Equals initializer syntax

[oowekyala]

• Introduce the new = initializer syntax to replace the parens syntax in all targets except C++. This works exactly like parentheses do, except in C++ it calls the assignment operator.
• Using the parens syntax emits a warning now
• The formatter converts to the new syntax automatically for ease of migration (for C++ it doesn't touch anything).

Updated 2023-03-02 This does in fact add a syntax for C array initializers (which can be used also for struct initializers). Summary:

• Using a: int(0) or a: int(0,0) produces a warning (except in C++). they can be replaced with a: int = 0 resp. a: int = {0, 0} (the latter only in C/C++).
  • the braced list syntax is not available in Python or TS, so if you were declaring a python list/tuple with a(1,2), you can convert it to a = {= [1, 2] =} to remove the warning, or just wait until we allow writing a = [1, 2] directly.
• In all targets, reactor instantiations have to use the equal syntax: write new R(a = 1) instead of new R(a (1)).
  • In C++, if you want to call the constructor with parameters, then you have to write out the type of the constructed object explicitly: new R(a = {= std::vector<int>(1,2) =} or so (whereas you don't need to repeat the type when you write a default parameter value or a state var initializer). This is so because we would get problems with generic type parameters otherwise (and C++ itself has the same behavior).
  • In C, if you were writing new R(a(1,2)) or new R(a = (1,2)) to mean an array, then use new R(a = {1, 2})
  • In Python and TS, the same caveat applies, that you can't write a python/TS bracket-delimited list without target escaping yet.

Remaining action items:

• make build pass
• update doc and website pages see
  • Update doc for lf-lang/lingua-franca#1580 lf-lang.github.io#130
  • Format files for lf-lang/lingua-franca#1580 playground-lingua-franca#24
• @cmnrd write a test for C++
• Fix issues in test/Cpp/src/target/InitializerSyntax.lf where {} needs to be escaped to not be omitted.

Fixes #1650 and #1651

[lhstrh]

Only 654 files changed? 🤣

[lhstrh]

This looks good to me, though it would be great to get some more 👀 on this. My main concern is that this does not only affect our tests, but also our examples. We should probably set up CI in https://github.com/lf-lang/examples-lingua-franca first, and prepare a PR that makes all the required adjustments...

[lhstrh]

Just to clarify, does using braces now trigger a warning or an error?

[edwardalee]

This also requires an update of all the examples and text on the website.

[oowekyala]

Just to clarify, does using braces now trigger a warning or an error?

Using parentheses triggers a warning (except in C++), braces is still an error in all targets except C++. Also in all targets including C++, writing new R(a(1)) or a {1} instead of a = 1 or eg a = {= std::vector<int>(1) =} is now a warning.

This also requires an update of all the examples and text on the website.

I've been working on this and will submit a PR today.

[cmnrd]

Looks like all the tests pass now. This should be ready to be merged. Should we go ahead or does anyone else want to review the code?

[petervdonovan]

I do not object to merging this right away even though I have not reviewed the code. If other people think that more review is important and can wait for a few days for me to do it, I can, but I also do not mind trusting you and Clement so that this big PR does not linger any longer.

[petervdonovan]

The one failing test, NativeListsAndTimes, relies on a C23 feature, so it shouldn't be surprising that it does not pass on Windows. (The failing CLI tests are just a network issue.)

In C, I am not aware of any use cases for zero-length arrays except in the case when you are code-generating C, using macros, etc., and the array of length zero is the continuation of some pattern. I am not sure I can envision such situations cropping up in LF programs. For sure it is awkward and weird that C does not permit initializers of length zero before 2023, but maybe that is C's fault, not ours. We have the ability to patch it up in LF, but I am not sure that that is our responsibility.

[petervdonovan]

LGTM, I have no important comments. Looking forward to seeing this get merged soon!