{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ECUG（Effective Cloud User Group，實效雲計算用戶組）今日在上海舉辦，會上，七牛雲 CEO 許式偉以“數據科學與Go+”爲主題發表了主題分享，講述了對數據科學變遷的理解，對新語言 Go+ 的設想和規劃，並大膽指出數據科學正迎來爆發期，像字節跳動一樣的新型公司只會越來越多。以下爲演講內容整理，Enjoy~"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"剛纔在閒聊說ECUG變得越來越高大上，其實我也變得越來越像一個單純的講師。ECUG到現在也非常多年了，剛纔說辦了13屆，今年剛好是第14個年頭。其實這一屆本來應該在去年辦，但因爲疫情延後了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"其實，我在ECUG一直貫徹的理念有兩個："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第一，讓自己持續地寫代碼。因爲每一次來ECUG我都很緊張，不能什麼都沒有呀。所以這也是挺好的機會，能讓自己持續留在技術一線；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"另外，其實我每年分享的主題都有一定的延續性，會有一個關於我自己對未來思考的脈絡。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"去年開始，我在聊數據科學，前面有三年是聊在端上的一些實踐。原因是，我認爲雲計算的第一個時代應該是屬於機器計算，也就是虛擬機；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二代就是雲原生，我認爲這是一場被稱爲“基礎架構”的革命。也就是說，第一階段是資源，第二階段是基礎架構；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三個階段，我之前判斷爲應用計算，這會涉及前端和後端的協同。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"從去年開始，我的分享轉向了數據科學，爲什麼呢？一個很重要的因素和趨勢是，數據時代的到來。尤其是2017年之後，數據大量地被數字化以後，在各行各業，其實都會有涉及數據科學的廣泛應用。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"去年也是蠻巧的，我腦子一熱就搞了一個語言出來。我以前搞過蠻多語言，受衆也蠻多的。但是那個很明確，從來沒有想過有一天能夠商業化。也許碰巧有一些公司用它來做商業化，但是基本上從出生那一刻開始，就不是衝着商業化去的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"2012年我花了很多精力在佈道Go，因爲當時作爲一個初創公司 CEO，招人太難。好的招人邏輯就是讓別人覺得你有趣，公司業務清晰。Go+是我第一個認認真真希望能夠把它商業化的語言，但目前宣傳得還不多，1.0還沒發佈。我想講講我自己對Go+和數據科學的一個思考，爲什麼認爲Go+有商業化的機會。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我今天聊的話題大概有四個方面："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"語言的發展；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據科學的發展；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Go+的設計理念；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Go+現實的迭代。"}]}]}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"語言的發展"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"首先，我們講講語言的發展，程序員對這個話題非常感興趣。我把語言的發展史分爲三個部分："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第一，靜態語言的發展史。我選的是TOP20的語言，這個是根據現在最火的語言排行榜排名選的，前20名的語言我排了一下大概是這樣的，最早發佈的是C，到現在其實還在排行榜前三的位置。第二是C++，Objective-C、Java、C#、Go，Swift、Go+。我們可以看到一個比較有趣的現象，差不多每6-8年會出現一輪新的、具備影響力的靜態語言，這是生產力迭代的表徵。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二，腳本語言發展的TOP20。你會發現它們非常不一樣。最早是Visual Basic，然後是Python、PHP、JavaScript、Ruby，腳本語言是集中大爆發的，差不多全在Java出現的前後，來自90年代的創辦者。這是非常有趣的一件事情，也是非常值得思考的，背後一定有一些內在的原因。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三，數據科學相關的語言發展史，但數據科學我選的是TOP50，因爲TOP20實在太少了。也蠻有意思的，最早的是SQL，第二個SAS，MATLAB、Python、R、Julia，Python最早從來沒想過自己是數據科學語言，但最終變成了這樣。Go+也是拿來湊數的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這裏又存在一個很明顯的特徵：它的跨度跟靜態語言一樣大，所以數據科學發展其實是古老而漫長的，發展得沒有那麼快。