librelist archives

Re: [efene] Tests for constants

From:

Mariano Guerra

Date:

2011-11-25 @ 10:15

On Fri, Nov 25, 2011 at 1:44 AM, Volker Mische <volker.mische@gmail.com> wrote:
> Hi,
>
> as I patched the constants to allow all uppercase ones, I thought it's a
> good idea to write some tests. But I hit some problems. The AST looks
> different as it uses the line number of the original line where the
> constant was defined, not where it was used.
>
> Here's an example of what I tried:
>    tu:test_mod_ast("$const = 3\n\nbar = fn () {\n io.format(\"~p\",
> [$const])\n}\n",
>        "-module(temp).\n-export([]).\nbar() ->\nio:format(\"~p\",
> [3]).\n", "temp")
>
> The important difference in the result is:
> {cons,4,{integer,1,3},{nil,4}}
> vs.
> {cons,4,{integer,4,3},{nil,4}}
>
> I'm not sure what the correct solution for the problem is :)

the correct solution wasn't implemented :D

I just commited a fix for that


https://github.com/marianoguerra/efene/commit/2539233a3e8f9d228908ea48eaa3545913d080ca

when it replaces a constant it replaces the line in the constant
expression for the line where it's being inserted

there are still complex expressions that aren't handled (functions,
list comprehensions, if/try/case etc) they are easy to add but I'm
abusing work time :P

regarding tests, we should move to a more automatized testing, until
now I run the tests and grep for "error:" I think we can do one of two
things:

 * create a simple tests lib for efene
 * move the tests to erlang and use eunit

eunit can't be used in efene because it uses macros
moving to eunit would be easy but I don't know what it says about the
language if we don't test it in itself :D

what do you think?

example code:

$number = 1
$string = "asd"
$tuple = (1, true, atom, "string", [], [1, 1.2])

@public
run = fn ()
    io.format("~p, ~p ~p~n", [
        $number,
        $string,
        $tuple])

mariano@ganesha:~$ fnc test.ifn
Compiling test.ifn

mariano@ganesha:~$ fnc -r test run
1, "asd" {1,true,atom,"string",[],[1,1.2]}

mariano@ganesha:~$ fnc -t mod test.ifn
[{attribute,1,module,test},
 {attribute,2,export,[{run,0}]},
 {function,7,run,0,
     [{clause,7,[],[],
          [{call,8,
               {remote,8,{atom,8,io},{atom,8,format}},
               [{string,8,"~p, ~p ~p~n"},
                {cons,8,
                    {integer,9,1},
                    {cons,10,
                        {string,10,"asd"},
                        {cons,11,
                            {tuple,11,
                                [{integer,11,1},
                                 {atom,11,true},
                                 {atom,11,atom},
                                 {string,11,"string"},
                                 {nil,11},
                                 {cons,11,
                                     {integer,11,1},
                                     {cons,11,{float,11,1.2},{nil,11}}}]},
                            {nil,11}}}}]}]}]}]

mariano@ganesha:~$ fnc -t erl test.ifn
-module(test).

-export([run/0]).

run() ->
    io:format("~p, ~p ~p~n",
              [1, "asd",
               {1, true, atom, "string", [],
                [1, 1.19999999999999995559]}]).

On 11/25/2011 11:15 AM, Mariano Guerra wrote:
> On Fri, Nov 25, 2011 at 1:44 AM, Volker Mische <volker.mische@gmail.com> wrote:
>> Hi,
>>
>> as I patched the constants to allow all uppercase ones, I thought it's a
>> good idea to write some tests. But I hit some problems. The AST looks
>> different as it uses the line number of the original line where the
>> constant was defined, not where it was used.
>>
>> Here's an example of what I tried:
>>    tu:test_mod_ast("$const = 3\n\nbar = fn () {\n io.format(\"~p\",
>> [$const])\n}\n",
>>        "-module(temp).\n-export([]).\nbar() ->\nio:format(\"~p\",
>> [3]).\n", "temp")
>>
>> The important difference in the result is:
>> {cons,4,{integer,1,3},{nil,4}}
>> vs.
>> {cons,4,{integer,4,3},{nil,4}}
>>
>> I'm not sure what the correct solution for the problem is :)
> 
> the correct solution wasn't implemented :D
> 
> I just commited a fix for that
> 
> 
https://github.com/marianoguerra/efene/commit/2539233a3e8f9d228908ea48eaa3545913d080ca
> 
> when it replaces a constant it replaces the line in the constant
> expression for the line where it's being inserted
> 
> there are still complex expressions that aren't handled (functions,
> list comprehensions, if/try/case etc) they are easy to add but I'm
> abusing work time :P

Thanks for fixing it.

> regarding tests, we should move to a more automatized testing, until
> now I run the tests and grep for "error:" I think we can do one of two
> things:
> 
>  * create a simple tests lib for efene
>  * move the tests to erlang and use eunit
> 
> eunit can't be used in efene because it uses macros
> moving to eunit would be easy but I don't know what it says about the
> language if we don't test it in itself :D
> 
> what do you think?

Wouldn't it make sense to have macro support in Efene? I think one goal
should be full interoperability with Erlang. Other than that, have you
heard of Etap before [1]? It's used by CouchDB and GeoCouch.

I'm not sure how much work a testing framework in Efene would be. I
don't want to waste time, but on the other hand it would be to find bugs
in the language itself.

[1] https://github.com/ngerakines/etap

Cheers,
  Volker