Add gradualizer

[arcusfelis]

Added gradualizer, tweaked types a bit to pass the checks.
Mainly we have to constrain the result from a function before returning it (if we don't use the same return type)

It also adds more dialyzer checks (which produce similar warnings).

MIM-1981

[codecov]

Codecov Report

Patch coverage: 100.00% and project coverage change: +0.33% 🎉

Comparison is base (be80c30) 90.69% compared to head (b872029) 91.03%.

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #11      +/-   ##
==========================================
+ Coverage   90.69%   91.03%   +0.33%     
==========================================
  Files           8        8              
  Lines         430      435       +5     
==========================================
+ Hits          390      396       +6     
+ Misses         40       39       -1     

Files Changed	Coverage Δ
src/cets_ack.erl	100.00% <ø> (ø)
src/cets.erl	99.42% <100.00%> (+0.01%)	⬆️
src/cets_call.erl	84.61% <100.00%> (+4.61%)	⬆️
src/cets_discovery.erl	56.36% <100.00%> (-0.78%)	⬇️

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[erszcz]

Suggested change

	assert_integer(X) when is_integer(X) -> X.
	-spec assert_integer(any()) -> integer().
	assert_integer(X) when is_integer(X) -> X.

We have to add a spec to let Gradualizer now that after the guard succeeds the return value is actually an integer(). Otherwise, it would assume -spec assert_integer(any()) -> any() and therefore treat the return value as any() further on, which significantly relaxes any further checks against it.

[arcusfelis]

oh, interesting, dialyzer would still know that assert_integer returns an integer :)
But not gradualizer

Line 604 Column 2: Invalid type specification for function cets:handle_get_info/1. The success typing is (#{'ack_pid':=_, 'opts':=_, 'other_servers':=[pid()], 'tab':=atom() | ets:tid(), _=>_}) -> #{'ack_pid':=_, 'memory':=integer(), 'nodes':=[atom()], 'opts':=_, 'size':=integer(), 'table':=atom() | ets:tid()}

Will add

[arcusfelis]

ok, we don't really need to add assert, just spec is enough.

[erszcz]

If we don't want/need the guard, i.e. the dynamic check (which I think is better to have in general, unless we're 100% the result cannot be anything else), we can also do:

-include_lib("gradualizer/include/gradualizer.hrl").

...
size => ?assert_type(ets:info(Tab, size), integer())
...

[arcusfelis]

oh, actually runtime check

-spec assert_non_neg_integer(any()) -> non_neg_integer().
assert_non_neg_integer(X) when is_integer(X), X >= 0 -> X.

makes dialyzer unhappy

src/cets.erl
Line 620 Column 2: Type specification cets:assert_non_neg_integer(any()) -> non_neg_integer() is a supertype of the success typing: cets:assert_non_neg_integer(non_neg_integer()) -> non_neg_integer()

:)

[arcusfelis]

also, don't wanna add gradualizer as a build dep just for the header, to make compilation faster :D

[erszcz]

Yeah, we might create a gradualizer_support package at some point to address the compilation issue.

[erszcz]

Same story, if there's no spec then Gradualizer will assume -spec pid_or_port(_) -> any().