靜態語言差不多每6-8年有一個迭代，但數據科學語言不是，中間跨度特別大。我覺得現在正進入數據科學的加速期。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"你可能會想，爲什麼我要分析語言發展史呢？有幾個結論是關鍵。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"首先，我認爲腳本語言是特定歷史階段下的產物，長期來看，靜態語言更有生命力。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二，數據科學是計算機的最初需求，最早計算機就是用來做計算的。它歷史悠久但進步緩慢，因爲數據大爆發的時代一直沒有到來。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三，數據科學的發展。數據科學其實分爲幾個階段，第一個階段我叫做“原始時期”，也可以叫“數學軟件時代”，這些軟件基本上可歸納爲兩個特徵，第一個是在有限領域裏，最典型的是BI（商業智能）；第二個有限數據規模，典型就像Excel，行列數都是非常有限的，其他的軟件也基本上是這樣的。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據科學的發展"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這個時期的數據科學特點是什麼？首先它不是一個基礎設施，實際上是數學應用軟件，但能力非常全，很強大，包括了統計、預測、洞察、規劃、決策等等。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二個時期我叫做“數據科學的基建時期”，真正成爲了數據科學的基礎設施，最典型的是大數據的興起。Map\/Reduce是Google 2004年發佈的一篇論文，2006年就出現了Hadoop，2009年出現了Spark。我認爲這算是大數據興起的一個階段，也是數據科學基礎設施化的開始。這個時期跟剛纔的數學軟件不一樣，是以大規模處理能力爲先，並不是以功能強大爲先，它的功能相對侷限。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"深度學習的興起和大數據的興起間隔時間比較長，深度學習2015年開始有TensorFlow，2017年開始有Torch，這是兩個知名度最高的深度學習框架，深度學習本質就是y=F（x），程序員實現這個F，但最核心的概念是如何讓機器自動產生這個F，來達成最佳曲線。它其實是基於測量結果的自動計算。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"假設今天沒有牛頓三大定理，但我有一堆測量數據，理論上應該能夠發現牛頓三大定理，這就是深度學習的核心邏輯。它跟大數據並不是相互取代的關係，而是一種能力的加強，更多其實是如何讓大數據的能力更進一步，更強悍。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"有種看法認爲，今天經濟發展背後的驅動因子其實只有兩個，一個是計算，另外一個是數據。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據核心就是我們今天聊的數據科學，數據科學其實是到了一個新的範式，有一個詞叫“第四範式”，中國有一個公司也叫第四範式，我們認爲數據是基因的一種生產能力，其實跟計算是站在兩個或是更高層次的維度。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"前面是數據科學的兩個階段，到第三個階段。我覺得是數據科學的大爆發時期，也就是今天，用馬雲的話說是“DT時代”，原始時期是在有限的領域，有限的數據規模下去做的一種能力。未來首先是全領域的，首先領域不再侷限於所謂的BI範疇，第二個是大規模的數據，第三個是隨處可見，隨處可見包括雲、智能手機、嵌入式設備等，這些都會植入我們所謂的智能。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這就意味着，今天移動互聯網的興起已經讓很多公司非常牛，互聯網的平民化或互聯網應用的誕生，催生了BAT，但是我們知道，現在新興的、比較牛的公司，像字節跳動這種，其實不是互聯網的成功，而是數據科學的成功。今天仍然不能說，數據科學是平民化的，它的門檻非常高。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但是我們看到，智能應用已經產生了，智能應用不會只侷限於抖音這樣一個局部領域的生產力放大，各行各業都會被數據智能，也就是剛剛我們提到的第四範式所影響。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據和數據科學，一定會成爲下一代生產力的支撐，今天產生了字節跳動、快手這樣的新興的公司，但他們只是一個開始，絕對不是結尾。數據只是副產品，大家想象一下，BI數據只是一個副產品，只是用於後期的運營決策。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但是今天我們看到在大量的應用裏，數據就是原材料。這是非常不一樣的狀態，這也是爲什麼，我把它叫做數據科學大爆發時期，這是我覺得今天爲什麼需要 Go+ 的原因，也是其背後的歷史背景。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據科學的未來一定是通用語言和數學軟件的融合，從而完成真正意義上的，數據科學的基礎設施化。但在今天，數據科學的基礎設施化還沒有完全完成，這是我自己的判斷。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"今天的Python已經很好了，爲何需要Go+？