速なフィードバックとインタラクティブな開発を可能にしています。また、言語の安定性とコミュニティの活発さは、発展の鍵とされています。

50:13

🔧 クロージャー言語の開発者体験とツールの進化

クロージャー言語の開発者体験が向上し、ツールが進化していると報告されています。特に、データファインナビゲーション、プログラミング言語との統合、リモートツールとの連携などが挙げられます。

🌐 クロージャー言語の多様な展開とコミュニティイベント

クロージャー言語は多様な環境で展開されており、コミュニティイベントも活発に行われています。クロージャー言語のイベントであるClosure Conが開催され、またStrange Loopという別のイベントも紹介されています。

Keywords

💡クロージャー

クロージャーは、関数とその関数にアクセス可能な変数をグループ化したデータ構造です。このビデオでは、クロージャー言語の強みとコミュニティの活発さを強調しており、言語の使いやすさと柔軟性に焦点を当てています。例として、ビデオではクロージャー言語がどのようにして問題を解決し、開発者体験を向上させるかについて話されています。

💡迷路アルゴリズム

迷路アルゴリズムとは、コンピュータグラフィックスやゲーム開発などで使用される、迷路を自動生成するアルゴリズムのことです。ビデオでは迷路生成とクロージャー言語の応用について議論しており、迷路を通してクロージャー言語の柔軟性と表現力を説明しています。

💡ユニカーサル迷路

ユニカーサル迷路とは、入り口と出口が一定の迷路のタイプです。ビデオでは、ユニカーサル迷路を生成するアルゴリズムについて説明しており、これはクロージャー言語を使って迷路を表現し、解決する能力を例示するのに役立ちます。

💡再帰的バックトレーキング

再帰的バックトレーキングは、迷路生成アルゴリズムの1つで、再帰関数を使って迷路を生成する方法です。ビデオではこの手法を使って迷路を生成し、クロージャー言語での再帰的な問題解決プロセスを説明しています。

💡関数型プログラミング

関数型プログラミングは、変数の変更を避け、関数としてデータを扱うプログラミングパラダイムです。ビデオではクロージャー言語が関数型プログラミングの原則を採用しており、それがどのようにソフトウェア開発の複雑さを軽減するかを強調しています。

💡不変データ

不変データとは、一度生成されると変更できないデータ構造のことです。ビデオではクロージャー言語が不変データを採用しており、これによりソフトウェアのバグやエラーを減らし、コードの予測可能性を高める利点を説明しています。

💡コミュニティ

コミュニティは、特定の分野や興味を持つ人々が集まって交流する集団です。ビデオではクロージャー言語のコミュニティが活発であり、言語のサポートや発展に貢献していると強調しています。また、コミュニティが提供する様々なツールやライブラリについても触れています。

💡柔軟性

柔軟性とは、変化に応じて適応する能力を指します。ビデオではクロージャー言語の柔軟性が強調されており、さまざまなプラットフォームや要件に応じてコードを展開・変更できるという利点を説明しています。

💡ツール

ツールとは、ソフトウェア開発において役立つプログラムやアプリケーションのことです。ビデオではクロージャー言語とそのコミュニティが提供するツールの豊富さと、それらが開発プロセスを効率化し、簡素化するのに役立つと紹介しています。

💡データサイエンス

データサイエンスは、統計学、コンピュータサイエンス、およびドメイン知識を組み合わせてデータから洞察を得るプロセスです。ビデオではクロージャー言語がデータサイエンス分野でより好まれる言語になる可能性があると示唆しており、その適切なライブラリや機能の強化が行われていると語っています。

Highlights

Speaker expresses appreciation for the community and the Big Ten approach of Recur closure.

Introduction of the talk 'The Amazing Closure' which was initially planned to be in the form of a maze.

Decision to not proceed with the maze format for the talk due to difficulty in building a narrative structure.

Talk divided into three parts: mazes, why closure is awesome, and future closure stuff.

Mazes are defined using 1913 Webster's Dictionary, highlighting the confusion and perplexity associated with them.

Introduction of maze algorithms from the book 'Mazes for Programmers'.

Demonstration of a simple maze generation algorithm using a random walk.

Use of Rebel to create and visualize mazes with perfect properties.

Explanation of Wilson's algorithm for maze generation and its properties.

Recursive backtracker algorithm for generating mazes with pleasing curvy paths.

Challenge of generating a talk in the form of a maze and the switch to unicursal mazes.

Technical difficulties with generating an 8x8 universal maze and the switch to a smaller example.

Dexter's algorithm for maze solving and its implementation.

Closure's 15th anniversary and its focus on reducing programming complexity.

Closure's stability and its cultural value of avoiding breaking changes.

Closure's flexibility with deployment on different platforms like JVM, JavaScript, and CLR.

