[oil-language] Implement command sub in Oil expressions.

One more spec test passes.  And we can even interoperate!

bin/oil.py

@@ -501,7 +501,7 @@ def ShellMain(lang, argv0, argv, login_shell):
   bool_ev = osh_expr_eval.BoolEvaluator(mem, exec_opts, word_ev, errfmt)
   exec_deps.bool_ev = bool_ev
 
-  expr_ev = expr_eval.OilEvaluator(mem, errfmt)
+  expr_ev = expr_eval.OilEvaluator(mem, ex, errfmt)
   exec_deps.expr_ev = expr_ev
 
   tracer = dev.Tracer(parse_ctx, exec_opts, mem, word_ev, trace_f)

oil_lang/expr_eval.py

@@ -6,7 +6,7 @@
 
 from _devbuild.gen.id_kind_asdl import Id
 from _devbuild.gen.syntax_asdl import (
-    expr_e, word_part_e
+    expr_e, expr_t, word_part_e
 )
 from _devbuild.gen.runtime_asdl import (
     lvalue, value_e, scope_e,
@@ -23,8 +23,9 @@ class OilEvaluator(object):
   2. Look up variables and evaluate words.
   """
 
-  def __init__(self, mem, errfmt):
+  def __init__(self, mem, ex, errfmt):
     self.mem = mem
+    self.ex = ex
     self.errfmt = errfmt
 
   def LookupVar(self, var_name):
@@ -71,6 +72,7 @@ def EvalWordPart(self, part):
     raise NotImplementedError(part.__class__.__name__)
 
   def EvalRHS(self, node):
+    # type: (expr_t) -> Any
     """
     This is a naive PyObject evaluator!  It uses the type dispatch of the host
     Python interpreter.
@@ -90,6 +92,9 @@ def EvalRHS(self, node):
     if node.tag == expr_e.Var:
       return self.LookupVar(node.name.val)
 
+    if node.tag == expr_e.CommandSub:
+      return self.ex.RunCommandSub(node.command_list)
+
     if node.tag == expr_e.DoubleQuoted:
       s = ''.join(self.EvalWordPart(part) for part in node.parts)
       return s

spec/oil-assign.test.sh

@@ -33,9 +33,11 @@ x=7
 
 #### command sub $(echo hi)
 var x = $(echo hi)
-echo x=$x
+var y = $(echo '')
+# Make sure we can operate on these values
+echo x=${x:-default} y=${y:-default}
 ## STDOUT:
-x=hi
+x=hi y=default
 ## END
 
 #### shell array @(a b c)