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"當然很多人會有疑問：今天的Python已經很好了，在深度學習領域已經被非常廣泛地使用，爲什麼Python還不夠，需要Go+？其實我是認爲，Python是成不了基礎設施的。首先它是一個腳本語言，我認爲僅僅是特定歷史階段的需要。數據科學本身是一種算力革命，哪怕在芯片領域，數據也能幹翻計算，上層應用領域更加如此，一定有一個新的基礎設施承載者需要出現。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"算力本質上是一種計算密集型業務，Python的背後是C，只靠Python還是不行。今天是C和Python支撐了整個深度學習，但數據科學一定還有進一步下沉，下沉的結果是什麼？所以我覺得今天需要Go+！前面主要講我自己爲什麼認爲Go+有商業化的機會，當然我所說的商業化不一定是賺錢，大家不要誤會這一點，語言可能在大多數人心目中是一個不賺錢的東西，但是不代表不重要，它非常重要。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三，Go+的設計理念。Go+爲什麼是今天這個樣子？計算背後要的是程序員，而數據科學背後要的是數據科學家或者叫分析師。這兩個角色其實還是不一樣的，雖然都是技術工作。我認爲培養程序員是相對容易的，今天程序員的數量是非常龐大的，但數據科學家的數量相對較少，這也是爲什麼前幾年深度學習興起以後，所謂的AI工程師薪資被炒翻了，比程序員貴很多。其實就是因爲數據科學家不容易找。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這個角色承載着技術和商業的連接，要找到同時具備兩種能力的人是很難的。數據科學首先是一個技術工作，要的是技術能力，又要懂商業。今天仍然沒有非常體系化的培養數據科學家的能力，沒有這樣一個體系方法論。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"那麼Go+的核心理念又是什麼呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第一個，我們試圖用Go+來統一程序員和數據科學家，他們之間要有共同話語，讓雙方能自然對話，我覺得這是Go+最核心的一個思考點。Go+很重要的一個核心邏輯，是用一門語言去讓兩個角色能對話。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"在這個基礎上，我們延伸了一些設計邏輯。首先，Go+是一個靜態語言，語法是完全兼容Go的；第二，形式上要比Go更像腳本，有更低的學習門檻。Go雖然在靜態語言裏，可能學習門檻是低的，但還不夠低，沒有Python那麼低；第三，很自然的，因爲我們要做一個數據科學的語言，所以它必然要有更簡潔的、數學運算上的語言支持；第四是雙引擎，同時支持靜態編譯爲可執行文件，也支持編譯成字節碼來解釋執行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"爲什麼我們會選擇語法完全兼容Go呢？首先我個人很堅定的認爲，靜態語言擁有更強的生命力，能更加穿越歷史的高梯。大家也都很容易理解，語言其實是需要翻譯週期的，語言的生命週期非常長，我們不能很侷限的說，當前是什麼東西，我就如何決定語言的設計，實際上我們要找到跨週期的元素。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二，爲什麼是Go？我個人認爲，在靜態語言裏，Go的語法設計是最爲精簡、學習門檻也是最低的，哪怕你以前沒有學過靜態語言，也很容易學會Go。我們公司最早是用Go招聘，但大部分招進來的人都不會Go。我們用Go的時候，世界上真沒多少人認爲Go是未來的流行語言。我們自己實踐Go，兩週的學習基本上夠了，是門檻非常低的一門靜態語言。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但從數據科學語言來講，Go的門檻還不夠低，Go+完全兼容Go，我們希望它比Go的門檻還要有更低。所以它形式上要比Go更像腳本，因爲腳本往往更容易理解，學習門檻和Python處於同一個層次。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"去年8月份Go+剛誕生，差不多10月份左右，我就開始讓13-14歲，六年級到初一這個階段的三個小孩學習Go+。至少我自己實踐證明，這個事情是可行的。他們能理解Go+的整個設計，並且能夠自如的使用Go+寫代碼。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這證明了我們在Go的基礎上做的所有簡化其實是划算的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我這裏簡單列了一些Go+的語法，當然不是全部，只是一些我認爲還是相對比較簡潔的表達。有理數Python裏面沒有，我們認爲有理數在數據科學裏面還是非常成功的，在無損運算裏還是會非常常見。我們其實有內置的有理數的技術，MAP、SLICE基本上Python都有。列表理解其實也是Python有的，但我們對列表支持還是非常的完整，基本上理解了For循環怎麼寫也就理解了列表，更多的還是數據科學的一些常規操作的簡潔表達。以上是一個大概語法，如果有朋友沒看過Go+，希望可以大概對Go+有個理解。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Go+是非常有意思的，唯一一個選擇了雙引擎的語言，既支持靜態編譯，也支持可解析執行。爲什麼要做雙引擎呢？因爲我認爲程序員和數據科學家的訴求是不一樣的，數據科學家喜歡單步執行，大家可以在心目中想象一下數學軟件，包括SAS、MATLAB，數學軟件都是單步執行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這並不是因爲它們爛，程序員理解程序邏輯是可以放在腦子裏的，我們腦子裏知道寫的對不對。