erlang作为一个为电信级别而出现的语言，热更新是其最重要的特性之一
热代码升级-Erlang允许程序代码在运行系统中被修改。旧代码能被逐步淘汰而后被新代码替换。在此过渡期间，新旧代码是共存的。
下面我们以最典型的gen_server为例子，讲解一下这个BT的功能
1 -module(tt13). 2 -behaviour(gen_server). 3 4 -export([test/0]). 5 -export([start_link/0, stop/0, init/1, handle_call/3, handle_cast/2, handle_info/2, terminate/2, code_change/3]). 6 7 -record(state, {cnt}). 8 9 -define(SERVER, ?MODULE). 10 11 %%-------------------------------------------------------------------- 12 %% Function: start_link() -> {ok,Pid} | ignore | {error,Error} 13 %% Description: Starts the server 14 %%-------------------------------------------------------------------- 15 start_link() -> 16 gen_server:start_link({local, ?MODULE}, ?MODULE, [], [{debug, [trace]}]). 17 18 test() -> 19 gen_server:call(?SERVER, test). 20 21 stop() -> 22 gen_server:cast(?SERVER, stop). 23 24 %%-------------------------------------------------------------------- 25 %% Function: init(Args) -> {ok, State} | 26 %% {ok, State, Timeout} | 27 %% ignore | 28 %% {stop, Reason} 29 %% Description: Initiates the server 30 %%-------------------------------------------------------------------- 31 init(_) -> {ok, #state{cnt=1}}. 32 %%-------------------------------------------------------------------- 33 %% Function: %% handle_call(Request, From, State) -> {reply, Reply, State} | 34 %% {reply, Reply, State, Timeout} | 35 %% {noreply, State} | 36 %% {noreply, State, Timeout} | 37 %% {stop, Reason, Reply, State} | 38 %% {stop, Reason, State} 39 %% Description: Handling call messages 40 handle_call(test, _From, #state{cnt=Cnt} = State) -> 41 {reply, {ok, Cnt}, State#state{cnt=Cnt+1}}; 42 43 handle_call(stop, _From, State) -> 44 {stop, normal, ok, State}; 45 46 handle_call(_Unrec, _From, State) -> 47 {reply, {error, invalid_call}, State}. 48 49 %%-------------------------------------------------------------------- 50 %% Function: handle_cast(Msg, State) -> {noreply, State} | 51 %% {noreply, State, Timeout} | 52 %% {stop, Reason, State} 53 %% Description: Handling cast messages 54 %%-------------------------------------------------------------------- 55 handle_cast(_Msg, State) -> 56 {noreply, State}. 57 58 %%-------------------------------------------------------------------- 59 %% Function: handle_info(Info, State) -> {noreply, State} | 60 %% {noreply, State, Timeout} | 61 %% {stop, Reason, State} 62 %% Description: Handling all non call/cast messages 63 %%-------------------------------------------------------------------- 64 65 handle_info(_Info, State) -> 66 {noreply, State}. 67 68 %%-------------------------------------------------------------------- 69 %% Function: terminate(Reason, State) -> void() 70 %% Description: This function is called by a gen_server when it is about to 71 %% terminate. It should be the opposite of Module:init/1 and do any necessary 72 %% cleaning up. When it returns, the gen_server terminates with Reason. 73 %% The return value is ignored. 74 %%-------------------------------------------------------------------- 75 76 terminate(_Reason, _State) -> 77 io:format("hello gen server: terminating~n"). 78 79 %%-------------------------------------------------------------------- 80 %% Func: code_change(OldVsn, State, Extra) -> {ok, NewState} 81 %% Description: Convert process state when code is changed 82 %%-------------------------------------------------------------------- 83 84 code_change(_OldVsn, State, _Extra) -> 85 {ok, State}. 86 87 %%==================================================================== 88 %%other fun 89 %%====================================================================
编译运行结果
1 1> c(tt13). 2 {ok,tt13} 3 2> tt13:start_link(). 4 {ok,<0.39.0>} 5 3> tt13:test(). 6 *DBG* tt13 got call test from <0.32.0> 7 *DBG* tt13 sent {ok,1} to <0.32.0>, new state {state,2} 8 {ok,1} 9 4> tt13:test(). 10 *DBG* tt13 got call test from <0.32.0> 11 *DBG* tt13 sent {ok,2} to <0.32.0>, new state {state,3} 12 {ok,2} 13 5> tt13:test(). 14 *DBG* tt13 got call test from <0.32.0> 15 *DBG* tt13 sent {ok,3} to <0.32.0>, new state {state,4} 16 {ok,3}
如果修改了函数，可以直接运行
1 -module(tt13). 2 -version("1.1"). 3 -behaviour(gen_server). 4 5 %........... 6 %...........省略若干行 7 %................ 8 9 handle_call(test, _From, #state{cnt=Cnt} = State) -> 10 {reply, {ok, Cnt}, State#state{cnt=Cnt*2}}; 11 12 %........... 13 %...........省略若干行 14 %................