The thriving Closure community and its continuous growth and support.

Closure's strengths in stability, flexibility, community, and interactive development.

Opportunities for Closure in data science and the importance of interactive development.

Core team priorities for Closure focusing on interactive development, tooling, and performance.

Encouragement for the community to continue moving fast and making stuff with Closure.

Transcripts

uh thank you for the introduction and I

it's a pleasure to be here I really

enjoy uh getting together with closures

and and being able to do this and I

really appreciate the Big Ten approach

that Rec closure has done with bringing

in all the community groups and

everything else

um that's fantastic and and it's it's

been so cool to hear about all the

different

stuff and all the people who are uh

working so hard to make closure in the

closure Community a great place to be so

uh so my talk is called the amazing

closure

which you might also call the uh maze

enclosure

and I had this idea while reading this

book earlier this year

of doing a taught which is a cool book

so you should go check it out

um so I had this idea like what if I

could generate mazes with closure

what if I could solve masses with

closure

what if each cell of a maze represented

a set of slides and then the talk order

was generated from May's solution

this is a really cool idea and it's a

really bad idea also so uh I did this

and it was the results were not good so

it is very hard to build a narrative

structure uh when you don't know the

order that anything's going to be in

um so I I didn't end up doing this so

sorry about that or you're welcome

whichever one is appropriate

um

and so instead this is going to be a

talk in three parts the first part is

going to be about mazes because I did I

wrote a bunch of Base code and it was

fun and so uh I heard yesterday on a

keynote that you're supposed to do

things that are fun so we're gonna do a

little fun first and then we're going to

talk about closure and why closure is

awesome and some of what you've heard

today and we're going to talk a little

bit about uh future closer stuff

so I'm gonna go back to the vehicle the

amazing closure the amazing closure The

Mazes for programmers book do that this

stuff part one mazes all right we're all

caught up so what is a maze so if you

could actually install the 1913

Webster's Dictionary in your dictionary

which I recommend doing

um because it has really cool stuff in

it that's much more interesting than the

one than the dictionaries that are

installed by default so uh these are

some definitions for Maze that were that

are in the Webster's Dictionary uh from

1913 which really gives you a lot of the

confusion perplexity bewilderment

baffling a lot of good words there

um so uh that was indeed my experience

of trying to build a talk that was in

the form of a maze so so but I found a

solution which was to not do that so

um there are a bunch of maze algorithms

um there are probably I don't know 10 15

different ones that are in the book

maybe more if you take some of the

exercises and and ideas in there

um I'm going to talk about two of them

um this is a really simple one I think

this might be one if not the first one

that's in the book

um so the basic idea is starting from

the top left

that you move right

and then uh do so do the top row and

then do the second row third row from

left to right and at each step you

randomly decide to either extend the run

that your the passage you're curving to

the right or end the run that you're on

and carve a random passage down from any

cell in the run that you just made

um so

and I have slides here but that's no

that's no fun right let's do some stuff

and let's just let's do some code stuff

that's more fun so

um

and I'm using Rebel here

and so I have a representation of a map

of a maze which is just a map and it's

got like an adjacency Matrix kind of

thing in it and some other information

[Music]

um but that that's not fun right we need

to actually show this so I wrote a

little viewer for rebel

um to show me that maze so this is just

uh so we can make new mazes lots of

Mazes and all the mazes that we're

talking about are are almost all the

mazes we'll talk about today are perfect

mazes which generally you can pick an

entry and an exit Point anywhere on the

Maze and there should be a path there so

um so you can you can try to pick that

out yourself if you like but

um this is uh just a typical algorithm

and and then I wanted to show I wanted

to show how we're carving which I have

slides for but like I said that's no fun

so here's some code and I thought it'd

be useful if I just went into here and

put a little tap in there

that seems like a good idea right

all right so I made a new one oh look

there's all these mazes here

I'll convenient

all right so I'm gonna

um I'm gonna browse off of that tap

and then we can walk into that and we

can actually step through all the steps

so you can see that as we move to so

we're going to start in the upper left

and in each cell so the first cell we're

going to randomly decide should we carve

down or to the right so we decided to

carve right right right and then we got

to the end of a run so in this step we

carved down instead

um and that ended that run and then so

we start over at the next cell which is

this one right here

lost my there

um so these two they they both basically

carved down so we didn't so several

times I don't know we didn't we just

carved down instead of going to the

right

and we're just going to keep going

so I'm just I'm just re-walking all the

code that had all the things that

happened while I was generating that

maze right

oops Yeah so there we got to the end so

we ended up with a maze right so that's

a really simple algorithm there's some

other ones like that

um you might see that um if you look at

a few of these in a row

so let me go back to

so if you look at a few of these in a

row I'm going to make a few a bunch of

them tell me if you see any similarities

here

look at the bottom row

because the only choice when you get to

the bottom row is going right you can't

carve down all of these have a bottom

row that's open and often you end up

with a right a right column that's

mostly open so this is not there this

can't generate every possible base right

it's it has some biases to it and

there's several algorithms like that

that have some kind of bias built into

it and then there are other ones that do

not

um so this is

there's one called Wilson that makes

mazes

so that's cool

those don't have that property or

there's one called hunt and kill

so these are cool the one I like the

best is actually

um

this one which is a recursive

backtracker algorithm

so this one just tends to make things

that are pleasingly curvy uh if that

makes sense

[Music]