但數據科學家做計算的時候，不能知道計算結果對不對，因爲人的計算能力比計算機弱太多了，所以一定要單步執行看到計算結果，才能知道自己下一步應該怎麼辦，這是數據科學家和程序員工作模式完全不同的一個點。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"因爲他是在做計算而不是在做一種程序邏輯，所以他很難不去做單步執行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但當數據科學家建了一種模型，最終要使用了，這時他仍然希望最終交付的是最大化的執行效率，他一定不希望代碼很爛。所以這個時候他就又需要靜態編譯執行，這也是爲什麼Go+希望設計成雙引擎，因爲調試階段和生產使用階段，工作模式完全不一樣了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第四，Go+的迭代，當前做到什麼份上？Go+雖然還沒有推出1.0版本，但是語法目前支持百分之六七十肯定有了，語法完成度還是不錯的。Go+的源代碼，通過掃描器轉成一個Go+的Token，再通過一個parser變成Go+的抽象的語法數，常見語言都是這麼幹的。Go+的抽象語法數有兩個分支，一個生成Go的代碼從而使其可以靜態編譯，另外一個分成字節碼解析執行，中間的分支是通過引入了一個執行規範，其實是一個抽象的接口。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"當前，我個人在迭代的過程中發現了一個問題，對另外一個人來說，是需要一段時間熟悉整個業務的。Go+執行規範的部分，其實是一種抽象的SAX接口，基於事件驅動，我有一個事件發送給接受方，接受方按自己的需要理解這個事件，這在文本處理裏面比較常見。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我們之前設計的接口基本上是用事件驅動的一種模式來把不同組件完成，這是一個事件的接口，編譯器把抽象語法數解析完發生一些事件，這些事件被兩個模塊分別接收，按照自己的需求去幹活，這個模式代碼還是有點難理解，尤其是編譯器裏面又做一些比較複雜的事情，大家如果瞭解過Go背後的實現邏輯，類型推導在Go裏面比較複雜，其實我們編譯器的複雜性也是由類型推導導致的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我其實是試圖重構這個邏輯，想把執行規範部分變得不再是一個抽象的接口，而是一個標準實現的DOM，這個DOM本身包含了類型推導的能力，從而使得編譯器相對比較簡單，這是目前正在做的一個重構。剛纔講實現我沒法講的特別細，我講一下我們Go+下一步做的重心是什麼。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"首先，最核心的邏輯，還是希望今年能夠發佈1.0版本，而1.0版本最重要的事情是把用戶的範式做最大化的確認，1.0以後我希望和Go差不多，後面的語法變更是比較少的。當前最重要的工作，是明確Go+需要哪些最核心的語法，並且在1.0版本就儘量去支持，除非有一些特定的考量比如說像Go，放棄一些特別複雜的語法特性。Go+也是類似的，我們可能會放棄一些特別複雜的語法特性，但是基本上儘可能把大部分我們需要的語法特性在1.0版本里確定下來。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"1.0我們會先進行單引擎的迭代，先做好靜態編譯的引擎，等1.0發佈以後再迭代腳本的引擎。歡迎大家加入Go+團隊，我們希望用商業化的方式來去運作，也會招聘Go+的團隊成員。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我認爲Go+的核心是首先統一了程序員和數據科學家的語言，讓雙方能夠自然對話。另外我非常堅定的相信Go+會是數據科學的下一個變革，我自己非常興奮能夠做這樣一件事情，也非常歡迎認可這件事的人加入我們。這是聯繫我們的方法，第一個是項目的地址（"},{"type":"link","attrs":{"href":"https:\/\/github.com\/goplus","title":null,"type":null},"content":[{"type":"text","text":"https:\/\/github.com\/goplus"}],"marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}]},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"），第二個投簡歷的郵箱（[email protected]），第三個是我推特的地址（@xushiwei）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]}]}
許式偉：相比Python，我們可能更需要Go+
{"type":"doc","content":[{"type":"blockquote","content":[{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","text":"ECUG（Effective Cloud User Group，實效雲計算用戶組）今日在上海舉辦，會上，七牛雲 CEO 許式偉以“數據科學與Go+”爲主題發表了主題分享，講述了對數據科學變遷的理解，對新語言 Go+ 的設想和規劃，並大膽指出數據科學正迎來爆發期，像字節跳動一樣的新型公司只會越來越多。以下爲演講內容整理，Enjoy~"}]}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"剛纔在閒聊說ECUG變得越來越高大上，其實我也變得越來越像一個單純的講師。ECUG到現在也非常多年了，剛纔說辦了13屆，今年剛好是第14個年頭。