Make assignment (match) in comprehension work as strict generator

[richcarl]

Implements EEP 77 (erlang/eep#74)

It would be useful to be able to easily bind variables in the qualifiers of a comprehension, for example:

[Z || X <- Ls,
      {foo, Y} = g(X),
      Z <- f(Y, h(Y))].

You can do this today by writing a singleton generator:

[Z || X <- Ls,
      {foo, Y} <- [g(X)],  % produce single element
      Z <- f(Y, h(Y))].

but this has some drawbacks: the intent is not clear to the reader; you have to remember to write a single element list around the right hand side argument (which itself could be a list); you probably want it to be a strict generator so typos don't just silently yield no elements; and someone might edit the code later and accidentally add extra elements to the right hand side, causing unintended Cartesian combinations.

It is in fact already allowed syntactically to have a = match expression in the qualifiers, but this is just interpreted as any other expression - thus expected to produce a boolean value, and if false, the current element will be skipped. Hence, any qualifier to the right of a V = Expr match will only execute if V is true. We can therefore expect that no such uses exist in practice. (The OTP code base has been checked and does not contain any.) To be completely safe, matches in a comprehension qualifier position can first be made an error, and then allowed with these binding semantics in the following major release.

This is related but orthogonal to #9134. It would make those few cases where variables have been exported from subexpressions of qualifiers much nicer to rewrite than requiring them to be singleton generators.

[github-actions]

CT Test Results

     6 files     571 suites   1h 59m 16s ⏱️
 4 449 tests  4 320 ✅ 129 💤 0 ❌
10 629 runs  10 473 ✅ 156 💤 0 ❌

Results for commit 1c1c525.

♻️ This comment has been updated with latest results.

To speed up review, make sure that you have read Contributing to Erlang/OTP and that all checks pass.

See the TESTING and DEVELOPMENT HowTo guides for details about how to run test locally.

Artifacts

• Complete CT logs (Download Logs)
• HTML Documentation (Download HTML Docs)
• Windows Installer

// Erlang/OTP Github Action Bot

[richcarl]

Correction: 1 occurence in the OTP test suites, in singleton_inference.erl, which is weird enough that I won't even try to rewrite it myself. Maybe @jhogberg who wrote it (commit ebc461e) can help?

[richcarl]

Turned into a feature. Becomes an error if it occurs when the feature is not enabled.

Also fixed multiple problems with the features implementation.

[richcarl]

Rebased on #10376 to make the Features fix a separate thing.

[richcarl]

Rebased again and updated feature introduced version to 29.