um and this is a very simple algorithm

actually to go in and

and where I'm at

I can't

the stupid things in the way

all right

let's

see if I could find my back tracker in

here somewhere

so same thing here oh I do I was

actually already tapping it so how

convenient

how convenient

I got lots of stuff in the backtrack now

tap now

friends that

um

starting to get a lot of data in here so

all right so the way recursive

backtracking works is uh actually you

get to the right place in here

all right so recursive backtracking

we're just going to carve as far as we

can until we run into a dead end so

do keep carving keep carving keep going

oh wait a dead end right so then you

just pick a random new place that you've

never seen before and start crying new

at it and we picked a place right in

here there wasn't far to go so now we're

going to find a new place which was up

here and now we're going to find a new

place and here we're gonna get a lot of

room to run

then you're going to keep going and

sometimes it takes a while to get

through all of this stuff but in the end

you end up with a cool maze so that's

cool so I'm gonna these are my backup

slides which we don't care about because

everything worked flawlessly

all right

um so an interesting thing I ran into as

I was doing the trying to make a talk

out of this was

um I immediately found when you looked

at the at a maze that um

that often not very many of them would

have uh not very many of the cells would

be involved in the solutions so I guess

I haven't shown this yet

but it's more fun if you solve these

mazes first so let's do that so this is

just amazed with the solution in it

but you notice that the packs the path

length here is 27 there's 64 total

squares so I'm only using like you know

maybe a third or somewhere in the

quarter to a half range in in most in

most mazes so there's a big one but it

seemed like a lot of effort for me to

make slides for all those cells and then

not use a third of them or a half of

them so that made me disappointed so

what I did and said was

um there's another kind of maze called a

unicursal maze and that is basically a

space filling curve

that

um

that will let's see this is so uni

can I do that oh I use the wrong one

[Music]

