Erlang Abstract Format并不难懂,只是枯燥一点罢了,如果把Abstract Format的文档翻译出来,其实就是Erlang教科书中语法入门的部分. Erlang Abstract Format实际上是用Erlang代码的AST,下面通过一些真切的实例代码了解一下它的一些细节.
首先,Erlang Abstract Format里面包含一些概念,我会在下面的描述中把涉及到的概念字体加粗.请注意概念之间的层次关系.Erlang代码本身使用非常扁平的module组织,每一个module是由一系列forms组成的.这些forms分成两大类:attributes和函数声明 见下图.
Attribute
attributes相对比较简单,我们先从一个只有attributes没有任何函数声明的module开始
-module(k). -compile(export_all). -compile({parse_transform,print_form}). -export([test/0]). -url("http://cnblogs.com/me-sa/"). -record(student,{class,id}). -record(player,{id=0,name=[],level}).
上面代码对应的Abstract Format如下:
[{attribute,1,file,{"k.erl",1}},
{attribute,1,module,k},
{attribute,2,compile,export_all},
{attribute,4,export,[{test,0}]},
{attribute,5,url,"http://cnblogs.com/me-sa/"},
{attribute,6,record,
{student,[{record_field,6,{atom,6,class}},
{record_field,6,{atom,6,id}}]}},
{attribute,7,record,
{player,[{record_field,7,{atom,7,id},{integer,7,0}},
{record_field,7,{atom,7,name},{nil,7}},
{record_field,7,{atom,7,level}}]}},
{eof,10}]
我们可以一一对照上面思维导图,可以看到上面每一行代码包括record定义在内在Abstract Format层面看都是attribute.里面不断出现的数字是代码所在行数,这个信息是非常重要的,在编译时提示代码出错行,运行时报错包含代码行数都要用到它.Erlang最新版报错的时候已经包含了出错的代码所在行,而之前这个功能是没有的,工作在蛮荒纪的erlanger,搞了一个smart_exceptions的项目来实现这个功能: https://github.com/thomasl/smart_exceptions/tree/master/stable 言归正传,上面需要细说的form是record的定义,在有record字段初始值的情况,结构会稍复杂一点:{record_field,LINE,Name,Value}.比如player的id=0,对应的Form是{record_field,7,{atom,7,id},{integer,7,0}},id的初始值0是一个字面常量,记作{integer,7,0},在Erlang Abstract Format文档里面字面常量被称为Atomic literals,包含四种:atom integer string float,见下图
Function declarations
说过了attribute,下面我们看下函数声明部分,这里有一个概念Patterns,它特指function或fun的参数列表.我们做几个简单的方法出来:
-module(a). -compile(export_all). -export([test/0]). -record(student,{class,id}). -record(player,{id=0,name=[],level}). test()-> "hello world!". test(a,[1,2]) -> "a:[1,2]"; test(12.5,100)-> "test". test([]) -> empty; test(abc) -> "atom test". foo(a)-> {b,100}. bar({1,2},12)-> [1,2,3,4,5,6]. k(Num) when Num >1000 -> bigger_than_100; k(Num) -> whatever. call(1000)-> k(1000); call(1002)-> erlang:now().
它对应的Abstract Format是:
[{attribute,1,file,{"a.erl",1}},
{attribute,1,module,a},
{attribute,2,compile,export_all},
{attribute,4,export,[{test,0}]},
{attribute,6,record,
{student,
[{record_field,6,{atom,6,class}},{record_field,6,{atom,6,id}}]}},
{attribute,7,record,
{player,
[{record_field,7,{atom,7,id},{integer,7,0}},
{record_field,7,{atom,7,name},{nil,7}},
{record_field,7,{atom,7,level}}]}},
{function,10,test,0,[{clause,10,[],[],[{string,11,"hello world!"}]}]},
{function,13,test,2,
[{clause,13,
[{atom,13,a},
{cons,13,{integer,13,1},{cons,13,{integer,13,2},{nil,13}}}],
[],
[{string,14,"a:[1,2]"}]},
{clause,15,
[{float,15,12.5},{integer,15,100}],
[],
[{string,16,"test"}]}]},
{function,18,test,1,
[{clause,18,[{nil,18}],[],[{atom,19,empty}]},
{clause,20,[{atom,20,abc}],[],[{string,21,"atom test"}]}]},
{function,24,foo,1,
[{clause,24,
[{atom,24,a}],
[],
[{tuple,25,[{atom,25,b},{integer,25,100}]}]}]},
{function,27,bar,2,
[{clause,27,
[{tuple,27,[{integer,27,1},{integer,27,2}]},{integer,27,12}],
[],
[{cons,28,
{integer,28,1},
{cons,28,
{integer,28,2},
{cons,28,
{integer,28,3},
{cons,28,
{integer,28,4},
{cons,28,
{integer,28,5},
{cons,28,{integer,28,6},{nil,28}}}}}}}]}]},
{function,30,k,1,
[{clause,30,
[{var,30,'Num'}],
[[{op,30,'>',{var,30,'Num'},{integer,30,1000}}]],
[{atom,31,bigger_than_100}]},
{clause,32,[{var,32,'Num'}],[],[{atom,33,whatever}]}]},
{function,36,call,1,
[{clause,36,
[{integer,36,1000}],
[],
[{call,37,{atom,37,k},[{integer,37,1000}]}]},
{clause,38,
[{integer,38,1002}],
[],
[{call,39,{remote,39,{atom,39,erlang},{atom,39,now}},[]}]}]},
{eof,41}]
看上面的函数声明,function里面包含一个或者多个function clauses,比如:
{function,18,test,1,
[{clause,18,[{nil,18}],[],[{atom,19,empty}]},
{clause,20,[{atom,20,abc}],[],[{string,21,"atom test"}]}]},
这段代码就很好玩了,它在形式上已经非常贴近Clojure(好吧,Lisp行不行)里面函数定义的语法了:
Clojure 1.4.0
user=> (defn make_a_set
([x] #{x})
([x,y] #{x,y}))
#'user/make_a_set
user=> (make_a_set 12)
#{12}
user=> (make_a_set 12 23)
#{12 23}
user=>
再看一下函数调用,erlang:now().这一句的调用对应的是: [{call,39,{remote,39,{atom,39,erlang},{atom,39,now}},[]}]}]} 看到这里能想起来Erlang里面对Local Call和Remote Call的定义了吧. 上面的代码里面还可以看到Guard的表达形式:
[[{op,30,'>',{var,30,'Num'},{integer,30,1000}}]].