其實這一屆本來應該在去年辦，但因爲疫情延後了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"其實，我在ECUG一直貫徹的理念有兩個："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第一，讓自己持續地寫代碼。因爲每一次來ECUG我都很緊張，不能什麼都沒有呀。所以這也是挺好的機會，能讓自己持續留在技術一線；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"另外，其實我每年分享的主題都有一定的延續性，會有一個關於我自己對未來思考的脈絡。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"去年開始，我在聊數據科學，前面有三年是聊在端上的一些實踐。原因是，我認爲雲計算的第一個時代應該是屬於機器計算，也就是虛擬機；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二代就是雲原生，我認爲這是一場被稱爲“基礎架構”的革命。也就是說，第一階段是資源，第二階段是基礎架構；"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三個階段，我之前判斷爲應用計算，這會涉及前端和後端的協同。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"從去年開始，我的分享轉向了數據科學，爲什麼呢？一個很重要的因素和趨勢是，數據時代的到來。尤其是2017年之後，數據大量地被數字化以後，在各行各業，其實都會有涉及數據科學的廣泛應用。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"去年也是蠻巧的，我腦子一熱就搞了一個語言出來。我以前搞過蠻多語言，受衆也蠻多的。但是那個很明確，從來沒有想過有一天能夠商業化。也許碰巧有一些公司用它來做商業化，但是基本上從出生那一刻開始，就不是衝着商業化去的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"2012年我花了很多精力在佈道Go，因爲當時作爲一個初創公司 CEO，招人太難。好的招人邏輯就是讓別人覺得你有趣，公司業務清晰。Go+是我第一個認認真真希望能夠把它商業化的語言，但目前宣傳得還不多，1.0還沒發佈。我想講講我自己對Go+和數據科學的一個思考，爲什麼認爲Go+有商業化的機會。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我今天聊的話題大概有四個方面："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"numberedlist","attrs":{"start":null,"normalizeStart":1},"content":[{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":1,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"語言的發展；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":2,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據科學的發展；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":3,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Go+的設計理念；"}]}]},{"type":"listitem","attrs":{"listStyle":null},"content":[{"type":"paragraph","attrs":{"indent":0,"number":4,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Go+現實的迭代。"}]}]}]},{"type":"heading","attrs":{"align":null,"level":3},"content":[{"type":"text","text":"語言的發展"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null}},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"首先，我們講講語言的發展，程序員對這個話題非常感興趣。我把語言的發展史分爲三個部分："}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第一，靜態語言的發展史。我選的是TOP20的語言，這個是根據現在最火的語言排行榜排名選的，前20名的語言我排了一下大概是這樣的，最早發佈的是C，到現在其實還在排行榜前三的位置。第二是C++，Objective-C、Java、C#、Go，Swift、Go+。我們可以看到一個比較有趣的現象，差不多每6-8年會出現一輪新的、具備影響力的靜態語言，這是生產力迭代的表徵。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二，腳本語言發展的TOP20。你會發現它們非常不一樣。最早是Visual Basic，然後是Python、PHP、JavaScript、Ruby，腳本語言是集中大爆發的，差不多全在Java出現的前後，來自90年代的創辦者。這是非常有趣的一件事情，也是非常值得思考的，背後一定有一些內在的原因。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三，數據科學相關的語言發展史，但數據科學我選的是TOP50，因爲TOP20實在太少了。