this is really cursing

news have been here

I bet you I I bet you I know exactly

what happened so I think this threw an

exception which you can't see because

you can't see all of my screen

I think this one

expects there to be a starting cell

which I defaulted other other places

hmm

I don't know what I did

not sad

oh I know exactly what I did so this is

the problem with universal mazes

um so the way you make them unicorn maze

is basically the same as recursive

backtracking

um and then there's uh well so there's

there's two ways to make it one is

actually to take

um do I have

so one is actually to

I was very

I actually missed a slide here I missed

David Bowie so

um Universal mazes are also also called

labyrinths in some places so

um you know had to get a picture David

Bowie in here

um so there's two two ways to make a

universal maze that I know of so one is

to take any perfect mace you already

have and effectively cut the pat cut all

of the paths that you have in half

um

and so as an example here here is a

perfect maze that a 4x4 perfect maze

and so basically I drew a half line

between all of the all of the existing

things and you'll see and you have to

add an extra little close you have to

close off the first spot here to make

the entry point and you then always will

end up with a universal maze that exits

right next to the entrance because we

basically have stolen you know that same

uh entrance path

um this always works uh it's fast and

it's really boring and so that's no fun

um the other way to do it is with

recursive backtracking and so just like

we did uh recursive backtracking before

um you're gonna you wanna keep going

until you hit a dead end

but instead of starting at a random

point you then back up and undo your

path to the point where you can make a

different decision

um this is kind of what's called oh my

God complexity it's it's not it's not

actually exponentially bad it's actually

much worse than that so it's you know

some sort of factorial bad if if you're

if you're super unlucky

um there are tricks so you can do things

like memorize dead paths so that if you

return to them you

um you don't revisit them you can uh you

can think about how to detect the case

where you hit a wall

and you've got live cells on either side

of you you are not going to be able to

revisit both of them so you found a dead

end you don't have to go visit the whole

inner part and stuff like that

um but it's still bad so

um the reason things aren't working is

because uh my computer is thrashing

itself to death trying to make an 8x8

Universal maze right now

um which is bad

um eight by eights take uh can take a

very long time to make so and so so that

was the problem there was that uh I did

too much

um so I'm gonna

um

I am going to

to give up and rebuild my environment

here

this happens

even to me

foreign

yeah so if anybody else there has

imposter syndrome I managed to break my

environment yesterday in a way that

nothing that I had done for the last

several weeks worked

and it took me a good half hour before I

realized what it was

what was it I'm not gonna go into

because I haven't released the code yet

so

um

foreign

let's be a little less ambitious

great

this thing takes a starting

so

let's see if that works I think I'm

connected

foreign

sorry I'm running actually a fairly

large pile of unreleased various bits of

rubble so it's it's not all entirely

baked yet

it's still not connecting for me I don't

know it doesn't matter

um we're gonna go on so

I have slides this is why we prepare so

this is an example of a slide made with

universal recursive backtracking this is

an API 8x8 my computer just accidentally

happened to be very lucky and made this

in 11 milliseconds

um but most of them took more like 20

minutes when I was making them so uh so

this is an example of one that does not

have an entry and exit that both happen

right next to each other it's got

um

yeah here so it's upper left and upper

right are the entry and exit for it so

anyways that's cool so how do we solve

mazes

um Dexter's algorithm we basically you

basically flood the maze record the

distance

um and then from the end choose a path

with decreasing distance and then you

have basically a backwards path of that

point so you just reverse it you can

also do a depth first search with

backtracking and that's basically how

the universal solver it just follows it

uh this is a example of

uh what the distances look like so this

is a pretty big maze and it's flooded

from the upper left basically finding

distances so that the color here is

basically distance from uh zero zero uh

and then if you so wherever you want to

find a path from you basically you can

pick the bottom right and just follow a

decreasing gradient of color until you

get back to the beginning

let me move this out of the way

all right so that was amazed stuff uh

that was fun I I assume you probably are

asking that looks fun can I have that

code and the answer is yes but not now

because it's a big it's a big uh it's a

big mess but uh you know it would be

really cool if it would be to have

something in the clerk Garden that uh

showed off some maze stuff so uh there

maybe I'll do that but probably I won't

have time and I'll just release the base

code somebody else can do that

uh so closure let's talk about some

closure stuff uh recently A bunch of us

got together and celebrated closures

15th anniversary we had a little we also

had a video thing that uh went out from

new bank

um unfortunately Focus was not able to

be there on this weekend so he's not

pictured and also I want to say happy

birthday to focus yesterday was his

birthday

so I will get that in there

um and

I want to uh take this opportunity to

say thank you to rishiki so

um there's a bunch of cool people in the

closure Community

um there's a lot of Amazing Ideas

um a lot of the things that I think are

um are important about closure come come

from obviously from Richard's brain

um I didn't end up at working with Rich

by accident

I looked around and and tried to find

people that uh

um

I thought were doing important work

um and I that's how well one of the ways

that I that I got to closure and

um so I want to thank him for

um making closure and

um the way that he thinks about problems

uh are really important and I say I say

this all with a little bit of

trepidation because there is a lot of uh

pushback in places about the sort of

hero worship of rich or whatever uh and

I think those are fair criticisms

um and closure is not just Rich

obviously there is a huge Community

which we have seen all day yesterday and

today

um and so there are many many people

that make closure great

um but I personally

um think rich is one of the most

important computer scientists of the

last 20 years maybe ever and it's really