除了function clasuse之外,还有 if clauses, case clauses and catch clauses 就不再细说.深入函数体内部就是各种表达式, 方法体里面包含了一些表达式Expressions,Expressions和Patterns表示方式是一样的.比如A = lists:seq(1,10).这样一个表达式,它的Abstract Format其实很好的解释了这里等号其实是一个匹配运算.
7> E= fun(Code)-> {_,Tokens,_}=erl_scan:string(Code),rp(erl_parse:parse_exprs(Tokens)) end.
#Fun<erl_eval.6.80484245>
8> E(" A = lists:seq(1,10).").
{ok,[{match,1,
{var,1,'A'},
{call,1,
{remote,1,{atom,1,lists},{atom,1,seq}},
[{integer,1,1},{integer,1,10}]}}]}
ok
10>
除了上面这些"普通"的东西,当然还有列表解析和二进制数据处理相关的语法,下面通过两个简单的例子看下:
Eshell V5.10.2 (abort with ^G)
1> E= fun(Code)-> {_,Tokens,_}=erl_scan:string(Code),rp(erl_parse:parse_exprs(Tokens)) end.
#Fun<erl_eval.6.80484245>
2> E("[Item || Item<- [1,2,3,4],Item>2 ].").
{ok,[{lc,1,
{var,1,'Item'},
[{generate,1,
{var,1,'Item'},
{cons,1,
{integer,1,1},
{cons,1,
{integer,1,2},
{cons,1,{integer,1,3},{cons,1,{integer,1,4},{nil,1}}}}}},
{op,1,'>',{var,1,'Item'},{integer,1,2}}]}]}
ok
3>
7> E= fun(Code)-> {_,Tokens,_}=erl_scan:string(Code),rp(erl_parse:parse_exprs(Tokens)) end.
#Fun<erl_eval.6.80484245>
8> E("<<A:8,B/binary>> = <<1,2,3,4>>.").
{ok,[{match,1,
{bin,1,
[{bin_element,1,{var,1,'A'},{integer,1,8},default},
{bin_element,1,{var,1,'B'},default,[binary]}]},
{bin,1,
[{bin_element,1,{integer,1,1},default,default},
{bin_element,1,{integer,1,2},default,default},
{bin_element,1,{integer,1,3},default,default},
{bin_element,1,{integer,1,4},default,default}]}}]}
ok
对于Erlang数据结构中的王者List需要仔细观察下,它的表达形式是:
4> E("[a,b,c,d].").
{ok,[{cons,1,
{atom,1,a},
{cons,1,
{atom,1,b},
{cons,1,{atom,1,c},{cons,1,{atom,1,d},{nil,1}}}}}]}
ok
说到这里我们可以把概念之间的层次关系梳理出来了:
Run! Run!
上面啰嗦了那么多细节,那从Abstract Format如何到可执行的代码呢?下面我们就完成这个过程:
3> {ok, MTs, _} = erl_scan:string("-module(t)."). {ok,[{'-',1}, {atom,1,module}, {'(',1}, {atom,1,t}, {')',1}, {dot,1}], 1} 4> {ok, ETs, _} = erl_scan:string("-export([say/0])."). {ok,[{'-',1}, {atom,1,export}, {'(',1}, {'[',1}, {atom,1,say}, {'/',1}, {integer,1,0}, {']',1}, {')',1}, {dot,1}], 1} 5> {ok, FTs, _} = erl_scan:string("say() -> \"hello_world!!\"."). {ok,[{atom,1,say}, {'(',1}, {')',1}, {'->',1}, {string,1,"hello_world!!"}, {dot,1}], 1} 6> Forms= [begin {ok,R}=erl_parse:parse_form(Item),R end || Item<-[MTs,ETs,FTs]]. [{attribute,1,module,t}, {attribute,1,export,[{say,0}]}, {function,1,say,0, [{clause,1,[],[],[{string,1,"hello_world!!"}]}]}] 7> {ok, t, Bin} = compile:forms(Forms). {ok,t, <<70,79,82,49,0,0,1,224,66,69,65,77,65,116,111,109,0,0,0, 45,0,0,0,5,1,116,...>>} 8> code:load_binary(t,"nofile",Bin). {module,t} 9> t:say(). "hello_world!!" 10>
好吧,好多好玩的东西还没有说,暂时到这里,下回继续......
最后小图一张,经历了25年,13季,大侦探波洛的故事结束了,"女士们,先生们,该收场了........."