也蠻有意思的，最早的是SQL，第二個SAS，MATLAB、Python、R、Julia，Python最早從來沒想過自己是數據科學語言，但最終變成了這樣。Go+也是拿來湊數的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這裏又存在一個很明顯的特徵：它的跨度跟靜態語言一樣大，所以數據科學發展其實是古老而漫長的，發展得沒有那麼快。靜態語言差不多每6-8年有一個迭代，但數據科學語言不是，中間跨度特別大。我覺得現在正進入數據科學的加速期。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"你可能會想，爲什麼我要分析語言發展史呢？有幾個結論是關鍵。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"首先，我認爲腳本語言是特定歷史階段下的產物，長期來看，靜態語言更有生命力。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二，數據科學是計算機的最初需求，最早計算機就是用來做計算的。它歷史悠久但進步緩慢，因爲數據大爆發的時代一直沒有到來。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三，數據科學的發展。數據科學其實分爲幾個階段，第一個階段我叫做“原始時期”，也可以叫“數學軟件時代”，這些軟件基本上可歸納爲兩個特徵，第一個是在有限領域裏，最典型的是BI（商業智能）；第二個有限數據規模，典型就像Excel，行列數都是非常有限的，其他的軟件也基本上是這樣的。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據科學的發展"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這個時期的數據科學特點是什麼？首先它不是一個基礎設施，實際上是數學應用軟件，但能力非常全，很強大，包括了統計、預測、洞察、規劃、決策等等。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二個時期我叫做“數據科學的基建時期”，真正成爲了數據科學的基礎設施，最典型的是大數據的興起。Map\/Reduce是Google 2004年發佈的一篇論文，2006年就出現了Hadoop，2009年出現了Spark。我認爲這算是大數據興起的一個階段，也是數據科學基礎設施化的開始。這個時期跟剛纔的數學軟件不一樣，是以大規模處理能力爲先，並不是以功能強大爲先，它的功能相對侷限。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"深度學習的興起和大數據的興起間隔時間比較長，深度學習2015年開始有TensorFlow，2017年開始有Torch，這是兩個知名度最高的深度學習框架，深度學習本質就是y=F（x），程序員實現這個F，但最核心的概念是如何讓機器自動產生這個F，來達成最佳曲線。它其實是基於測量結果的自動計算。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"假設今天沒有牛頓三大定理，但我有一堆測量數據，理論上應該能夠發現牛頓三大定理，這就是深度學習的核心邏輯。它跟大數據並不是相互取代的關係，而是一種能力的加強，更多其實是如何讓大數據的能力更進一步，更強悍。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"有種看法認爲，今天經濟發展背後的驅動因子其實只有兩個，一個是計算，另外一個是數據。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據核心就是我們今天聊的數據科學，數據科學其實是到了一個新的範式，有一個詞叫“第四範式”，中國有一個公司也叫第四範式，我們認爲數據是基因的一種生產能力，其實跟計算是站在兩個或是更高層次的維度。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"前面是數據科學的兩個階段，到第三個階段。我覺得是數據科學的大爆發時期，也就是今天，用馬雲的話說是“DT時代”，原始時期是在有限的領域，有限的數據規模下去做的一種能力。未來首先是全領域的，首先領域不再侷限於所謂的BI範疇，第二個是大規模的數據，第三個是隨處可見，隨處可見包括雲、智能手機、嵌入式設備等，這些都會植入我們所謂的智能。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這就意味着，今天移動互聯網的興起已經讓很多公司非常牛，互聯網的平民化或互聯網應用的誕生，催生了BAT，但是我們知道，現在新興的、比較牛的公司，像字節跳動這種，其實不是互聯網的成功，而是數據科學的成功。今天仍然不能說，數據科學是平民化的，它的門檻非常高。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但是我們看到，智能應用已經產生了，智能應用不會只侷限於抖音這樣一個局部領域的生產力放大，各行各業都會被數據智能，也就是剛剛我們提到的第四範式所影響。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據和數據科學，一定會成爲下一代生產力的支撐，今天產生了字節跳動、快手這樣的新興的公司，但他們只是一個開始，絕對不是結尾。數據只是副產品，大家想象一下，BI數據只是一個副產品，只是用於後期的運營決策。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但是今天我們看到在大量的應用裏，數據就是原材料。這是非常不一樣的狀態，這也是爲什麼，我把它叫做數據科學大爆發時期，這是我覺得今天爲什麼需要 Go+ 的原因，也是其背後的歷史背景。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"數據科學的未來一定是通用語言和數學軟件的融合，從而完成真正意義上的，數據科學的基礎設施化。