all about the way that he approaches

problems and problem solving and

um making that the focus of everything

so this is my mini appreciation for Rich

so thanks Rich

um and when he started closure

um

this is a picture that he's used in some

in some other other talks and things

like that

um the important thing here was trying

to think about what are the problems

that programmers face when we build

systems and the two biggest ones were

domain complexity which is something

that is built into the domain so that's

not something that we inherently can fix

right

um and the second one is misconception

which is just working on the wrong thing

or doing solving the wrong problem

solving the problem wrong there's a lot

of different variants of this again this

is not a programming problem this is a

human problem

um but it is something to keep in mind

as you think hard about you know what am

I doing every day

um and then there's a big slew of

problems that are really important and

are the domain of programming languages

and so

um just looking at a few of these like

uh Place oriented programming the sort

of centrality of mutable places

um and mutable databases both of those

are sort of what he calls place oriented

programming and

um there are so many problems that come

out of that and so his search really was

for how do we how do we work in a way

that is not that and there were a lot of

there was already a lot of work on

immutable data structures and things

like that uh okasaki and all those kinds

of things and

it wasn't until he found the the the

hashary map tree work of uh Bagwell that

this really became a thing that he could

address in the programming language and

I would argue that he has you know

extended that to a large degree into

daytonic which takes the same sort of

approach to data in the database

um weak support for information this is

really the idea here is

um information is sparse it is uh it has

partial information

it has

um increasing information over time as

we work on things we tend to add

attributes to it

um open so open Maps open polymorphism

all those kinds of things it all comes

out of this

um and I'm not going to talk about all

these weak support for names is another

really important one

um but a lot of these are very visible

in design exposure they're still visible

today there's still things we talk about

every day on the closure team or these

exact things so

um that's if if anybody you know has

enjoyed using closure to work on their

systems the focus on these problems is

why that I mean that is the big thing

um so

uh closure work so I want to go through

just some reasons that that I love

closure so closure works and I think it

is clear from the state of closure

survey this is from

um two different questions on there that

people are using closure more and more

for work they're using it more and more

in big organizations

um there are more and more downloads We

just recently crossed two million

downloads a month for the main closure

artifact

um you'll see the unique IPS are not are

are a more complicated picture

um there's a big reason for that which

is that in April we released closure 111

and in June we released closure 112

Alpha One so those two things make

people download new stuff so but you

know there's 250 000 IPS downloading and

closure from somewhere every month so

there's a lot of people out there using

closure

um there are now a bunch of closure

unicorns um new bank I'm somewhat

familiar with

um but other ones that have happened

over the last year or two and imperative

reefy Health apps flyer tax bet I'm sure

there are others that I don't even know

that are on this list

and newbank now has more than 1200

closure developers more than a thousand

microservices I don't even know how much

lines of code millions of lines of code

of closure code so they're a huge a huge

user of both closure and daytonic

um

but also I know these other companies

are as well and there's lots of other

small companies

um

this uh there's this post that Ed Weibel

wrote when

um cognitek joined new bank and

I think I revisit it periodically

because I think it's really it was

really important

um I think it really highlighted a lot

of stuff that newbank found useful

um because they they really made a very

considered choice about which language

to use

um based on managing complexity

um so this is a quote from with a few

minor edits from from that post which is

down down in the footer

um but they knew they wanted to build

something with scale they knew that the

biggest threat to scale was complexity

and they wanted to constrain and

mitigate categories of complexity and so

they were you know reading out of the

tarpet looking at databases that led to

the atomic which led them to closure so

um this is a I think a fascinating way

that a company the size of newbank you

know got started with it so it's a it's

very interesting and I don't think it's

that much different from from some of

the things that Bradford was saying in

his talk

um it we have you know closure is

squarely focused at how do we reduce the

complexity of writing and evolving our

code over time and that's

um that is our uh damper in many cases

that's the thing that stops us from

moving forward

uh closure makes us happy in the stack

Overflow survey closure is one of the

most loved languages

it is also this study that people did of

uh I think it was of

um words in

um different GitHub repos or words or

comments and issues I'm not really sure

I don't remember now exactly what the

details were but you'll see closure here

had the the most uses of you know

awesome cool fun happy helpful and

interesting and and get up repos and

issues and things like that so

um we we like stuff that's cool and it's

a happy community

to in most ways uh closure pays well

um closure has topped this survey I

think every year that we have been

included in it

um there is at least one or two years

where we were not

um so thank you to everybody who keeps

filling it out it uh is really important

for us to show up and be visible in

surveys like this and so I know it's a

pain to go fill out surveys and

sometimes they're long and all that but

it really does help in making us visible

um closure is stable so

um Rich has talked about this a link

breaking changes are breaking so don't

do that instead do things like accretion

which is adding new things relaxation

which is being more tolerant in what you

accept or make a new name instead of

breaking a thing that you have this has

become a cultural value both in the

language and in the libraries

um and it is to all of our benefits so

keep doing it it's awesome

um there are some language features that

help with this things like open Maps

open polymorphism avoiding concretions

all those are things that help with

um help with making closure stable

