Support OTP 28

[paulo-ferraz-oliveira]

I add the minimum requirements to have elvis_core not throw (and parse) for upcoming rule change operator_spaces (around <:= and <:-, for lists, binaries and maps)

We could think about moving minimum_otp_vsn to 26, now that 28 is out.

I changed the scope of the PR (was what is now struck through; now is what the title states) following discussions around preferences. I'll move it to draft until I have time to work on more changes.

Fix #89.

The process

For future reference:

1. I copied a formatted version of epp_dodger over a formatted version of ktn_dodger and adjusted the latter accordingly (take care to not ruin the "format" in place - eases review?).
2. I made sure rebar3 test was ✅
3. I added some tests for newer syntax
4. I checked changes to absform.md (https://github.com/erlang/otp/blob/master/erts/doc/guides/absform.md) and the mentioned modules:

• epp
• erl_eval
• erl_lint
• erl_parse
• erl_pp
• absform.md ℹ️ this was possibly the most useful of the bunch

5. I checked changes to:

• erl_syntax
• erl_prettypr

6. I ran a test, mentioned below, on the whole of OTP's codebase to find exceptions

Interesting commits:

• 46f1495: strict generators
• 899322d: zip generators
• f2f48e3: moduledoc/doc

[paulo-ferraz-oliveira]

I don't have permissions to change

but feel free to do it, if you like the change.

[elbrujohalcon]

My goal with #89 was to make these changes but also compare OTP28's epp_dodger against OTP27's epp_dodger and adjust ktn_dodger accordingly… like we usually do with every OTP version update.
Then, based on that, adjust ktn_code, the tests, and everything else here.
I believe we can wait a little bit more and do the whole move to OTP28 properly instead of introducing just what we need for Elvis to handle OTP28 files.

[paulo-ferraz-oliveira]

My goal with #89 was to make these changes but also compare OTP28's epp_dodger against OTP27's epp_dodger and adjust ktn_dodger accordingly… like we usually do with every OTP version update. Then, based on that, adjust ktn_code, the tests, and everything else here. I believe we can wait a little bit more and do the whole move to OTP28 properly instead of introducing just what we need for Elvis to handle OTP28 files.

Sure, or we can incrementally improve this via the open issue. I'm not sure how big the PR will end up being, but I must say I prefer smaller scope changes. The same reasoning applies to the Elvis Core change: better to do it incrementally. In any case, I'll separate the other one in two so it's easier to reason on.

[elbrujohalcon]

My perspective is that I prefer a single PR, with all the changes, to be sure we don't miss anything.

[paulo-ferraz-oliveira]

My perspective is that I prefer a single PR, with all the changes, to be sure we don't miss anything.

I won't argue with this preference, as long as when the final PR comes you don't ask me to split it up 😆

[paulo-ferraz-oliveira]

@elbrujohalcon, shall we up the minimum to OTP 26 when this pull request is ready for review?

[paulo-ferraz-oliveira]

I tentatively assigned this to 2.4.0.

[elbrujohalcon]

My perspective is that I prefer a single PR, with all the changes, to be sure we don't miss anything.

I won't argue with this preference, as long as when the final PR comes you don't ask me to split it up 😆

Pinky promise.

[elbrujohalcon]

@elbrujohalcon, shall we up the minimum to OTP 26 when this pull request is ready for review?

YES!

[paulo-ferraz-oliveira]

@elbrujohalcon, I think I was pretty thorough in the code comparisons (and we seem to cover absform.md 100%)... I'm gonna write a local test to analyse all the .erl from the Erlang codebase and see if something surfaces.

Done: test code was

parse_erl_for(Dir) ->
    case file:list_dir(Dir) of
        {ok, Entries} ->
            lists:foreach(
                fun(Entry) ->
                    FullPath = filename:join(Dir, Entry),
                    case filelib:is_dir(FullPath) of
                        true ->
                            parse_erl_for(FullPath);
                        false ->
                            filename:extension(Entry) =:= ".erl" andalso
                                try
                                    {ok, Binary} = file:read_file(FullPath),
                                    ktn_code:parse_tree(Binary)
                                catch
                                    Class:Reason:S ->
                                        erlang:display({exc, Class, Reason, FullPath, S})
                                end
                    end
                end,
                Entries
            );
        {error, _Reason} ->
            error
    end.

3801 analysed, and the output is

expected {error,function_clause,"<path_to_otp>/lib/xmerl/doc/examples/sdocbook2xhtml.erl"}
handled  {throw,{unhandled_abstract_form,{error,{{6,2},ktn_dodger,{warning,"-warning ( \"a warning\" )."}}}},"<path_to_otp>/lib/tools/test/make_SUITE_data/test1.erl"}
expected {error,{badmatch,{error,{{38,12},erl_scan,{illegal,string}},{38,12}}},"<path_to_otp>/lib/wx/api_gen/wx_extra/wxTaskBarIcon.erl"}
expected {error,{badmatch,{error,{{72,12},erl_scan,{illegal,string}},{72,12}}},"<path_to_otp>/lib/wx/api_gen/wx_extra/wxPrintout.erl"}
expected {error,{badmatch,{error,{{32,12},erl_scan,{illegal,string}},{32,12}}},"<path_to_otp>/lib/wx/api_gen/wx_extra/wxListCtrl.erl"}
expected {error,function_clause,"<path_to_otp>/lib/edoc/test/edoc_SUITE_data/un1.erl"}
expected {error,function_clause,"<path_to_otp>/lib/edoc/test/edoc_SUITE_data/un3.erl"}
expected {error,function_clause,"<path_to_otp>/lib/ssh/examples/ssh_sample_cli.erl"}
handled  {throw,{unhandled_abstract_form,{error,{{32,2},ktn_dodger,{warning,"-warning ( \"Ignore me -- testing that the debugger can handle warnings\" )."}}}},"<path_to_otp>/lib/common_test/test/ct_auto_compile_SUITE.erl"}
handled  {throw,{unhandled_abstract_form,{error,{{34,2},ktn_dodger,{warning,"-warning ( \"Ignore me -- testing that the debugger can handle warnings\" )."}}}},"<path_to_otp>/lib/debugger/test/andor_SUITE.erl"}
expected {error,function_clause,"<path_to_otp>/lib/compiler/test/compile_SUITE_data/bad_enc.erl"}

I'm gonna check if any of this is fixable, but these seem like corner cases, basically.

Yeah, so this results in a further commit for a fix, but all the other cases are basically the UTF-8 -related oddities being tested.

[elbrujohalcon]

I know it sucks and I'm truly sorry to request this but… I would like changes to ktn_dodger to be just the new additions (i.e., no formatting, no function renaming, no atom enclosing, etc…).
The docs as comments is clever. I didn't do it last time because it didn't occur to me that I could just put the new -doc attributes in comments. Those ones should stay :)

[paulo-ferraz-oliveira]

I would like changes to ktn_dodger to be just the new additions (i.e., no formatting, no function renaming, no atom enclosing, etc…).

How so? If you do the procedure above it's much simpler to find the differences by copy-pasting over the file. And then we don't have to deal with all the function name changes, etc. What's your rationale, though?

No a big issue; I'll spend some time moving this back, but the list of changes will probably be smaller, in the end.

[paulo-ferraz-oliveira]

erlfmt, even if asked to not format something, decided to remove 2 consecutive blank lines. Hope that's Ok.

[paulo-ferraz-oliveira]

Feel free to merge and release; I think I addressed all your comments. I'll then import the new version next to elvis_core.