但在今天，數據科學的基礎設施化還沒有完全完成，這是我自己的判斷。"}]},{"type":"heading","attrs":{"align":null,"level":2},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"今天的Python已經很好了，爲何需要Go+？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"當然很多人會有疑問：今天的Python已經很好了，在深度學習領域已經被非常廣泛地使用，爲什麼Python還不夠，需要Go+？其實我是認爲，Python是成不了基礎設施的。首先它是一個腳本語言，我認爲僅僅是特定歷史階段的需要。數據科學本身是一種算力革命，哪怕在芯片領域，數據也能幹翻計算，上層應用領域更加如此，一定有一個新的基礎設施承載者需要出現。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"算力本質上是一種計算密集型業務，Python的背後是C，只靠Python還是不行。今天是C和Python支撐了整個深度學習，但數據科學一定還有進一步下沉，下沉的結果是什麼？所以我覺得今天需要Go+！前面主要講我自己爲什麼認爲Go+有商業化的機會，當然我所說的商業化不一定是賺錢，大家不要誤會這一點，語言可能在大多數人心目中是一個不賺錢的東西，但是不代表不重要，它非常重要。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第三，Go+的設計理念。Go+爲什麼是今天這個樣子？計算背後要的是程序員，而數據科學背後要的是數據科學家或者叫分析師。這兩個角色其實還是不一樣的，雖然都是技術工作。我認爲培養程序員是相對容易的，今天程序員的數量是非常龐大的，但數據科學家的數量相對較少，這也是爲什麼前幾年深度學習興起以後，所謂的AI工程師薪資被炒翻了，比程序員貴很多。其實就是因爲數據科學家不容易找。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這個角色承載着技術和商業的連接，要找到同時具備兩種能力的人是很難的。數據科學首先是一個技術工作，要的是技術能力，又要懂商業。今天仍然沒有非常體系化的培養數據科學家的能力，沒有這樣一個體系方法論。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"那麼Go+的核心理念又是什麼呢？"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第一個，我們試圖用Go+來統一程序員和數據科學家，他們之間要有共同話語，讓雙方能自然對話，我覺得這是Go+最核心的一個思考點。Go+很重要的一個核心邏輯，是用一門語言去讓兩個角色能對話。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"在這個基礎上，我們延伸了一些設計邏輯。首先，Go+是一個靜態語言，語法是完全兼容Go的；第二，形式上要比Go更像腳本，有更低的學習門檻。Go雖然在靜態語言裏，可能學習門檻是低的，但還不夠低，沒有Python那麼低；第三，很自然的，因爲我們要做一個數據科學的語言，所以它必然要有更簡潔的、數學運算上的語言支持；第四是雙引擎，同時支持靜態編譯爲可執行文件，也支持編譯成字節碼來解釋執行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"爲什麼我們會選擇語法完全兼容Go呢？首先我個人很堅定的認爲，靜態語言擁有更強的生命力，能更加穿越歷史的高梯。大家也都很容易理解，語言其實是需要翻譯週期的，語言的生命週期非常長，我們不能很侷限的說，當前是什麼東西，我就如何決定語言的設計，實際上我們要找到跨週期的元素。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第二，爲什麼是Go？我個人認爲，在靜態語言裏，Go的語法設計是最爲精簡、學習門檻也是最低的，哪怕你以前沒有學過靜態語言，也很容易學會Go。我們公司最早是用Go招聘，但大部分招進來的人都不會Go。我們用Go的時候，世界上真沒多少人認爲Go是未來的流行語言。我們自己實踐Go，兩週的學習基本上夠了，是門檻非常低的一門靜態語言。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但從數據科學語言來講，Go的門檻還不夠低，Go+完全兼容Go，我們希望它比Go的門檻還要有更低。所以它形式上要比Go更像腳本，因爲腳本往往更容易理解，學習門檻和Python處於同一個層次。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"去年8月份Go+剛誕生，差不多10月份左右，我就開始讓13-14歲，六年級到初一這個階段的三個小孩學習Go+。至少我自己實踐證明，這個事情是可行的。他們能理解Go+的整個設計，並且能夠自如的使用Go+寫代碼。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這證明了我們在Go的基礎上做的所有簡化其實是划算的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我這裏簡單列了一些Go+的語法，當然不是全部，只是一些我認爲還是相對比較簡潔的表達。有理數Python裏面沒有，我們認爲有理數在數據科學裏面還是非常成功的，在無損運算裏還是會非常常見。我們其實有內置的有理數的技術，MAP、SLICE基本上Python都有。列表理解其實也是Python有的，但我們對列表支持還是非常的完整，基本上理解了For循環怎麼寫也就理解了列表，更多的還是數據科學的一些常規操作的簡潔表達。以上是一個大概語法，如果有朋友沒看過Go+，希望可以大概對Go+有個理解。