um some some examples that people talk

about a lot

um riches Rich wrote uh this demo of

ants to a sort of demo some of the

concurrency features of closure which

was I don't know the code dates till you

know 2009 or 10 or something like that

you can find it out there and adjust and

it still runs today uh and the exact

same code runs today

um in current closure just like it did

in closure pre 1.0 wherever that was

um Stu's also famously mentioned that

the code for programming closure which

was the first closure book

continues to uh run with every new

release and and he hasn't modified it

too much I think that's actually

slightly wrong I think there were some

examples that actually did break with

the 1.3 numerix changes but uh in

general that is largely true still

um you may have seen this graph floating

around it's been posted on Twitter and

things like that uh it was this is from

the history of closure paper

um I did the work to make this graph

which involved a pile of python and go

stuff

um and it ironically I went back and

tried to give that to somebody not too

long ago and everything about the python

and go environments that I had had bit

rotted so bad that I could not get any

of it to run anymore it was all

completely broken so it's a little bit

of ironing for you

um just for comparison I wanted to

include a few other ones I think are

interesting so like this is get I think

you should not be surprised by this uh

git is a very stable code base Linux

also a very stable very well-managed

code base uh this is a Scala you might

notice it looks a little different we

call this the slumping cake graph

um and then this is rails which you

might might also suspect would look kind

of similar

um so closure is flexible you can take

closure and you can you can deploy it on

the jvm or to JavaScript or to the CLR

which actually just this week the

closure CLR 1120 Alpha One compatibility

was announced so the closure CLR still

exists I don't know that almost anyone

actually uses it but it still exists

um and David Miller continues to to bash

away at it

um babashka's brought this into the

world of scripting

um and they're you work with graul has

been really amazing and I think has

really pushed that team in interesting

and useful ways and so I'm really glad

that all of that stuff exists

um we've seen closure flutter come out

this year with uh from uh

Kristoff and Baptiste and so again

that's awesome

um and I would also include not as a

host but as sort of a

um access point uh the clj python stuff

which is uh totally awesome to get

access to the python environment

um sort of as an escape hatch

um the the syntax uh you know James

Gosling talked about this yesterday is

kind of interesting is the it separates

syntax and semantics and that ha that

has lots of Pros things like the

homologonicity and macros and all that

kind of stuff but it has the downside of

that you're never exactly sure what

you're seeing what it means and so with

lisp especially you know you can build

these towers of abstraction that are

um that can be hard for people to come

into and understand so you can build a

world that's all your own

um and the problem is that it's all your

own and so it's hard to understand I

think closure is interesting in that

they did not we did not sign up for all

of that flexibility so

um things like uh reader macros custom

reader macros are a thing that

um definitely Fosters this in the in the

commonless world but doesn't exist in in

the closure world so that's that's an

interesting case where we've not not

taking on all of that flexibility

um there are tons of tools um and I want

to mention like we've seen all through

this conference if you've been if you've

watched any part of it you have seen

tons of tools and they're awesome and

you've seen tools talking to each other

in environments tools embedded in

environments and environments embedded

in tools and all this stuff and the cool

thing to think about is what makes all

that work and the answer is Eden like

everybody meets at Eden like you can

just pass these literal data structures

around and know that like whatever tool

you give it to like if it's a closure

aware thing like it's talking Eden right

so I can give it maps and vectors that

and uh and probably even tagged literals

and things like that and have it work so

um so that is the the secret sauce in

the middle of all this stuff uh and so

keep keep building cool things on top of

that

it's an awesome substrate for that

uh the closure Community is lovely

you've seen two days of that here at

this conference

um there so thank you to all of those

people who are doing all this work many

of them are on this call

um many of them are not on the call

um there's tons of people doing stuff

it's amazing when I do the draf every

week it is just shocking to me every

week that there's you know a dozen

videos a dozen blog posts there's dozens

of libraries that are being updated

every single week literally all of the

time so there is no break from it it's

uh

it is a thriving community and anybody

who thinks closure's dead is just

completely not trying to to find it or

trying not to find it because uh uh it

is uh throbbing and alive I would say so

um that's silly

um I I sometimes I think maybe I should

just write a bot that goes out on Hacker

News every six weeks or so and suggests

that some other language is dead just so

we'll throw some more food in the water

um don't do that I'm just kidding

um oh it says sponsorship I think

closure is is maybe not unique there are

definitely

um other language environments that have

different levels of Open Source

sponsorship many of those are are driven

by big companies that have foundations

and things like that

um but but closure really does have an

amazing amount of funding for open

source work

um and I think that is uh pretty great I

I don't think the sort of Grassroots

nature of it is matched by many other

communities that I can think of

um closure us together you know I had a

had some had some time yesterday but

just to resurface that

um they are doing a ton of different

kinds of funding they have there are

dozens of companies and hundreds of

developers that are supporting it uh

that's awesome if you're not supporting

it even you know to a very small amount

and you have the ability to do so I

would encourage you to do so or if your

company uh is not currently supporting I

would encourage you to encourage them to

do that it's good for all of us

um we and we have an open source that

this sort of culture of Grassroots

support of Open Source developers as

well newbank is supporting about 40

developers on a monthly basis which we

don't really talk about enough so I'm

going to write a post about that soon

but

um we are supporting a lot of people I

know that there are lots but when I go

