Allow blocks on LHS of pipe expression

[nkrkv]

Close #178

I’m tempted to generalize pipe expression as seen in #176 and #177. But to me, they are too wide, producing extraneous grammar conflicts for constructs not legal actually in ReScript.

So, as an alternative proposal, can we have a nailed-down addition to fix the bug, but not go too wide?

/cc @aspeddro

[aspeddro]

We can remove all rules on rhs and use only choice($.primary_expression, $.block)?

There will be minor changes in testing. But I think it's more consistent.

Example:

1 -> #A("foo")

Is parsed as:

yarn run v1.22.19
$ /home/pedro/Desktop/projects/tree-sitter-rescript/node_modules/.bin/tree-sitter parse ../../test-filetypes/rescript_print.res
(source_file [0, 0] - [1, 0]
  (expression_statement [0, 0] - [0, 14]
    (call_expression [0, 0] - [0, 14]
      function: (pipe_expression [0, 0] - [0, 7]
        (number [0, 0] - [0, 1])
        (polyvar_identifier [0, 5] - [0, 7]))
      arguments: (arguments [0, 7] - [0, 14]
        (string [0, 8] - [0, 13]
          (string_fragment [0, 9] - [0, 12]))))))
Done in 0.47s.

With this change:

yarn run v1.22.19
$ /home/pedro/Desktop/projects/tree-sitter-rescript/node_modules/.bin/tree-sitter parse ../../test-filetypes/rescript_print.res
(source_file [0, 0] - [1, 0]
  (expression_statement [0, 0] - [0, 14]
    (pipe_expression [0, 0] - [0, 14]
      (number [0, 0] - [0, 1])
      (polyvar [0, 5] - [0, 14]
        (polyvar_identifier [0, 5] - [0, 7])
        (arguments [0, 7] - [0, 14]
          (string [0, 8] - [0, 13]
            (string_fragment [0, 9] - [0, 12])))))))
Done in 0.05s.

[aspeddro]

Actually in another PR. See #182

[nkrkv]

Yep. Fair point, thank you for the PR!

[aspeddro]

So, as an alternative proposal, can we have a nailed-down addition to fix the bug, but not go too wide?

Sure. Looks great.