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"Go+是非常有意思的，唯一一個選擇了雙引擎的語言，既支持靜態編譯，也支持可解析執行。爲什麼要做雙引擎呢？因爲我認爲程序員和數據科學家的訴求是不一樣的，數據科學家喜歡單步執行，大家可以在心目中想象一下數學軟件，包括SAS、MATLAB，數學軟件都是單步執行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"這並不是因爲它們爛，程序員理解程序邏輯是可以放在腦子裏的，我們腦子裏知道寫的對不對。但數據科學家做計算的時候，不能知道計算結果對不對，因爲人的計算能力比計算機弱太多了，所以一定要單步執行看到計算結果，才能知道自己下一步應該怎麼辦，這是數據科學家和程序員工作模式完全不同的一個點。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"因爲他是在做計算而不是在做一種程序邏輯，所以他很難不去做單步執行。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"但當數據科學家建了一種模型，最終要使用了，這時他仍然希望最終交付的是最大化的執行效率，他一定不希望代碼很爛。所以這個時候他就又需要靜態編譯執行，這也是爲什麼Go+希望設計成雙引擎，因爲調試階段和生產使用階段，工作模式完全不一樣了。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"第四，Go+的迭代，當前做到什麼份上？Go+雖然還沒有推出1.0版本，但是語法目前支持百分之六七十肯定有了，語法完成度還是不錯的。Go+的源代碼，通過掃描器轉成一個Go+的Token，再通過一個parser變成Go+的抽象的語法數，常見語言都是這麼幹的。Go+的抽象語法數有兩個分支，一個生成Go的代碼從而使其可以靜態編譯，另外一個分成字節碼解析執行，中間的分支是通過引入了一個執行規範，其實是一個抽象的接口。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"當前，我個人在迭代的過程中發現了一個問題，對另外一個人來說，是需要一段時間熟悉整個業務的。Go+執行規範的部分，其實是一種抽象的SAX接口，基於事件驅動，我有一個事件發送給接受方，接受方按自己的需要理解這個事件，這在文本處理裏面比較常見。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我們之前設計的接口基本上是用事件驅動的一種模式來把不同組件完成，這是一個事件的接口，編譯器把抽象語法數解析完發生一些事件，這些事件被兩個模塊分別接收，按照自己的需求去幹活，這個模式代碼還是有點難理解，尤其是編譯器裏面又做一些比較複雜的事情，大家如果瞭解過Go背後的實現邏輯，類型推導在Go裏面比較複雜，其實我們編譯器的複雜性也是由類型推導導致的。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我其實是試圖重構這個邏輯，想把執行規範部分變得不再是一個抽象的接口，而是一個標準實現的DOM，這個DOM本身包含了類型推導的能力，從而使得編譯器相對比較簡單，這是目前正在做的一個重構。剛纔講實現我沒法講的特別細，我講一下我們Go+下一步做的重心是什麼。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"首先，最核心的邏輯，還是希望今年能夠發佈1.0版本，而1.0版本最重要的事情是把用戶的範式做最大化的確認，1.0以後我希望和Go差不多，後面的語法變更是比較少的。當前最重要的工作，是明確Go+需要哪些最核心的語法，並且在1.0版本就儘量去支持，除非有一些特定的考量比如說像Go，放棄一些特別複雜的語法特性。Go+也是類似的，我們可能會放棄一些特別複雜的語法特性，但是基本上儘可能把大部分我們需要的語法特性在1.0版本里確定下來。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"1.0我們會先進行單引擎的迭代，先做好靜態編譯的引擎，等1.0發佈以後再迭代腳本的引擎。歡迎大家加入Go+團隊，我們希望用商業化的方式來去運作，也會招聘Go+的團隊成員。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"我認爲Go+的核心是首先統一了程序員和數據科學家的語言，讓雙方能夠自然對話。另外我非常堅定的相信Go+會是數據科學的下一個變革，我自己非常興奮能夠做這樣一件事情，也非常歡迎認可這件事的人加入我們。這是聯繫我們的方法，第一個是項目的地址（"},{"type":"link","attrs":{"href":"https:\/\/github.com\/goplus","title":null,"type":null},"content":[{"type":"text","text":"https:\/\/github.com\/goplus"}],"marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}]},{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":"），第二個投簡歷的郵箱（[email protected]），第三個是我推特的地址（@xushiwei）。"}]},{"type":"paragraph","attrs":{"indent":0,"number":0,"align":null,"origin":null},"content":[{"type":"text","marks":[{"type":"color","attrs":{"color":"#494949","name":"user"}}],"text":" "}]}]}
發表評論
所有評論
還沒有人評論，想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章