look at those they're you know often 5

10 20 other people supporting

um those those people on projects as

well so uh it's it's certainly not just

newbank there are a lot of people

helping helping each other to uh make

great stuff

so the future

um I should have brought my shades to

where those for that

um so

I wanted to take a few slides just think

about obstacles and spread and some of

these James talked about a little bit

yesterday but

language choices are entrenched you know

it's a big thing to choose a language

and it's very difficult for a company to

change to a new language

um or to have that you know decision

point where they even really consider

another language

um and it's very difficult to get into

schools and boot camps there's there's a

feedback cycle here that you know

schools want to teach languages that

people use people want to use languages

that are taught so it's very difficult

to break into that cycle once you do you

get a virtual cycle there and I think

python definitely you know achieved that

they're they're probably the poster

child of that

um that managed to break into that cycle

enough

um to feed sort of a flywheel of of uh

continued uh people who know the

language but

um it's it's uh it's difficult to break

into that

um we need to be doing what we can to we

the closure Community does a great job

talking to each other

um things like this event and other

events I think we are pretty well

connected in the community to each other

I think we are

um not great at sort of seeking out

external communities to talk to Rich

certainly did a lot of this in the early

days going to talk to you know python

Ruby communities or info queue or you

know strange Loop or things like that

um and we we need to keep doing that

um that's the only way that we're gonna

if you want to grow grow the grow the

bubble you got to go talk to people

outside the bubble so

um as you think about if you if you want

to help with that submitting talks at

conferences is a fantastic way to go

talk to other people that

um that haven't haven't seen closure

um familiarity is a big thing the thing

that which would call easiness

um it's things that are close at hand

right closure is not close at hand for a

lot of people who've been through

um various you know computer science or

Boot Camp or learning

um paths so the prefix style is

difficult for the reasons James talked

about yesterday

um the the parens tend to be a blocker

for some people even if it's silly it is

um functional programming and immutable

data if you go look at uh the I didn't

put it in here but the closure survey

for like what are what are your

obstacles to learning closure in

particular you'll find that functional

programming and immutable data are at

the top

um it is a different Paradigm it's a

different way of thinking about things

and thinking in recursion uh also is

another thing that's difficult getting

out of the loop mindset and into the

working at The Collection level

transforming one collection into another

collection or uh thinking about

recursive algorithms those are things

that it's very those are very natural

ways to structure programs enclosure and

they're very unnatural for for a lot of

other languages so especially the ones

that are taught so

um so there is definitely a learning

curve there

um and I think

a lot of that is unlearning you know so

there's a lot of unlearning that has to

happen uh people who pick up closure is

their first language which is not really

that common but people who do I think

find it to be not that hard so I've

heard that from a number of people who

went that path but

um because they have a little less on

learning to do

um so some ways to turn this into a

positive

um student talk a few years ago about

alien technology and I think that's a I

think talked about that or talking about

it as a superpower talking about it as a

secret leverage things like Paul

Graham's essays about lisp and how that

was their secret weapon and you know

being productive against other you know

similar competitors using other

languages like that the idea here is

there is a more productive way to work

and

um it's not a silver bullet like you're

still writing code you're still fighting

the same problems that you fight in any

other time so it's not uh you know a 10x

kind of a thing but it can definitely be

uh transformative to make this leap and

think in a different way and so

um finding the right way to talk about

that that is alluring and uh brings

people in

um and uh encourages them is the is the

trick so we need to keep doing that I

think that's a cool way to talk about it

um jobs this is a been for any small

Tech anything that's not super widely

spread

um the jobs are not

uh evenly distributed among the people

either by geography or skill level so

the people the things that are you know

there are just hot spot areas where uh

companies that work with closure are and

they typically want more senior level

people

um that's just uh that's just the nature

of this problem so the only way to

really solve that problem is to have

more companies and more developers so

um one way to do that is to hire

non-closure developers and train or

Mentor them so I mentioned before that

newbank has 1200 plus closure developers

um we didn't go out and hire 1200

closure developers right that's just

there there weren't 1200 enclosure

developers to go higher so there were

you know we hired people who were smart

and were wanting to uh wanted to do cool

stuff and we we taught him closure so

um you can do that too

um things that are not obstacles in my

opinion and you're free to fight with me

about this but

um there are so many learning materials

out there for closure people are

continually making them there are really

good old ones

um on the closure team it's people are

like people often ask us on the closure

team why don't you write more learning

material like

Stu and I wrote program enclosure a

corrode closure applied Focus co-wrote

Joy of closure like we've collectively

have written three three I think pretty

good books that if you read all those

and you'd be you'd know a lot at that

point

um they're they're and this is just

scratching the surface there are tons of

other great books out there there are

tons of great online resources out there

um same thing with tools and editors it

used to be a problem to get a good tool

it's not a problem there are there are

plenty of good tools out there and I'm

not saying they're all perfect I'm not

saying they don't all need more things

of course they do that's always the case

um the biggest problem for me is it's

hard to find the right one the one

that's right for you so they're not all

good for people from a Java background

or a python background or a data science

background or whatever so

um I think there's a lot we can do to

