Skip to content

Latest commit

 

History

History
151 lines (121 loc) · 5.24 KB

README.md

File metadata and controls

151 lines (121 loc) · 5.24 KB
Raw

FancyFlow

FancyFlow is an experimental Erlang library to bring convenience of things like the elixir pipe operator into Erlang, without needing to introduce new syntax (although we do play with existing stuff to avoid semantic confusion). It also allows some more flexibility by allowing the user to choose the placement of the state being weaved through.

It's a toy, but I'm open to feedback.

Usage

Add the fancyflow_trans to your modules or applications, and use any of the control flow functions:

[pipe](InitialState, Exp1, Exp2, ..., ExpN)
[maybe](InitialState, Exp1, Exp2, ..., ExpN)
[parallel](Exp1, Exp2, ..., ExpN)

Where each expression can be a valid Erlang expression, with the state substituted by the _ variable.

For example:

-module(fancyflow_demo).
-export([sans_pipe/0, pipe/0,
         sans_maybe/0, maybe/0]).

-compile({parse_transform, fancyflow_trans}).

-spec sans_pipe() -> string().
sans_pipe() ->
    String = "a b c d e f",
    string:join(
      lists:map(fun string:to_upper/1, string:tokens(String, " ")),
      ","
    ).

-spec pipe() -> string().
pipe() ->
    [pipe]("a b c d e f",
           string:tokens(_, " "),
           lists:map(fun string:to_upper/1, _),
           string:join(_, ",")).

-spec sans_maybe() -> {ok, non_neg_integer()} | {error, term()}.
sans_maybe() ->
    case file:get_cwd() of
        {ok, Dir} ->
            case file:read_file(filename:join([Dir, "demo", "data.txt"])) of
                {ok, Bin} ->
                    {ok, {byte_size(Bin), Bin}};
                {error, Reason} ->
                    {error, Reason}
            end;
        {error, Reason} ->
            {error, Reason}
    end.

-spec maybe() -> {ok, non_neg_integer()} | {error, term()}.
maybe() ->
    [maybe](undefined,
            file:get_cwd(),
            file:read_file(filename:join([_, "demo", "data.txt"])),
            {ok, {byte_size(_), _}}).

-spec sans_parallel() -> [term() | {badrpc, term()}].
sans_parallel() ->
    R1 = rpc:async_call(node(), lists, seq, [1,10]),
    R2 = rpc:async_call(node(), filelib, wildcard, ["*"]),
    R3 = rpc:async_call(node(), erlang, apply,
                        [fun() -> timer:sleep(10), slept end, []]),
    R4 = rpc:async_call(node(), ets, all, []),
    [rpc:yield(R1), rpc:yield(R2), rpc:yield(R3), rpc:yield(R4)].

-spec parallel() -> [{ok, term()} | {error, term()}].
parallel() ->
    [parallel](lists:seq(1,10),
               filelib:wildcard("*"),
               begin timer:sleep(10), slept end,
               ets:all()).

The pipe() function reworks the sans_pipe() function to be equivalent. The maybe() one gives the same result as the sans_maybe() one, although it ignores the initial state altogether by using undefined. The sans_parallel() one shows traditional rpc-based handling of parallel operations whereas parallel() shows a simpler syntax, without need of work-arounds for unexported calls.

The expressions must be literal, and may be nested although this isn't always super clear:

-spec nested() -> [{ok, _} | {error, _}].
nested() ->
    [parallel](
        %% first operation reads ./demo/data.txt
        [maybe](undefined,
                file:get_cwd(),
                file:read_file(filename:join([_, "demo", "data.txt"]))),
        %% second parallel op makes a filename and reads its size if any
        [pipe]("a b c d e f",
               string:tokens(_, " "),
               lists:map(fun string:to_upper/1, _),
               string:join(_, ","),
               %% Maybe the file doesn't exist
               [maybe](_, % the string from [pipe] is a filename here
                       file:read_file(_),
                       {ok, {byte_size(_), _}})
              )
    ).

With a result set possibly looking like:

1> fancyflow_demo:nested().
[{ok,{ok,<<"124567890\n">>}},
 {ok,{error,enoent}}]

Showing what happens if the first file exists (contains 1234567890\n) and the second one does not (filename A,B,C,D,E,F), with both reads happening in parallel.

It might be a good idea not to nest the expressions too much.

How it works

Each form of

[pipe](InitialState, Exp1, Exp2, ..., ExpN)
[maybe](InitialState, Exp1, Exp2, ..., ExpN)
[parallel](Exp1, Exp2, ..., ExpN)

is translated to:

fancyflow:pipe(InitialState, [fun(Var) -> Exp1 end,
                              fun(Var) -> Exp2 end,
                              ...,
                              fun(Var) -> ExpN end])
fancyflow:maybe(InitialState, [fun(Var) -> Exp1 end,
                               fun(Var) -> Exp2 end,
                               ...,
                               fun(Var) -> ExpN end])
fancyflow:parallel([fun() -> Exp1 end,
                    fun() -> Exp2 end,
                    ...,
                    fun() -> ExpN end])

Which internally, runs a fold over the functions based on the state. The variable used as a function head is generated dynamically (looks like _#Ref<0.0.3.1555>) such that they never conflict with the surrounding scope, and never complain if they are not used.

the [Function](...) syntax has been chosen to not look like a regular function call and so that nobody mistakes its unorthodox (and otherwise illegal) usage of free variables with normal Erlang code.