可以看到我们修改了计数的方法，而且修改了版本号，然后我们继续运行
1 6> c(tt13). %编译新的代码 2 {ok,tt13} 3 7> tt13:test(). 4 *DBG* tt13 got call test from <0.32.0> 5 *DBG* tt13 sent {ok,4} to <0.32.0>, new state {state,8} 6 {ok,4} 7 8> tt13:test(). 8 *DBG* tt13 got call test from <0.32.0> 9 *DBG* tt13 sent {ok,8} to <0.32.0>, new state {state,16} 10 {ok,8} 11 9> tt13:test(). 12 *DBG* tt13 got call test from <0.32.0> 13 *DBG* tt13 sent {ok,16} to <0.32.0>, new state {state,32} 14 {ok,16}
可以看到代码就直接替换了，注意编译的时候会用新的代码替换下一次运行的结果，正在运行还是old code，所以不要编译多次（一般在测试环境先进行热更新测试）。
如果要替换init/1里面的代码？这个方法肯定是不行的，因为init/1代码只运行一次,比如我要修改state结构体，那要怎么弄呢
1 -module(tt13). 2 -version("2.0"). 3 -behaviour(gen_server). 4 5 -record(state, {testcheck, cnt}). 6 %........... 7 %...........省略若干行 8 %................ 9 10 init(_) -> {ok, #state{testcheck='chk', cnt=1}}. 11 %%-------------------------------------------------------------------- 12 %% Function: %% handle_call(Request, From, State) -> {reply, Reply, State} | 13 %% {reply, Reply, State, Timeout} | 14 %% {noreply, State} | 15 %% {noreply, State, Timeout} | 16 %% {stop, Reason, Reply, State} | 17 %% {stop, Reason, State} 18 %% Description: Handling call messages 19 handle_call(test, _From, #state{cnt=Cnt} = State) -> 20 {reply, {ok, Cnt}, State#state{cnt=Cnt+1}}; 21 22 %........... 23 %...........省略若干行 24 %................ 25 26 code_change("1.1", {state, Cnt}, _Extra) -> 27 {ok, {state, chk, Cnt}}; 28 29 code_change(_OldVsn, State, _Extra) -> 30 {ok, State}. 31 32 %........... 33 %...........省略若干行 34 %................
我们修改了state结构体，修改了init/1函数，而且重写了code_change/3，下面我们运行如下
1 10> compile:file(tt13). /*编译代码 2 {ok,tt13} 3 11> sys:suspend(tt13). /*暂停服务 4 ok 5 12> code:purge(tt13). /*清除旧的代码，如果有运行的化 6 false 7 13> code:load_file(tt13). /*加载新的代码 8 {module,tt13} 9 14> sys:change_code(tt13,tt13,"1.1",[]). /*修改state状态 10 ok 11 15> sys:resume(tt13). /*恢复服务 12 ok 13 16> tt13:test(). 14 *DBG* tt13 got call test from <0.32.0> 15 *DBG* tt13 sent {ok,32} to <0.32.0>, new state {state,chk,64} 16 {ok,32} 17 17> tt13:test(). 18 *DBG* tt13 got call test from <0.32.0> 19 *DBG* tt13 sent {ok,64} to <0.32.0>, new state {state,chk,128} 20 {ok,64} 21 18> tt13:test(). 22 *DBG* tt13 got call test from <0.32.0> 23 *DBG* tt13 sent {ok,128} to <0.32.0>, new state {state,chk,256} 24 {ok,128}
整个替换过程是10-15步，注意这个过程中的tt13服务是hang住的，如果这时候使用服务，会出现timeout，所以一般这5步都是同时执行。
后面可以看到state状态已经改变