um sort of set up starting points and

paths particularly for people coming

from different contexts so that is a

place where I think we could do some

really good work

um and I think it's also so those are

some obstacles but why is closure

compelling if focuses on solving real

problems you get fast feedback

interactive development you can move

fast uh it's the thing that Bradford

talked about

um it's stable so you don't waste time

turning your environment and fixing

stuff that's needlessly broken which is

what happens to me every time I use any

other language pretty much

uh it is flexible you can respond to

requirements all those things about open

systems and evolution over time it's

designed to be flexible you can do a ton

of stuff the community is great you can

get out so many places now uh and the

people are awesome you can have fun

doing it uh it pays well it is low

friction these days you can get started

really really quickly I think I don't

think we do as good a job as we can

explaining that to people so I'll take

that one but uh yes I think it's uh it's

not that hard to get started these days

and

um you can deploy and reach everywhere

like the amount of ecosystem that you

can reach is is really like all of the

major ecosystems today pretty much so

it's pretty awesome

um

some key opportunity areas so it is

ridiculous that closure is not the

favored language for data science

the data language it's

um we are we are just sitting smack dab

in every place where we should be for

this and so

um my Supreme thanks go out to the

cyclos group and all the people who are

working on all those libraries and it

this is not my area of expertise I don't

know this stuff but

um the the work is getting done and we I

super support all that rich supports all

of it rich has said multiple times to me

this year closure should be the best

language for data science the end that

that and I agree so

um the more we can do on that I think

that's a place where uh data science is

not going away it's going to keep

evolving

um we need to get our foot in the door

there

um and make sure the python is not the

only option on the table because

um I want for me closure lets you get

started quickly and make stuff which is

true of python as well

um but I would much rather take the

thing that I made in closure and make

that a production worthy thing like the

path to that is way shorter than doing

it through python I just in my opinion

so

um you're going to end up with a much

better piece of software at the end uh

in that path

my opinion and uh interactive

development closure has always had this

sort of Rapid feedback cycle do the

ruffle

um the tools that people are putting

together today things like you know data

rabbit and the clerk stuff and there's

so many people building amazing tools

with closure it's not accidental it's

closures designed for this thing you

know

um we are this is this is where

um this is where we have always shined

and we should we should continue to push

it as a key thing for closure or a key

selling point for closure it it lets you

work interactively with your program uh

you know as a living thing and it is the

best part of

um of lisp and small talk and all of

that tradition and so

um we are going to keep working on that

so

uh core team priorities

um this is not coming for Rich this is

not a road map this is me just

um sort of reflecting over the things

that we have worked on in the last year

a lot of which is not out yet

um but things around interactive

development we continue to think that's

important we're continuing to do work

with datify nav

um with prouple uh Rebel is sort of a

great tool for uh doing that stuff there

are others now like portal and

um

my mind is losing me but I'm someone's

out there screaming reveal

reveal thank you sorry too many things

in my head

um so all of these take advantage of

data fine nav

um we have identified some issues around

um being able to connect Rubble remotely

I don't know if anybody caught that

earlier but I was actually connecting

Rubble remotely not in proc to calva

when I was doing the maze work

um so we have a working version of that

uh it has some really interesting

reusable Tech underneath it which we're

going to release

um and that should be useful to lots of

other people too not just to Rubble

um I've been doing a lot of work on

ad-libs recently so it's coming we're

going to get it integrated into closure

itself in some way we're trying to work

out the best ways to do that

um and some of that also hooks into

closure CLI a little bit

um so we will always continue working on

improving interop with jvm host

both to get access to new functionality

things like um Loom is a really

interesting and a lot of the value-based

stuff that's coming out in the newest VM

jvms is really interesting and it both

to get us access to those new features

where they make sense for us and also

for performance so that's that work is

never going to stop it's just going to

be a continuing ongoing thing we've done

some some air op work this year which is

not also we had not quite gotten it to

the release point but working on it for

112 time frame

um we'll continue working on tooling

um done a lot this year continue pushing

that pushing that up the hill

um and I think uh working on the new

closure developer experience that really

that's a lot of really website and and

tooling and a lot of different stuff

it's it's just uh like I said it's just

connecting the dots that already exist a

lot of a lot of places uh and uh

thinking about that warm and then uh

I'll include spec we have not sort of

actively working on it the second but we

have been talking about it lately and

and uh hopefully that will get some

Cycles in the new year which I'm sure

lots of people are happy to know

um would like to get the spec 2 work at

least to a point where it's released

even if it's maybe not you know the end

point we would like like to at least get

that to a point where it's out there so

uh so I'm gonna leave you with this uh

move faster Break Stuff but we don't

break things right so we're not going to

do that we're gonna move fast we're

gonna make stuff so I'm making that the

uh and I'm sure Bradford would sign up

for this

um let's keep moving fast and making

stuff and I want to give a big thanks to

the re-closure team uh they've done a

great job and I really appreciate

everything they did putting this

together

and if you're interested in more closure

stuff closure conj is coming April 27th

28th it will be in person in Durham it

will also be streamed online

um so you should be able to get access

to that from anywhere and we would love

to see some of you there and Jordan

mentioned strangely the last stranger

will be September 21st and 22nd next

year in St Louis so

that's all I have